Ryzen vs i7 (Mainstream); Threadripper vs i9 (HEDT); X299 vs X399/TRX40; Xeon vs Epyc

This is the way with life. All laptop are smaller less powerful than the desktop offerings at the time. So if you are waiting for the laptops to offer the power and performance of current desktop offerings it is a never ending cycle of wishing and waiting.

 

Putting a smack down on Intel;

 

8% IPC with 200mhz, i'd say its more 5% for most application and 100mhz at best for zen3.

combine the L3 for 8 cores could be good for games and video/audio/photo editing but thats about it? this should increase latency for L3 yet again though

 

I doubt Zen 3 will be the major uptick that Zen 2 was. That is ok for now, just having forward progression is what is important.

 

While decreasing latency of hopping to the I/O die for any comms to the other CCX on die and providing a larger pool to reduce memory calls with the larger latency hit.

What's you basis for thinking 100MHz and 5 percent?

AMD was rumored and achieved the 13-15% they were rumored to hit on the IPC for Zen 2. Yet you make a claim on what? That is going by SPEC testing. Anand covered it.

If true, AMD Zen 3 could have 23-25% IPC over Zen, and more than that in floating point. And Zen4 is a larger rework.

But back up that belief...

 

i need a basis for making speculation? gotta stop your attacking attitude.

you really think 200mhz is possible? that 200mhz is either pushing beyond the limit or within efficiency clock range, which is to say below 4ghz on an ES sample. i have 100% NO DOUBT what so ever we'll barely see a 100mhz when overclocked on the high end. I would love to be proven wrong because I am waiting for the "5ghz" cpu that has yet to come out from AMD.

it's hard to believe IPC will go up 8% just by combine the L3 there must be more optimization made otherwise its just bs.

 

OK, first, a basis for speculation makes it a theory. Not having a basis is you talking out of your rear. Now I don't stop you from talking out of your rear, but I WILL point it out.

Second, 4 layers of the stack will be done with EUV, hence a process refinement and change. That change allows for an 18% or so area reduction. Considering they are considering further increasing the L3, they would be confident the added latency is less impactful than any increase, if increased at all.

Third, there is a 10% estimated power reduction on 7nm+. That means, using isopower/isoperformance graph, that the frequency will go up. You may have a point on it only amounting to 100MHz, but there is another possibility where you are wrong.

If the majority of the die shrink is left unused, but they spread out what they can within the same package size, although not much, it can allow for higher boosts.

Or they could use the space to go wider on the cores/pipeline, which would help to get that IPC gain.

The truth is, there is enough information out there on process change and changes on design to say that it is possible, if not to the level of probable.

Please, cite what underpins your belief put into form in that speculation from your rear.

 

From what I can see, @ole!!! doesn't need to respond to your navel-gazing logic.

From your own admission, he may be just as right as you at this point.

Time will tell.

 

again with your forced opinion.
remember the fiasco of glofo from zen to zen+?

https://www.anandtech.com/show/1490...nology-massive-performance-power-improvements
https://www.anandtech.com/show/1422...echnology-7-nm-with-higher-transistor-density

boy would you look at that, going from 14 to 12nm nets you 10% performance increase with a 15% density. 7nm to 7nm+ gives ~same performance but improves efficiency by 10% and ~17% density.

i mean 14nm to 12nm gets us a ~100mhz from zen to zen+, its a no brainer this 7nm to 7nm+ at best gives another 100mhz. of course TSMC might have better node when compared to glofo and end up with more mhz but i'll go with what we have here.

also remember adoretv's 5ghz? he used the number and to calculate where he got his 5ghz from and thats his downfall. the number being advertised by TSMC/ glofo/ samsung are all within "efficient range" which is to say probably around 1.5 to 2.2ghz ish.

once you go over 4ghz all those numbers such as 10% performance probably become like 1-2% so unless they 100% specify 10% performance improvement at 4.5ghz, i'll take it that its just another 50-100mhz.

i mean there are definitely more to be desired and my speculation in the end is just that.

given that we are enthusiasts after high ST performance who wouldnt want a 5ghz AMD CPU with another 10% IPC boost over current zen2. im talking in favor of AMD but he still manage to find something to be upset about.

 

And as you noted, tsmc is not GF. In fact the origen of the 5GHz numbers were using GF speed estimates. Also, remember my prediction, which was right overall on performance uplift, but wrong on frequency and IPC (I went high on frequency a little, low on IPC a little, but got damn loser to the money overall on performance uplift).

And this is fanboy crap, which you insinuated in your last paragraph. This is you literally spouting off without a basis. When you explained yourself further, I addressed it.

Also, do you remember the difference between TSMC 12nm and tSMC 16nm? It was pretty sizeable. Yet they are both 16nm iterations. So why are you using GFs nose shrink to see what TSMC's will be like? You could use TSMC's history, even looking at Maxwell to Pascal, to get a potential idea. But you chose the other.

Now, here is the other reason your example doesn't apply: AMD only did a node shrink with no other substantial mods to the CPU, kept the footprint the same, and got 3% IPC and like 100MHz. So with more substantial changes to the die plus the shrink, you think no way to hit 6-8%. That boggles my mind, especially getting a 52% IPC with Zen, even though bulldozer sucked. 3% with Zen+. Around 15% on Zen 2, but then are saying no way to 6-8% on Zen 3.

It's because of things like this I make you explain yourself. It shows where flaws in your rationality may exist. I explain myself for that same reason, so that my underlying facts supporting my claims can be challenged.

You actually dig deeper when pushed. tiller reverts to childish BS when pushed and, at times, semantic games to try to push his ideas. In that way, you get and have more respect than him.

Edit: forgot to mention you are talking two different libraries for the frequency pushed. The denser library, found in ARM chips, have lower frequencies, although for ARM server TSMC has shown 4GHz chips on the 7nm process, which is part of the speed difference. The HPC libraries are less dense. They also regularly achieve higher speeds. Now which do most of their customers use? Hmmm.

 

TSMC is not glofo, however the technological stuff behind them are similar, they are literally competition. is that the best you can come up with to justify yourself when you "call" me out? it wont matter even if TSMC is better than glofo, which i already stated in my statement that they are different. ultimately it'll run into the same problem once you overclock past a certain point, those numbers dont mean jack anymore.

again what you seem to fail at understanding is that i would LOVE to be proved wrong. i want AMD at 5.5ghz if i can, all the more better for me when i buy TR 24cores.

 

The problem with us overclockers is if we can get 5.5 GHz stable we then want 6.0 GHz. Enough is never enough!

 

Now toy moved the goal post. First it was discussing 8% IPC and 200MHz, now it's some crap about 5.5GHz. Are you kidding me. Even Intel couldn't hit 4GHz with their increase in IPC of supposedly 18%, all while ignoring the black mark cannon lake to make the claim.

And yes, considering I would be called out using GF's process to explain Intel going from 14nm to 14nm+ (skylake to Kaby) or 14nm+ to 14nm++ (kaby to coffee), you cannot use the argument of GF to talk about TSMC. It's the same thing. I could argue like you did that they use similar processes. But the truth is the execution by the companies is different and those differences matter enough to make your statement false.

So is that the best you can come up with? Slight of hand for a comparative argument justification followed by arguing the absurd with an injection of hyperbole?

 

not kidding you, if you can tell when im trying to make a point you may as well not reply at all.

gonna sum this up and not argue with you no more. 100mhz at best to 7nm+, if we can get more thats good, doubt it.

with the way cpu is right now 12-16 cores only at 4.2 - 4.3ghz at best all core OC with cancerous voltage like 1.4v, even 200mhz is nothing amazing. give that 4.6ghz @1.25v im still waiting.

 

Even Intel on high core counts chokes on about 4.4 GHz water cooled. Considering the higher IPC of zen 2 4.3 GHz is awesome. You want better than 5.0 GHz setup your LN2 and go ahead.

 

And I'm not kidding you, make hyperbolic comments in argumentation, expect to get called out on the absurdity, as it doesn't help your point at all.

Now, 10% power draw reduction is around 1.26V, assuming equal amps. So, the question is whether you can achieve 200MHz in the extra 140mV. It is possible, but looking at overclocks of Zen 2, it took around that much to go to 4.4 from 4.3 all core on a cherry chip, with that last 100MHz being where efficiency drops off a cliff. Had you argued this, without mentioning anything on IPC which you still haven't supported and have dropped from subsequent argumentation, then you'd have a point. This it's why I specifically agreed on the frequency assessment. There is datum to point to. It is this other braggadocio I'm challenging, as well as your use of deflection and trying to diminish or dismiss unfavorable facts instead of acknowledgment and addressing them. That behavior makes me mad, with its prevalence in politics, the news, and tech today.

 

I think what might be overlooked is with CCX at 8 cores on ZEN3 maybe we could see, albeit at lower clocks, a 32 core AM4 or even a 128 core Epyc?

 

I don't think we'll get that with Zen 3. Die size too large, power constraints, etc.

But you now don't have just the best core or two out of 3 to 4 per CCX, you have the top core or two out of 8 cores. Also, it was pointed out at Tom's hardware that the wrong cores are being chosen for the top boost. Because of this, by changing the top core binning process and Microsoft coming with a better scheduler to select the best boost core to keep the CPU fed, we could see higher single core speeds.

For Epyc, they are staying at 64-cores for Milan. It maybe higher on custom and semi-custom chips.

But, eventually, they are working toward using an active interposer. Think of the I/O due integrated into the interposer, which was tested to have optimal throughput with an 8-core die, while no need for the artificial bifurcation anymore of a CCX. With that as a base, then attacking more chiplets, ASICs, or HBM makes sense. But that is my hopes off of known information, not hard evidence.

 

you are entitled to your opinion and that 200mhz isnt happening here. the average uplift of 7nm+ vs 7nm is a measly 100mhz and even then i think thats too much given on the track record we saw from glofo's node. once again, i'll mention here since you seem to not read. TSMC MIGHT HAVE BETTER NODE THAN GLOFO SO WE MIGHT END UP MORE THAN 100MHZ.

there you happy?

 

Please stop the bickering over opinions, there is no silicon to test yet.

 

The "Amazing" new Intel 9900KS has been released, and it is much closer to the Ryzen 3900x... in price.

http://forum.notebookreview.com/thr...-lake-cpus-z390.811225/page-152#post-10962216

 

Interesting bit of vendor news, the AMD 5700/5700XT have been noticeable slowing sales of Nvidia GPU's in that price range:

" RTX 2070 Coming Back
Our board partner contacts have told us that they’ve ordered more RTX 2070 GPU dies, and that NVIDIA is reviving the previously-retired 2070 line because AMD RX 5700 XT sales are outpacing 2060 Super sales. NVIDIA had originally told us (on record) that the RTX 2070 would be replaced with the 2060 Super and 2070 Super, but the company looks to be prolonging the life of its 2070 to better compete with AMD."
https://www.gamersnexus.net/news-pc/3528-hw-news-14nm-shortage-worsens-nvidia-revives-rtx-2070

HW News - NVIDIA Brings Back RTX 2070 to Fight AMD, Worse 14nm Shortage
Gamers Nexus
Published on Nov 1, 2019
Hardware news this week notes NVIDIA's resurrection of the 2070 (to fight vs. 5700 XT)

IMHO, Nvidia should have brought back the GTX1080ti in the $400 - $450 range, anything *without* that RTX baggage.

 

Here's the AMD Epyc's 105 World Records list:

AMD EPYC Processor World Records
https://www.amd.com/en/processors/epyc-world-records

Spoiler: Current list of 105 Epyc World Records with source links...

Original world record list at launch (as of 8/7/19):
1. 1P CPU2017 Floating Point Rate Base is a(n) 1-socket world record (tied) on AMD EPYC™ 7742. ROM-223 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16292.html#

2. 1P CPU2017 Floating Point Rate Base is a(n) 1-socket world record (tied) on AMD EPYC™ 7742. ROM-98 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16244.html#

3. 2P CPU2017 Floating Point Rate Base is a(n) 2-socket world record on AMD EPYC™ 7742. ROM-94 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16241.html#

4. 1P CPU2017 Floating Point Rate Peak is a(n) 1-socket world record on AMD EPYC™ 7742. ROM-97 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16244.html#

5. 2P CPU2017 Floating Point Rate Peak is a(n) 2-socket world record on AMD EPYC™ 7742. ROM-93 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16241.html#

6. 1P CPU2017 Floating Point Speed Base is a(n) 1-socket world record on AMD EPYC™ 7742. ROM-105 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16289.html#

7. 2P CPU2017 Floating Point Speed Base is a(n) 2-socket world record on AMD EPYC™ 7742. ROM-103 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16240.html#

8. 1P CPU2017 Floating Point Speed Peak is a(n) 1-socket world record on AMD EPYC™ 7742. ROM-104 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16246.html#

9. 2P CPU2017 Floating Point Speed Peak is a(n) 2-socket world record on AMD EPYC™ 7742. ROM-102 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16240.html#

10. 1P CPU2017 Integer Rate Base is a(n) 1-socket world record on AMD EPYC™ 7742. ROM-96 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16290.html#

11. 2P CPU2017 Integer Rate Base is a(n) 2-socket world record on AMD EPYC™ 7742. ROM-92 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16242.html#

12. 1P CPU2017 Integer Rate Peak is a(n) 1-socket world record on AMD EPYC™ 7742. ROM-95 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16290.html#

13. 2P CPU2017 Integer Rate Peak is a(n) 2-socket world record on AMD EPYC™ 7742. ROM-91 as of 8/7/2019.
Source: http://spec.org/cpu2017/results/res2019q3/cpu2017-20190722-16242.html#

14. 2P SAP SD 2-tier is a(n) 2-socket world record on AMD EPYC™ 7702. ROM-129 as of 8/7/2019.
Source: https://www.sap.com/dmc/benchmark/2019/Cert19044.pdf

15. 2P SAP SD 2-tier is a(n) 2-socket world record on Windows® on AMD EPYC™ 7702. ROM-128 as of 8/7/2019.
Source: https://www.sap.com/dmc/benchmark/2019/Cert19044.pdf

16. 1P SPEC ACCEL OpenACC is a(n) Overall world record on AMD EPYC™ 7742. ROM-150 as of 8/7/2019.
Source: https://www.spec.org/accel/results/res2019q3/accel-20190716-00127.html#

17. 1P SPEC ACCEL OpenACC is a(n) 1-socket world record on AMD EPYC™ 7742. ROM-221 as of 8/7/2019.
Source: https://www.spec.org/accel/results/res2019q3/accel-20190716-00127.html#

18. 1P SPEC ACCEL OpenCL is a(n) Overall world record on AMD EPYC™ 7742. ROM-151 as of 8/7/2019.
Source: https://www.spec.org/accel/results/res2019q3/accel-20190716-00128.html#

19. 1P SPEC ACCEL OpenCL is a(n) 1-socket world record on AMD EPYC™ 7742. ROM-222 as of 8/7/2019.
Source: https://www.spec.org/accel/results/res2019q3/accel-20190716-00128.html#

20. 1P SPEC MPI 2007 Medium is a(n) 1-socket world record on AMD EPYC™ 7742. ROM-153 as of 8/7/2019.
Source: https://www.spec.org/accel/results/res2019q3/accel-20190716-00126.html#

21. 1P SPEC OMP(G) 2012 is a(n) 1-socket world record on AMD EPYC™ 7742. ROM-152 as of 8/7/2019.
Source: https://www.spec.org/mpi2007/results/res2019q3/mpi2007-20190716-00637.html#

22. 1P SPECjbb2015-Composite Critical is a(n) 1-socket world record on AMD EPYC™ 7702P. ROM-120 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00463.html#

23. 1P SPECjbb2015-Composite Critical is a(n) 1-socket world record on Linux® on AMD EPYC™ 7702P. ROM-197 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00463.html#

24. 1P SPECjbb2015-Composite Critical is a(n) 1-socket world record on Windows® on AMD EPYC™ 7742. ROM-155 as of 8/7/2019.
Source: http://www.spec.org/jbb2015/results/res2019q3/jbb2015-20190718-00484.html#

25. 2P SPECjbb2015-Composite Critical is a(n) Overall world record on AMD EPYC™ 7742. ROM-126 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00461.html#

26. 2P SPECjbb2015-Composite Critical is a(n) 2-socket world record on AMD EPYC™ 7742. ROM-122 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00461.html#

27. 2P SPECjbb2015-Composite Critical is a(n) 2-socket world record on Linux® on AMD EPYC™ 7742. ROM-198 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00461.html#

28. 1P SPECjbb2015-Composite Max is a(n) 1-socket world record on AMD EPYC™ 7702P. ROM-120 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00463.html#

29. 1P SPECjbb2015-Composite Max is a(n) 1-socket world record on Linux® on AMD EPYC™ 7702P. ROM-119 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00463.html#

30. 1P SPECjbb2015-Composite Max is a(n) 1-socket world record on Windows® on AMD EPYC™ 7742. ROM-199 as of 8/7/2019.
Source: http://www.spec.org/jbb2015/results/res2019q3/jbb2015-20190718-00484.html#

31. 2P SPECjbb2015-Composite Max is a(n) Overall world record on AMD EPYC™ 7742. ROM-146 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00461.html#

32. 2P SPECjbb2015-Composite Max is a(n) 2-socket world record on AMD EPYC™ 7742. ROM-111 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00461.html#

33. 2P SPECjbb2015-Composite Max is a(n) 2-socket world record on Linux® on AMD EPYC™ 7742. ROM-121 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00461.html#

34. 1P SPECjbb2015-Distributed (1n) Critical is a(n) 1-socket world record on Linux® on AMD EPYC™ 7742. ROM-118 as of 8/7/2019.
Source: https://www.spec.org/jbb2015/results/res2019q3/jbb2015-20190718-00482.html

35. 2P SPECjbb2015-Distributed (1n) Critical is a(n) 2-socket world record on AMD EPYC™ 7702. ROM-190 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00470.html#

36. 2P SPECjbb2015-Distributed (1n) Critical is a(n) 2-socket world record on Linux® on AMD EPYC™ 7702. ROM-110 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00470.html#

37. 1P SPECjbb2015-Distributed (1n) Max is a(n) 1-socket world record on AMD EPYC™ 7702P. ROM-117 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00465.html#

38. 1P SPECjbb2015-Distributed (1n) Max is a(n) 1-socket world record on Linux® on AMD EPYC™ 702P. ROM-193 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00465.html#

39. 2P SPECjbb2015-Distributed (1n) Max is a(n) 2-socket world record on AMD EPYC™ 7702. ROM-157 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00471.html#

40. 2P SPECjbb2015-Distributed (1n) Max is a(n) 2-socket world record on Linux® on AMD EPYC™ 7702. ROM-157 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00471.html#

41. 1P SPECjbb2015-MultiJVM Critical is a(n) 1-socket world record on Linux® on AMD EPYC™ 7742. ROM-116 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190718-00480.html#

42. 1P SPECjbb2015-MultiJVM Critical is a(n) 1-socket world record on Windows® on AMD EPYC™ 7742. ROM-156 as of 8/7/2019.
Source: http://www.spec.org/jbb2015/results/res2019q3/jbb2015-20190718-00476.html#

43. 2P SPECjbb2015-MultiJVM Critical is a(n) 2-socket world record on Linux® on AMD EPYC™ 7742. ROM-108 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00462.html#

44. 1P SPECjbb2015-MultiJVM Max is a(n) 1-socket world record on AMD EPYC™ 7702P. ROM-115 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00467.html#

45. 1P SPECjbb2015-MultiJVM Max is a(n) 1-socket world record on Linux® on AMD EPYC™ 7702P. ROM-191 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00467.html#

46. 1P SPECjbb2015-MultiJVM Max is a(n) 1-socket world record on Windows® on AMD EPYC™ 7742. ROM-154 as of 8/7/2019.
Source: https://www.spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00474.html

47. 2P SPECjbb2015-MultiJVM Max is a(n) 2-socket world record on Linux® on AMD EPYC™ 7742. ROM-107 as of 8/7/2019.
Source: http://spec.org/jbb2015/results/res2019q3/jbb2015-20190717-00463.html#

48. 1P SPECpower_ssj2008 is a(n) 1-socket (2U) world record on Linux® on AMD EPYC™ 7742. ROM-186 as of 8/7/2019.
Source: https://www.spec.org/power_ssj2008/results/res2019q3/power_ssj2008-20190717-00986.html#

49. 1P SPECpower_ssj2008 is a(n) 1-socket (1U) world record on Windows® on AMD EPYC™ 7702P. ROM-148 as of 8/7/2019.
Source: https://www.spec.org/power_ssj2008/results/res2019q3/power_ssj2008-20190717-00984.html#

50. 1P SPECpower_ssj2008 is a(n) Overall world record on AMD EPYC™ 7742. ROM-185 as of 8/7/2019.
Source: https://www.spec.org/power_ssj2008/results/res2019q3/power_ssj2008-20190717-00986.html#

51. 1P SPECpower_ssj2008 is a(n) 1-socket (1U) world record on Linux® on AMD EPYC™ 7742. ROM-187 as of 8/7/2019.
Source: https://www.spec.org/power_ssj2008/results/res2019q3/power_ssj2008-20190717-00985.html#

52. 2P SPECpower_ssj2008 is a(n) 2-socket world record on AMD EPYC™ 7702. ROM-149 as of 8/7/2019.
Source: http://spec.org/power_ssj2008/results/res2019q3/power_ssj2008-20190716-00980.html#

53. 2P SPECpower_ssj2008 is a(n) 2-socket (2U) world record on Linux® on AMD EPYC™ 7702. ROM-183 as of 8/7/2019.
Source: http://spec.org/power_ssj2008/results/res2019q3/power_ssj2008-20190716-00980.html#

54. 2P SPECpower_ssj2008 is a(n) 2-socket (2U) world record on Windows® on AMD EPYC™ 7702. ROM-147 as of 8/7/2019.
Source: http://spec.org/power_ssj2008/results/res2019q3/power_ssj2008-20190716-00982.html#

55. 2P SPECvirt_sc2013 is a 2-socket world record on AMD EPYC™ 7702. ROM-99 as of 8/7/2019.
Source: https://www.spec.org/virt_sc2013/results/res2019q3/virt_sc2013-20190726-00122-perf.html

56. 2P TPC-DS @ 10TB is an Overall performance world record on AMD EPYC™ 7502. ROM-158 as of 8/7/2019.
Source: 2S TPC-DS @ 10TB 4,418,054 QphDS@10,000GB, $0.12/QphDS@10,000GB = http://www.tpc.org/5003#

57. 2P TPC-DS @ 10TB is a 2-socket performance world record on AMD EPYC™ 7502. ROM-238 as of 8/7/2019.
Source: 2S TPC-DS @ 10TB 4,418,054 QphDS@10,000GB, $0.12/QphDS@10,000GB = http://www.tpc.org/5003#

58. 2P TPC-DS @ 10TB is a Overall price-performance 10TB world record on AMD EPYC™ 7502. ROM-230 as of 8/7/2019.
Source: 2S TPC-DS @ 10TB 4,418,054 QphDS@10,000GB, $0.12/QphDS@10,000GB = http://www.tpc.org/5003#

59. 1P TPC-E SQL Server is a 1-socket price-performance world record on AMD EPYC™ 7742. ROM-179 as of 8/7/2019.
Source: http://www.tpc.org/4085#

60. 1P TPC-E SQL Server is a 1-socket performance world record on AMD EPYC™ 7742. ROM-184 as of 8/7/2019.
Source: http://www.tpc.org/4085#

61. 1P TPC-H @ 1TB 1-node is a 1 Processor world record non-clustered at 1000GB on AMD EPYC™ 7702. ROM-180 as of 8/7/2019.
Source: http://www.tpc.org/3342#

62. 1P TPC-H @ 1TB 1-node is a Overall non-clustered price-performance world record at 1,000GB on AMD EPYC™ 7702. ROM-237 as of 8/7/2019.
Source: http://www.tpc.org/3342#

63. 1P TPC-H @ 1TB 4-node is a Overall price-performance world record (tie) on AMD EPYC™ 7702. ROM-239 as of 8/7/2019.
Source: http://www.tpc.org/3340#

64. 1P TPC-H @ 1TB 4-node is a 4 Processor world record (historical) cluster on AMD EPYC™ 7702. ROM-177 as of 8/7/2019.
Source: http://www.tpc.org/3340#

65. 1P TPC-H @ 1TB 8-node is a Overall world record on AMD EPYC™ 7702. ROM-140 as of 8/7/2019.
Source: http://www.tpc.org/3341#

66. 1P TPC-H @ 1TB 8-node is a Overall price-performance world record (tie) on AMD EPYC™ 7702. ROM-141 as of 8/7/2019.
Source: http://www.tpc.org/3341#

67. 1P TPC-H @ 1TB 8-node is a 8 Processor cluster world record (historical) cluster on AMD EPYC™ 7702. ROM-139 as of 8/7/2019.
Source: http://www.tpc.org/3341#

68. 1P TPCx-HS @ 10TB is a Overall 10TB world record on AMD EPYC™ 7502P. ROM-165 as of 8/7/2019.
Source: http://www.tpc.org/5532#

69. 1P TPCx-HS @ 10TB is a 1-socket 10TB world record on AMD EPYC™ 7502P. ROM-166 as of 8/7/2019.
Source: http://www.tpc.org/5532#

70. 1P TPCx-HS @ 10TB is a Overall price-performance 10TB world record on AMD EPYC™ 7502P. ROM-167 as of 8/7/2019.
Source: http://www.tpc.org/5532#

71. 1P TPCx-HS @ 30TB is a Overall 30TB world record on AMD EPYC™ 7502P. ROM-161 as of 8/7/2019.
Source: http://www.tpc.org/5533#

72. 1P TPCx-HS @ 30TB is a(n) 1-socket 30TB world record on AMD EPYC™ 7502P. ROM-162 as of 8/7/2019.
Source: http://www.tpc.org/5533#

73. 1P TPCx-HS @ 30TB is a(n) Overall price-performance 30TB world record on AMD EPYC™ 7502P. ROM-163 as of 8/7/2019.
Source: http://www.tpc.org/5533#

74. 1P TPCx-IoT 4-node is a(n) 1-socket 4P total world record on AMD EPYC™ 7502. ROM-189 as of 8/7/2019.
Source: http://www.tpc.org/5754#

75. 1P TPCx-IoT 4-node is a(n) Overall price-performance world record on AMD EPYC™ 7502. ROM-228 as of 8/7/2019.
Source: http://www.tpc.org/5754#

76. 2P TPCx-V is a(n) Overall world record on AMD EPYC™ 7702. ROM-106 as of 8/7/2019.
Source: http://www.tpc.org/5302#

77. 2P TPCx-V is a(n) 2-socket world record on AMD EPYC™ 7702. ROM-123 as of 8/7/2019.
Source: http://www.tpc.org/5302#

78. 2P TPCx-V is a(n) Overall price-performance world record on AMD EPYC™ 7702. ROM-159 as of 8/7/2019.
Source: http://www.tpc.org/5302#

79. 2P VMmark 3.1 classic is a(n) 2-host, matched pair 2-socket world record on AMD EPYC™ 7702. ROM-35 as of 8/7/2019.
Source: https://www.vmware.com/content/dam/digitalmarketing/vmware/en/pdf/vmmark/2019-08-07-HPE-ProLiant-DL385Gen10.pdf#

80. 1P VMmark 3.1 vSAN is a(n) 1-socket, 4-host world vSAN record on AMD EPYC™ 7702P. ROM-36 as of 8/7/2019.
Source: https://www.vmware.com/content/dam/digitalmarketing/vmware/en/pdf/vmmark/2019-08-07-HPE-ProLiant-DL325Gen10.pdf#



Updated world record list (since 8/7/19):

81. 1P 2nd Gen EPYC™ 7742 powered server has set the 1-socket world record on the SPECpower_ssj® 2008 benchmark https://www.spec.org/power_ssj2008/results/res2019q3/power_ssj2008-20190717-00986.html# as of 08/07/19. ROM-185

82. 2P 2nd Gen EPYC™ 7742 powered server has set the Overall world record on the Cinebench - R15 benchmark https://hwbot.org/submission/4216395_sth_patrick_cinebench___r15_2x_epyc_7742_11080_cb as of 09/11/19. ROM-297

83. 2P 2nd Gen EPYC™ 7742 powered server has set the 2-socket world record on the Cinebench - R15 benchmark https://hwbot.org/submission/4216395_sth_patrick_cinebench___r15_2x_epyc_7742_11080_cb as of 09/11/19. ROM-298

84. 2P 2nd Gen EPYC™ 7742 powered server has set the Overall world record on the Cinebench - R20 benchmark https://hwbot.org/submission/4216519_sth_patrick_cinebench___r20_2x_epyc_7742_31833_marks as of 09/11/19. ROM-299

85. 2P 2nd Gen EPYC™ 7742 powered server has set the 2-socket world record on the Cinebench - R20 benchmark https://hwbot.org/submission/4216519_sth_patrick_cinebench___r20_2x_epyc_7742_31833_marks as of 09/11/19. ROM-300

86. 2P 2nd Gen EPYC™ 7742 powered server has set the 2-socket world record on the SAP® Sales and Distribution (SD) 2-tier benchmark https://www.sap.com/dmc/benchmark/2019/cert19047.pdf as of 09/11/19. ROM-301

87. 2P 2nd Gen EPYC™ 7742 powered server has set the 2-socket world record on Linux® on the SAP® Sales and Distribution (SD) 2-tier benchmark https://www.sap.com/dmc/benchmark/2019/cert19047.pdf as of 09/11/19. ROM-302

88. 1P 2nd Gen EPYC™ 7742 powered server has set the 1-socket world record on the SPEC ACCEL® OpenMP® benchmark https://www.spec.org/accel/results/res2019q3/accel-20190813-00129.html as of 09/11/19. ROM-303

89. 1P 2nd Gen EPYC™ 7702P powered server has set the 1-socket world record on the SPECrate® 2017 Floating Point (Energy Base) benchmark http://spec.org/cpu2017/results/cpu2017-20190903-17793.html as of 09/11/19. ROM-304

90. 1P 2nd Gen EPYC™ 7702P powered server has set the 1-socket world record on the SPECrate® 2017 Floating Point (Energy Peak) benchmark http://spec.org/cpu2017/results/cpu2017-20190903-17793.html as of 09/11/19. ROM-305

91. 2P 2nd Gen EPYC™ 7702 powered server has set the Overall world record on the SPECrate® 2017 Floating Point (Energy Peak) benchmark http://spec.org/cpu2017/results/cpu2017-20190903-17796.html as of 09/11/19. ROM-306

92. 2P 2nd Gen EPYC™ 7702 powered server has set the 2-socket world record on the SPECrate® 2017 Floating Point (Energy Peak) benchmark http://spec.org/cpu2017/results/cpu2017-20190903-17796.html as of 09/11/19. ROM-307

93. 2P 2nd Gen EPYC™ 7702 powered server has set the 2-socket world record on the SPECrate® 2017 Floating Point (Energy base) benchmark http://spec.org/cpu2017/results/cpu2017-20190903-17796.html as of 09/11/19. ROM-308

94. 1P 2nd Gen EPYC™ 7542 powered server has set the Overall world record @ 1TB (SPARK - Framework 2) on the TPC® Express Benchmark HS @1TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-309

95. 1P 2nd Gen EPYC™ 7542 powered server has set the 1-socket @ 1TB world record (SPARK - Framework 2) on the TPC® Express Benchmark HS @1TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-310

96. 1P 2nd Gen EPYC™ 7542 powered server has set the Overall price-performance @ 1TB world record (SPARK - Framework 2) on the TPC® Express Benchmark HS @1TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-311

97. 1P 2nd Gen EPYC™ 7542 powered server has set the Overall world record @ 1TB (Map Reduce - Framework 1) on the TPC® Express Benchmark HS @1TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-312

98. 1P 2nd Gen EPYC™ 7542 powered server has set the 1-socket @ 1TB world record (Map Reduce - Framework 1) on the TPC® Express Benchmark HS @1TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-313

99. 1P 2nd Gen EPYC™ 7542 powered server has set the Overall price-performance @ 1TB world record (Map Reduce - Framework 1) on the TPC® Express Benchmark HS @1TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-314

100. 1P 2nd Gen EPYC™ 7542 powered server has set the Overall world record @ 3TB (SPARK - Framework 2) on the TPC® Express Benchmark HS @3TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-315

101. 1P 2nd Gen EPYC™ 7542 powered server has set the 1-socket @ 3TB world record (SPARK - Framework 2) on the TPC® Express Benchmark HS @3TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-316

102. 1P 2nd Gen EPYC™ 7542 powered server has set the Overall price-performance @ 3TB world record (SPARK - Framework 2) on the TPC® Express Benchmark HS @3TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-317

103. 1P 2nd Gen EPYC™ 7542 powered server has set the Overall world record @ 3TB (Map Reduce - Framework 1) on the TPC® Express Benchmark HS @3TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-318

104. 1P 2nd Gen EPYC™ 7542 powered server has set the 1-socket @ 3TB world record (Map Reduce - Framework 1) on the TPC® Express Benchmark HS @3TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-319

105. 1P 2nd Gen EPYC™ 7542 powered server has set the Overall price-performance @ 3TB world record (Map Reduce - Framework 1) on the TPC® Express Benchmark HS @3TB benchmark http://www.tpc.org/tpcx-hs/results/tpcxhs_advanced_sort.asp?version=2 as of 09/11/19. ROM-320

106. 1P 2nd Gen EPYC™ 7742 powered server has set the 1-socket world record on the TPC® Express Benchmark V benchmark http://www.tpc.org/results/t5303.asp as of 09/11/19. ROM-321

107. 1P 2nd Gen EPYC™ 7742 powered server has set the Overall price-performance world record on the TPC® Express Benchmark V benchmark http://www.tpc.org/results/t5303.asp as of 09/11/19. ROM-324


FOOTNOTES
SPEC®, SPECvirt®, SPEC CPU®, SPECrate®, SPECjbb®, SPECpower®, SPEC ACCEL®, SPEC MPI®, SPEC OMP®, and SPECspeed® are registered trademarks of Standard Performance Evaluation Corporation. Learn more at spec.org. TPC and TPC benchmark are trademarks of the Transaction Processing Performance Council. VMmark and VMware vSAN are trademarks of VMware in the US or other countries.

Here's an older graphic when AMD Epyc only had 100 World Records:

 

Today was a wild ride for Intel, why oh why did Intel release on the same day as AMD? In the past Intel did this to have stolen sparkle from AMD, but today it really back fired on Intel, and showed how far Intel is behind AMD...

Ryzen 9 3950X vs. Entire Intel Cascade Lake-X Lineup, When Price Cuts Aren't Enough
Nov 26, 2019
Hardware Unboxed
Ryzen 9 3950X vs. Intel's Entire Cascade Lake-X HEDT Lineup

 

Jay gives up on Intel and Switches to AMD... not a big surprise, but it's a nice build - ThreadRipper Orange!!

AMD won us back! Switching back to AMD
Nov 25, 2019
JayzTwoCents

Spoiler: Parts used in Build:

►Main Parts List
Phanteks Enthoo 719 (Formerly Luxe 2)
ASUS Zenith II Extreme Motherboard
AMD Threadripper 3970X CPU
64GB Corsair Dominator Platinum RGB Memory 3600MHz
EVGA RTX 2080Ti XC GPU
Corsair 1TB MP600 M.2 Gen 4 NVME SSD
Kingston Now 1TB Sata SSD Scratch Disk
10TB x2 Seagate Ironworf Enterprise HDDs

►Cooling Gear
Phanteks Glacier X399 AMD Waterblock
Phanteks Glacier Founders Edition 2080/2080Ti Waterblock
Phanteks 16mm/10mm Soft Tube Compression Fittings
Singularity Computers Protium Reservior/D5 Combo
AlphaCool 480mm Radiator
AlphaCool 360mm Radiator
10x Corsair LL120 RGB Fans
PrimoChill 16mm/10mm LRT Clear Soft Tubing
Corsair XL5 Premix Coolant Mayhems Orange Dye

Jay drops more stuff, again...

Jay: "I became so mad at Intel I didn't even fire it up [10980xe]" - 05:40

Jay: "Intel has been ripping you off, all along" - 05:42

Intel REALLY needs to rethink it's priorities...
Nov 27, 2019
JayzTwoCents
The Intel Core i9-10980XE is here... but it wasn't even worth turning on...


As I've been saying for a very long time, Intel is squandering their position in the market and being poor stewards of the technology and it's going to come back on them one day.

That day has been here for almost 3 years now, and a lot of people have already woken up and dropped Intel. Is today the day for you to drop Intel? Better earlier than later...

Even if Intel is a couple of FPS ahead in particular games, is it really visible or noticeable in everyday play? Is it worth feeding the beast - rewarding Intel for their bad behaviors? Can you really feel good about this and still enjoy gaming on Intel? I know that if I have a choice to choose AMD, then I can't choose Intel today and the options to choose AMD over Intel are increasing...

Awards: Best CPUs of 2019 (Gaming, Production, & Disappointment)
Nov 26, 2019
Gamers Nexus
Our round-up of the best CPUs of 2019 focuses on gaming, video editing and production, 3D modeling and animation (Blender), overclocking, budget class, and more!
Article: https://www.gamersnexus.net/guides/35...

Spoiler: Details...

This recap of 2019 CPU reviews covers a few key category, including best CPU for video editing, best CPUs for Blender and 3D Modeling, best CPUs for gaming and overclocking, and most well-rounded CPUs. We also talk about top processors for small business and hobbyist production workloads and, counter to that, for ultra-budget gaming PCs. Of course, it wouldn't be a GamersNexus video without also talking about the biggest disappointment of the year and the worst trends of the year. Use this guide to determine which CPU you should buy for Photoshop, Adobe Premiere, Blender, gaming, work, or value / budget. Conveniently, we're also posting this just before Black Friday and Cyber Monday, so CPU sales and deals should start soon on older hardware.

CPUs FEATURED (in order of appearance) - See Video and Youtube channel listing for more Details...

 

Another year gone, another CPU review of the year...not a big surprise, there's nothing much from Intel to recommend this year.

Top 5 Best CPUs of 2019, Gaming & Productivity
Dec 4, 2019
Hardware Unboxed


It's gotta be higher than 60% in the US, right?

60% of European PC Enthusiasts Prefer AMD CPUs, According to EHA Study
UPDATED by Raevenlord Monday, 03:30 Discuss (31 Comments)
https://www.techpowerup.com/261732/...efer-amd-cpus-according-to-eha-study#comments

"An independent study conducted by the European Hardware Association (EHA) has revealed that AMD now ranks higher than Intel in the CPU space. While we have seen this as recently as last week, where we reported on top sellers across some Amazon webstores in Europe, with AMD scoring most of the top sellers in both Germany and the UK, this is the first time a comprehensive study has put some verifiable, science-generated numbers for us to see.

According to the EHA, 60% of the European PC enthusiasts (in a sample of 10,000 respondents) showed a strong sentiment towards AMD as their favored manufacturer of CPUs, and would choose any sort of system with an AMD CPU over an Intel one (including APU, AMD + Radeon graphics cards and AMD + NVIDIA graphics cards).

This is a far cry from the same time around last year, where AMD only held 40% of a similar sample's preferred buying intention, and up from the 50% shown in the same study, carried out in 2H2019.

The same survey also shows a slightly increased preference for AMD's graphics cards, with the 1H2019 showing 19% preference compared to 23% in this latest study."

...chart + more + comments on site...

the54thvoid Posted on Dec 2nd 2019, 4:31
"To be fair, most technophobic people don't favour either, but the point of this specific survey still stands; those who do know about CPUs and their branding have come to favour AMD."

 

The point of this study is that people that favored AMD platforms were more vocal than the Intel camp at this point in time.

Ignore simplistic studies such as this and buy what actually gives you better results for your workflow/workloads. Do top sellers matter? Who are these simpletons? They don't do what I do with my platforms.

Just because a billion+ flies eat $&^#, doesn't mean it tastes good.

The AMD marketing machine is in full swing, but I don't see much change here. Especially for the normal users and particularly for mobile users. Every single AMD system I have played with seems woefully 'un-snappy' to me compared to my Intel setups. Keep going AMD, you're getting closer... maybe next year.

Furthermore, having a preference for 'any sort of AMD CPU powered system over an Intel one' shows just how little their 'vote' counts. Going into any decision with that kind of bias makes you less than intelligent. And definitely doesn't show that one platform is superior, or not, for a given workflow.

 

As far as snap, AMD feels fine here on the HEDT front. I prefer my Mobile AMD to Intel offerings as well. But this is my use case. The study is too mainly show that the pre Ryzen stigma of having an inferior machine no longer exists, you have to get over that.

I am not happy with AMD marketing and lies but now they are just all on an even playing field in that manor. Overall I do not regret having AMD just not as happy as I was.

 

Curious about which Mobile AMD you prefer vs. which Intel? The 10th gen Intel (10nm) is (again) at a different level than where we were 'minutes' ago.

 

I do not prefer AMD mobile as they are comparable performance wise for my use. Once 7nm makes it to mobile then I would prefer that. My preference as it is now for the AMD 2500u, and usually the 3500u, would be for comparable performance cost is lower. Again my use case which is as a casual home, non gaming, user for a mobile device.

 

The points you are right on:

1) Buy for your need,

2) Intel is better on mobile, and

3) this is but one data point.

But you are wrong in most protestations on AMD. You cannot take 80% of retail DIY sales at one of the largest retailers in Europe with bad hardware.

Where Intel remains king:

1) pure game machines seeking the highest FPS, and

2) photoshop and specific tasks in adobe software.

AMD outperforms at almost every other price point at almost any other task with Zen 2 desktop and HEDT. You even have the 3960X outperforming the W3175X Xeon with 4 more cores in some cases, while the 3970X wraps up the rest in most cases, all at a lower price point!

So, except on some things, you are largely wrong.

 

My points stand. Your points need actual supply to become real.

And they still won't be useful or interesting to anyone in the 'normal user' category.

 

Are you kidding me? Look at no supply on cascade X. Look at Dell thinking of increasing AMD offerings because Intel's shortage is set to extend to H2 2020. Pot meet kettle.

And sales figures from GN affiliate links, the European hardware survey, and mind factory all say a different story. Granted, Intel still has the OEM pre-built market, but that will do very little if they have to switch due to supply.

Now, who do you think will get supply first? Intel is now back porting Willow cove to 14nm for rocket lake, allegedly. 14nm production is so backed up, Intel doesn't have a solution. Ice lake SP won't help as Cooper lake will be on 14nm and will be the premier chips for servers even with 10nm around.

So would you like to delve deeper into Intel's manufacturing woes?

 

I don't care about anyone's manufacturing woes. I care about products that I can buy now, if I'm buying now, and will only keep them if they're better than what I have. Having sampled both the latest available Intel and AMD products that interest me, AMD is still catching up. My general observations above were what was trickled down to 'normal' users, which you still think need the latest node to have the best experience (not).

Sales figures from 'affiliate links', 'hardware surveys' and 'mind factory' data points don't factor into this decision.

I have my own mind that I can make up, after all. No need to swallow or be spoonfed marketing fluff that is only tangentially true, at an absolute level, especially for the masses.

 

Your workload and uses benefit from Intel. Your snappiness suggestion is subjective garbage based on your sole experience and can even be based on software loaded on the OS, among other factors.

But I do give credence to your workloads FOR YOU being better on Intel. But there is a crap ton of other data points showing AMD doing better on other workloads. You not acknowledging that, then denigrating consumers that come to results and decisions different from yours is you ignoring their needs.

 

Yeah, the snappiness of the O/S, the programs I use and the time it takes the system to settle down from a full shutdown (with no hibernation file)/restart and recently, the startup using a hibernation file too (for my less stressed platforms) are all important to me.

Subjective? Sure. Garbage? If you don't care about it on your platforms; um, okay.

For the millionth time; I can and do see how normal people use their systems... Once again, this isn't about my workflows/workloads... this is what anyone else will get for a limited 'xx' dollars. AMD is deficient with regard to those points on the systems I've used/tested.

The workloads AMD is better at is deep multi-core compute (video 'editing'... for most consumer workload/users).

For most consumers though, that is not a thing for them - ever. And for those cases, a faster/more responsive computer is better. Not just having more cores thrown at an imaginary problem(s) as AMD would have us believe.

I've acknowledged that AMD in the multicore race is leading (and I've said that at least a year ago now...). That doesn't make them better where 99.9% of consumer loads sit though.

 

Since you claim the "comparative systems you are interested in" then mention what they are. as I fear performance per dollar AMD in most cases wins. On gen 1 and gen 2 Ryzen I can see you case on occasion, but not with 7nm situations.

Edit; I also have to take notice, you say for 99.9% of users. Not true. The average users system is bloatware loaded to the brim. Having a huge reserve of cores to tackle that while still being operable is a huge plus.

 

Looks like someone is reading the Intel propaganda on which apps mobile users use and applying it to desktops. Lmao.

 

Lol this thread cracks me up. I guess I don't notice any difference in "snappiness" of my Intel systems vs my AMD systems, and I've had a few over the past couple of years:
6700k
7700k
8700k
9700k
9900k
7920x
9800x
9920x
1700x
2700x
3900x

The only thing I miss about my x299 setup is the fun in overclocking it (and quad channel memory). But most of those systems were fast enough to do everything I wanted to do on them. But my workload is gaming, solid modeling, massive excel spreadsheets, and transcoding/decoding/encoding large video files. A SSD (sata or NVMe doesn't matter) really helps to provide that "snappiness" that we all enjoy and love.

 

It can be fun. Tiller promotes Intel to a fault. Hmscott promotes AMD to a fault (recently accused me of being an Intel shill, whereas others have accused me of being an AMD fanboy/shill; when accused by both sides, you must be doing something right).

But, yeah, it is fun when it doesn't get out of hand, and something for everyone to enjoy.

 

No Intel propaganda necessary, I use my own experience, including my interactions with 1,000's of 'normal' users too, over the years. And btw, this is a mobile forum, no?

Keep trying to side-skirt the real issues I'm highlighting. No worries. I'll try to keep you focused.

Intel may be stumbling for a few years now, comparing to ~2016 and earlier, but what it does offer, has real substance. I know I don't buy a new platform (Intel or not) without reason - others must (obviously) feel the same. And while it is not on its unreachable peak anymore, it is still comfortably above anything AMD offers in the mobile space.

To see for yourself where Intel is excelling at today, look at even the entry-level SP7 w/8GB RAM and 256GB SSD. Nothing AMD offers even comes close to that level of responsiveness in mobile and in my experience so far, even on the desktop space (vs. an appropriate desktop Intel platform, of course).

I don't expect everyone to appreciate these types of 'performance' gains, but just because one is not sensitive to these metrics, doesn't mean we're drinking the Koolaid either... from either side.

To give you an analogy to better understand my perspective, AMD is like a big dump truck that can handle a lot of 'load' at once. Intel is more like a sports ute that can handle everyday/normal 'loads' without waste (i.e. having 12+ extra cores sitting there idle).

I can appreciate the power of that dump truck with its high load limits that it offers. But I'd rather be more connected to the 'road' (i.e. real life) with something a little less powerful in an absolute sense, but much nimbler and satisfying each time I take it for a spin (whether that is just a little spin around the block, or, for a trip across the country).

And while I am more focused (in this forum) on mobile systems, I don't see that much difference on the desktop side either for normal, everyday users (because most of them are using a mobile platform by now too).

The propaganda that has permeated our society over the last three years is that 'moar cores' are the answer to everything. Not yet. There are other things just as important that have to be mastered still.

I'll once again thank AMD for giving Intel the push it needed to provide us with the choices (from both AMD and Intel) we all want and welcome.

 

First, my comment was a reference to Intel using how mobile users use their computers to say AMD is useless, putting it right beside Intel's HEDT presentation with no separation, which Intel was derided by the press for doing.

Second, this is a mobile forum, but look at the section this is in and the title of the thread. NOT TALKING MOBILE! It is literally all desktop and server. So you admit to being off topic.

Third, as countless reviewers point out, Intel can have the highest frame rate in gaming. For the average performance, AMD gives the performance per dollar, and with Zen 2 often the straight out performance, which is why they are recommended over all Intel desktop chips but the 9700K and 9900K, which are just recommended for gaming scenarios.

For mobile, I already conceded Intel has best performance. But trying to them say because they perform best in mobile, but their desktop systems is malarkey.

You are now moving away from saying your workloads to generalities. Guess what? Your generalities have been proven false by countless data points. Even your subjective "snappiness" standard has been refuted by countless reviewers, which have used WAY more variety of hardware each release cycle than you ever touch and consider.

Also, your reference to history MEANS ABSOLUTELY NOTHING. Nowhere in that time did AMD have an IPC lead, process lead, and ever have roughly the same performance or better than Intel's in the past 15 years or about that. So comparing things you heard or saw during the bulldozer days does not and cannot apply to Zen 2 chips.

That is how absurd you are being. Nearly as bad as Lyan Shrout and his bs.

 

I see you've already made up your mind and know me better than I know myself (or my experience). And, I haven't changed my stance at all, and I'm not talking 'just' gaming either (for myself). Sigh.

The entire forum here is 'notebookreview.com', that is why I question a mere thread talking about desktops.

Average performance means squat. I'll repeat it again: Intel gives the gains where it is actually needed, today (and still). 'Moar cores' (for normal, everyday users/workloads) are the Koolaid that AMD offers. I don't accept (but have tasted/tested it - and not just for my workloads/workflows (and just stop with that flimsy excuse, it's really getting old now)...).

...

Here you go, take it. You win. Happy?


Everyone else reading along can see my points clearly enough.

And here is exactly what I'm talking about.

Uh, where exactly are those applications for the masses... (the question I've been asking for over 2 years now...).

See:
https://www.tomshardware.com/news/a...ores-coming-in-the-era-of-a-slowed-moores-law

 

Pre Ryzen I was in a DTR culture. This was spurned by Intel as from all C2D's thru quad core I7's there were fairly powerful mobile offerings. I was fine for a long time on this model.

Then came the 16 core monsters for desktop's, all heavy workloads became menial. So here we are today.

 

We need more truth.

See:
https://www.tomshardware.com/news/reality-check-sorry-amd-hasnt-tripled-its-market-share-yet

But this is a good start.

The other 'data' presented as market share is similarly flawed, even if with different metrics.

 

Thos numbers are misleading. Since they encapsulate results from specific types of users and usually newer machines they may show a trend towards but not actual full typical user data.

 

So you seem to miss my explanation of retail numbers being DIY, which is a subsection of the consumer market. Never did I claim overall market share. Nor did I point to those two places, steam and passmark, due to sample size and polling procedures. A couple years back, steam would not ask to sample my AMD 1950X, but wanted to sample my laptop's 4790K whenever it connected every couple months.

Retail sales figures, Amazon's top selling skus, mindfactory's sales data as the largest retailer in Germany, or he EU hardware survey done by a company which polled 10,000 people, that has a very low error rate due to sample size, are very different than your article attacking those two specific sources. Polling on what consumers want at over 10,000 when the error rate is around 2% for 2,000 people polled says people want AMD. https://www.techspot.com/news/82972-survey-amd-cpus-preferred-60-europeans.html

If instead you want analyzed market share numbers, I can look those up, but the last time I did, AMD was mid-single digits in servers (Epyc 2 and subsequent Intel vulnerabilities may have changed that a bit), AMD had 12% of mobile, and 18% of desktop. Would you like me to check the update on those numbers?

As I said, Intel's manufacturing woes leading Dell to consider AMD for OEM machines would be the easiest way for Intel to lose market share. Intel will have chip shortages until H2 2020, continuing to shrink their presence in server and desktop, while they focus on keeping mobile.

Server sales take awhile to filter in due to test beds.

Edit: Here is Mercury Research, an oft cited company that analyzes market share. I usually use another company, but it was a quick search and I haven't had my coffee this morning.
https://www.tomshardware.com/news/a...-share-7nm-makes-landfall-as-price-war-begins

So according to them AMD is under 5% in server, which is about a percent off from what I've seen elsewhere, but not a large discrepancy. Mobile is over 14%, which my numbers were from Q2, so possible and inline with projections. And the 18% before really counting impact of Zen 2 chips also makes sense. Overall, it is about the same as the other company.

Now look at what people want versus what the market share is. That suggests the desktop segment will be shifting significantly. Mobile will take AMD releasing Zen 2 or Zen 3 chips for it to make a splash, which is early 2020 and in 2021. But Intel is also still cranking away with comet and ice lake ATM.

But enthusiasts and DIY types have spoken and Zen+ and Zen 2 CPUs really are flying off the shelves. As their mindshare changes, and that of tech reviewers which now recommend AMD over Intel, it is a matter of time.

That is why I point out that since Intel is still having manufacturing woes, unless they get their 7nm process working by 2021 (equivalent to TSMC 5nm) when AMD goes to 5nm TSMC, which a report yesterday suggested yields on 5nm at TSMC have already reached the yields seen on 7nm, then Intel will have even worse problems.

TSMC has greatly expanded their 5nm capacity. They have outspent Intel on expanding capacity and are the largest semiconductor fab in the world! Intel and Samsung usually tout their profits or revenue to say they are the largest, but that includes SELLING THEIR PRODUCTS. It looks at market float. Since TSMC is ONLY a pure play fab, they are not selling products. All they sell is fab time. So them drastically spending $11B to increase 7nm and 5nm capacity, along with their fab 18 for 5nm coming online for phase 2 next year and phase 3 in 2021 to do volume production, along with switching over some 7nm fabs and their 5nm fabs being able to switch to 3nm, which their 3nm fab in 2023 being a MONSTER size, suggests AMD's ability to provide more chips to market will only grow in the coming years while Intel can barely get 10nm to market STILL TO THIS DAY. How are they really going to do with Ice Lake SP?

 

for responsiveness, assuming zen 3 is 8% IPC over zen 2 as current rumor suggest, it will need ~4.3ghz to be equivalent to cascade lake at ~4.9 ghz.

the good news is that the large cache isn't really changing so maybe the improvements are from inter core and ccx latency reduction which would be nice to have, which actually means better IPC due to more efficient CORE instead of Cache and zen definitely needs that if it wants to beat intel with lower frequency.

i really want to see zen3 with an excellent IPC over zen2 while getting us 4.6ghz on all cores but I think intel will probably get there first with tiger lake and 10nm++. 4.6ghz doesn't seem all that unreachable with 10nm++ even considering how bad 10nm+ is right now, next year we'll have tigerlake mobile or possibly desktop chip at 4.4 to 4.6ghz with 20-25% IPC over my 8700k.

 

First, Intel is NOT doing a tiger lake desktop CPU. They are doing comet in Q1 if shortages allow, then rocket lake, which is Willow cove back ported to 14nm and due to core size will only allow for 8 core CPUs on mainstream. I won't argue that tiger on mobile will likely be better than the AMD Zen 2 mobile chips, because that likely is the case. But the optimism on desktop is foolhardy.

Now, let me cover the current rumors on Zen 3 and what they likely mean. The first rumor I'd like to address is IPC. The most recent rumor suggests around 10% on integer, up to 50% on floating point, and around 17% in mixed integer and floating point operations. No matter what, this means AMD is going wider on the CPU designs. One possibility for such a high floating point is AMD adding AVX512 support. That would mean either adding a single 512 pipeline or using 2x256 pipelines and some magic. That would give that large of a jump.

The other possibility is they are adding the similar AI inference tech that is also able to do some floating point, but would need some software mods for some programs to use. This is supported by Papermaster mentioning they do plan to support this in hardware at some point. The AVX512 argument and wider pipeline is supported by AMD's statements of waiting for software adoption of instruction sets before implementation.

For the integer IPC, part would come with change in pipelines, part in inter core comms, part from access to L3 increasing per core with less hassle to access it.

As to frequency, for servers, there is talk of an additional possible 200MHz. But this may not result in a boost to consumer offerings already clocked close to the curve on the frequency/voltage knee.

Also, where are you getting your bull from? Thinking Intel, who can't even provide sufficient supply of ice lake mobile chips to OEM laptop manufacturers and was in the low teens on yields on 10nm a year ago will magically get increased yields and over a 17% improvement in frequency in a single generation when nothing they have actually shown supports this. In fact, in many ways the ice lake mobile chips are MORE power hungry than the 14nm counterparts. Without solving that, you cannot improve the efficiency enough to get the higher frequency.

Further, your statement that the cache size is not changing shows you don't understand that cache systems are not just about the size of the cache, but the implementation. By removing the CCX constraint, where each core only had access to 16MB before having to travel off of the chip, it can now access 32MB, fed by work done by other cores as well, before jumping off chip regarding L3 cache access. That is a large change!

Further, AMD is already able to challenge Intel while having lower frequencies. Check out the reviews of the 3950X beating overclocked 18 core Intel CPUs.

So you seem to be misunderstanding and misinterpreting the data available!

 

@ole!!!, don't worry! We have someone to interpret and understand the data for us.