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

Of course they will add in the price differences and not the performance differences with equal core count cpu from Intel I would do the same. Read Endnotes slide 13: Where can I see price mentioned?


For the records... They prefer compare with lower Cinebench numbers for 10 cores as well. Wonder why.

 

That's because cost is more important than performance to 99% of the people that spend money on their own computer.

Adequate performance for the minimum of cost.

There is a halo effect of foolishness that compels people to lay out extraordinarily inordinate amounts of wasted time and money resources for maximum performance that they won't be able to use or if usable even notice in day to day use.

The carefully curated exceptions are the few benchmark tools that amplify the difference enough to seemingly justify the ridiculous expenses.

You can only run these synthetic aberrations so many times before you realize you are wasting your life. Finally realizing it is all a futile attempt to justify inducing your expenditures of wasted money and life opportunity.

Inducing you to spend beyond your limits to win at competitions carefully constructed to see how much more you can be induced to spend.

Spending more than you need for performance you don't need is wasted opportunity for other endeavors of equal or far more importance in your life.

That's why the Intel 7960 @ $1699 isn't paired up in comparison with an AMD 2950x that only costs $899; the 7960 is a waste of money in comparison.

 

I'll review after I wake up more. These for kids visiting can wear you out.

But, you need to compare test rigs, time taken, cooling, whether the all core enhancement was active as that controversy came up last fall, if I remember correctly, and all meltdown and spectre fixes. Until I review everything, I cannot give a good answer (on my phone, not at my desk).

Sent from my SM-G900P using Tapatalk

 

Yeah, many use software who use less treads/cores - single threaded workflow. This mean they should go for the cpu who offer best single threaded performance. And this ain't AMD Ryzen.

Yeah, The new Ryzen processors ain't out in hands to all yet.

 

They are using the comparison for two reasons, cost as mentioned is one. Tier level for consumer offerings is another. The 7900 right now is their mid offering, not the 7960. Now if the 28 core were being offered on the market to consumers then you might be able to argue the 7940 is the mid tier offering.

Funny now the Intel people want to only compare core count and possible frequency at those reduced core counts, not actual market offerings.

 

Although agreeing overall with the sentiment, benchmarking is still core for core, which is his competitive complaint.

This is multifaceted. Ryzen+ on AM4 has about a 1-4% IPC deficit if you look at hardware unboxed's comparison at clock for clock. With multithreaded workloads (smt & ht) and software optimization taken into account, we see AMD get closer than the single threaded IPC on some tasks.

Intel clocks higher. It goes further under LN2. But, for everyday builds, you don't get an extraordinary boss in clocks like at lower core count mainstream chips. Considering Intel only mentioned a 20 and 22 core skylake X chips, and 24-28 for cascade using 14nm++ like coffee and the upcoming 9000 series, we may see, for a year, the current offerings on HEDT going against the TR2 using zen+. The new 16 core should theoretically get to an extra 100-300mhz, but we'll just say 4.1-4.2GHz daily driving will be more common for AMD workstations.

If 3600 is more easily achievable on ram, then 3600 on CB will be easier to achieve (did it at 4.1GHz on my 1950X with 3600 ram tight timings). I do not know what the 16 core all core overclock average was or score ATM, but I'll see if silicon lottery still has that up or not or look on hwbot in q bit. But the all core OC closing is a good thing. At the price point, AMD is a no brainer this round.

(18C core versus 32C; if looking for benchmark records or for heavily multi threaded tasks, there is no reason to buy Intel 7980XE that I see)

But some software and benchmarks cannot yet use chips like this, which is a state of the overclocking circle which is slowly dying (meaning the competitive OCing, not the act of OCing, which will be here forever).

So, nuance. But on a per core rather than price point, sure, papusan gets it right and Intel wins. But with the new pricing difference, it is a pyrrhic victory.

Sent from my SM-G900P using Tapatalk

 

I can't see the huge performance gain from 2950X vs. 1950x. Around 2.9% increase due 1 bin higher base clocks.

The gain is single threaded performance. Maybe gen 2 overclock better. I don't know.



The interesting will be in this segment for most people (I don't talk about pro-sumers)... The 2920X and 2950X are targeted at high-end gaming builds, in which you're gaming, streaming, and doing a whole bunch of other things in tandem.

 

If you need single threaded, then buy mainstream. Seriously. As you go above 10-cores, you see it drop to 4.4GHz toward good OC on 18-core chips. Difference in single threaded between an AMD at 4.1-4.3 and Intel at 4.4 really isn't that much.

Plus, if using that rationale, the best intel will have is their 28-core chips going against the 32 core chips. Even if it can win in single threaded, for the expected cost, you could buy the 32 core PLUS and Intel mainstream rig, top end, and still come away with money over getting the 28-core. Many production studios, etc., would opt for that.

Now, because AMD forced the point, mainstream is growing to 8 cores with Intel's offerings, so FINALLY mainstream is more attractive, but still not there due to offerings on HEDT now. But, if you want single threaded, mainstream is where it is at.

Also, on the gimmick that was hardware acceleration for intel on Adobe media encoding.
https://www.pugetsystems.com/labs/a...on-in-Adobe-Media-Encoder---Good-or-Bad-1211/
(seems like snake oil now)

 

As I posted in bottom of my previous post, 12 and 16 cores will be more and more interesting. This for users who want high single threaded + multi threaded performance. Offer a bit more than mainstream and still can have high single core Clock speed. On top should be easy to add a nice overclock. Aka +-5.0GHz over all cores. Best from both worlds.

 

Now that I can agree with. And there is still the per task usage that needs considered. Some things still work better with Intel, others have been shown to work real well with AMD HEDT. At the new pricing tiers, it is just phenomenal, by bringing the per core cost down to $50-60 range for the entire AMD Threadripper lineup.

But we all already agree, on the price/performance, AMD is far and away the king. On performance, Intel is with single threaded and specific tasks (AMD got some of those tasks with TR, and the bit extra with TR2 might help a bit, but on a per core basis, still limited until you see inflection at a specific HCC or XCC chip, at least until the 14nm++ is used for the upcoming cascade-X chips, which we will see when they arrive).

Now, I agree 12 and 16 are more interesting, and am ecstatic AMD is pushing those for gaming. Why? Because it is to tell the companies doing games and single threaded apps to expect that to be where games are had (there are rumors of 12 or 16 core chips coming to mainstream AM4 socket). So this may be further signaling to them, and Intel, that this change is coming whether they like it or not. Think of the arguments made at going from dual to quad cores, or using 6 or 8 cores for gaming on HEDT just a couple years back. So, we may see new solutions to multithreaded apps soon.

I'm also staying away, as much as possible, rumors of speed boost on next year's chips.

 

CB will get little from just 100 MHz on TR2 increase. The fact it is still just 2 CCX complex's will help IPC more than Cinebench. It will still be below Intel, as is the 2700X. This is to be expected in games. As a 1440P gaming rig it will matter little if at all.

For now Intel has to give away the high end HEDT to AMD. With the 28 core, if prices are reasonable, they may be able to compete at least till the 7nm CPU's come around. Intel should have started over a year ago on a new group of CPU's. This will be getting harder as time goes to finance the R&D as AMD starts eating into Intel's deep enterprise hardware pockets.

Myself, I am not as keen on a 2990WX anymore. The price is a bit high for my liking, the performance increase not that high and I still can't justify the power I already have with the 1950X. TBH I would have been fine with just a 1700 and even then it would have been overkill. Other than benchmarking I never see more than 15% CPU usage (and rarely that) even encoding my DVD's for archive..

 

So, for time spy, I got 11628 with SMT off, but got 10983 with it on. For Firestrike, I got around 30K for CPU, but dropped to 24-25K with SMT off. I forgot to screenshot Time Spy Extreme, so will revisit. @Johnksss

 

Thanks for that.

 

OK, I saved the result, just not a screenshot. With SMT was 7060 for Time Spy Extreme, without SMT was 5707. So it seems only Time Spy original has the issue on cores, not firestrike/extreme/ultra or Time Spy Extreme.

 

I think a few people are in for a surprise. The 2990wx is definitely not a gaming friendly CPU from what it seems. The added latency of those two extra CCX's will be a detriment. Now if AMD has a way to skip those extra cores for gaming apps it might not be that bad. I think this is why AMD is not stressing gaming on the new top end CPU.

It also is why the 2950x is out there for review as it should be at least as good, if not better, than the 1950x. I think though gamers especially should hold off on the 2990x preorders for now. I mean if CB R15 is a clue then @6GHz a 100% scale would be 7,000 score but it seems only 6,000 is there. That is only 85% of the overall performance per thread of the 1950x. Agreed not a load that should force base clocks but still the 2950x has to be better than that..

 

About Intel’s 10nm Process Lead


Intel has been getting a lot of heat for failing to get yields up to the point where they could ship 10nm in volume. We have seen analysts talk about how Intel is now behind TSMC / GlobalFoundries / Samsung due to the fact that the latter plan to roll out their 7nm nodes in the 2nd half of 2018, while “Intel will still be at 10nm by 2019”. The problem with all this of course, is the assumption that the commercial nomenclature of these processes means anything – spoiler: it does not.

It's a reason AMD push out 2990WX and 2950X first. Their flagship for content creators + their best chips for gamers (2950X). They cover both ends.

 

Ah it does, the issue may be that 10nm is not till the h2 of 2019 letting 7nm be out as well. The issue here then is both 7nm and 10nm become new processes. While AMD may not be ahead then it will hopefully be on par. Till then the 2950x and 2990x will give Intels current line up fits, especially since AMD will be backwards compatible.

Edit; Speculation is with 7nm that AMD may move to 16c per CCX complex. Since AMD said somewhere TR3 would only have 2 active CCX complex's like the 2950x this gives credit to that rumor so that a TR3 with 32c would still be offered. So TR3 will have the not only the 32 cores but higher clock speed and sans the extra latency of the 2 non direct memory linked CCX's.

What will Intel do if Epyc offers 7nm at 64c per chip or 128c on a 2P?

 

If you are into LN2;
https://www.tomshardware.com/news/amd-threadripper_2-vs-intel-core_x,37550.html

 

For which CPU, 2950X? Because 2990WX was over 7K at 5.1GHz.Also, the work station part has the same features for gaming our current chips do, which is only using local memory and loading it on only one die. Then it comes down to frequency and IPC. The W is for workstation, meaning they are trying to specifically split the HEDT platform for their offerings into gaming and consumer, which still use only two active dies, and commercial, which doesn't care about NUMA as much and still has a large boost with the two extra cores, because that extra latency means nothing. So the stratification of product stacks is actually pretty good, as well is their marketing if people listen to it.

LMFAO. It actually does. But you only take one data point on density calculating it in the way that Intel said to calculate, then do not factor in it cannot be as dense as they quoted with the changed pitch. And, if the products on 7nm hit the market first, they lost the process lead on time, which is losing the process lead. If the new 10nm is closer to a 12nm process, as that article stated, then it isn't as dense as quoted and they have no room for error here. Did you even read all the links I gave on density calculations? Those were all before we found out about Intel abandoning what their original 10nm was due to defects. SemiAccurate, behind the paywall, suggested it is even worse than said there. If true, their process is yank. Alone, they only expected 25% performance or 50% energy reduction on 10nm. Now, those numbers are further reduced with the changes making Intel's new 10nm closer to 12nm. Here are two resources you should read on density to see what this author said is ignorant. No one with intelligence thinks that 7nm is better just because of nomenclature. They are looking at these specific density calculations:
https://www.semiwiki.com/forum/content/6895-standard-node-trend.html
https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-versus-globalfoundries-leading-edge.html

I've got a doctor's appointment this morning, so sorry if slow to respond.

 

@Papusan - Here is an in depth discussion of Intel's coming density, etc. https://forums.anandtech.com/threads/in-depth-intels-10nm-was-definitely-not-too-ambitious.2548698/ (they practically have an article's worth in that first post and it is worth EVERYONE reading).

Now, the rumor of what may be ditched is the COAG. So, whenever you see that used in density calculations, take it out. That is why 10nm is closer to a 12nm.

As to 7nm, Intel didn't plan on that until after Tiger Lake, likely until the new uARCH replacing iCore around 2021, or the following year. So, I did factor that in. EUV won't be volume until 1H for TSMC, Intel sometime in 2H of 2019 (which was why some, including me, suspected Intel would just wait for EUV to release their 10nm, but seems instead they are ditching an element to achieve it instead), and GF and Samsung around the time frame that Intel is adopting it for volume manufacturing. TSMC is the largest fab in the world, so them getting it first and up and running is no surprise, although Samsung trailing others in adoption is surprising since they are the number 2 fab in size (then GF is 3 and Intel is 4 in size, after the IBM fab was given to GF).

Edit: I also forgot to mention that Intel's 7nm will either go against other fabs 5nm or 3nm. Both TSMC and GF are considering going directly to 3nm for the main process nodes because 5nm only offers around 10% performance and 15% energy reduction, if I remember correctly. So, if Intel's 7nm, which is supposedly 2.4-2.7X density over their planned 10nm (which was around 2.7X density over their 14nm, but that is now likely around 1.6-1.8X density), is on time, it will be going against almost identical node shrinks on density. So, there is a chance they pull it back, but how close everyone will be at that point should show Intel has actual competition for the foreseeable future.

In addition, I don't blame Intel for this, other than them not changing course. They have been waiting on EUV instead of DUV since 2015 or 2016. By that coming late, Intel had no choice but to wait, until they had to adapt 10nm to DUV lithography when all other fabs did so for 7nm. So Intel refined 14nm into 14nm+ and ++. This is why I said Intel doesn't have a bag of tricks to pull from.

 

The 1950x, and I will presume the 2950x, at 4.0 GHz gives 3500 in CB R15, 100% scale they should give 7,000 at 32 cores but the 2990wx does not scale 100%. It needed 1.1 GHz extra to reach 7,000 score. It seems the 2970wx and 2990wx will not have that extra umph for gaming, and this is ok but watch for it.

 

Well, it hits around 6000 at 4GHz, so 86%. Also, 5.1GHz got around 7618, so you could probably hit 7000 on a chiller at like 4.7 or so, potentially.

 

Right, that 86% is a 14% loss, as it is the Ryzen is not as fast in gaming for whatever workload and to loose another 14% would be bad to say the least. Again if tricks are used to keep the load away from the extra cores that are slowing it down overall this could help immensely. Again a wait and see what is up, but beware there could be inherent performance issues.

Edit; it appears there is a compatibility mode that will allow just the 2 CCX's for gaming and there is no UMA just NUMA.

 

In the news, NDA is keeping it quiet out there;
https://www.ultragamerz.com/amd-win...-but-loosing-mid-range-to-intel-9th-gen-cpus/
https://appuals.com/amd-threadripper-2990wx-world-record/

 

AMD news/rumors;
http://www.nbc-2.com/story/38694429/amds-ryzen-desktop-cpus-for-2019-may-double-the-core-count

For Monday 10:30am CT
https://www.twitch.tv/amd

 

@Deks - You remember how I keep referencing the frequency vs power curve. Here is how that works in action:


Just to give you more information for when you see the stats on process for normalized power and normalized frequency.

 

Processors

• AMD Ryzen 3 2200G HDR247 (ar)
  
• AMD Ryzen Threadripper 2950X and 2990WXComptoir du Hardware (fr)
  
• AMD Ryzen Threadripper 2950X and 2990WX ComputerBase (de)
  
• AMD Ryzen Threadripper 2950X Guru3D
  
• AMD Ryzen Threadripper 2990WX Guru3D
• AMD Ryzen Threadripper 2990WX Hexus.net
  
• AMD Ryzen Threadripper 2950X and 2990WX HotHardware
  
• AMD Ryzen Threadripper 2990WX Legit Reviews
  
• AMD Ryzen Threadripper 2950X Phoronix
  
• AMD Ryzen Threadripper 2990WX Phoronix
• AMD Ryzen Threadripper 2990WX Tech Report
  
• AMD Ryzen Threadripper 2990WX TechGage
  
• AMD Ryzen Threadripper 2950X and 2990WX Tom's Hardware

 

One thing to note, did you see the differences between Linux and Win 10 from Phoronix? (not making excuses) Turns out M$ is supposedly going to work with AMD on the OS to better utilize the extra cores. Might help Intel as well considering any improvement to the scheduler for HCC chips will likely benefit all companies, even though this will focus on NUMA and scheduling, most likely. But, more proof of how ****ty Windows is, lol.
https://www.phoronix.com/scan.php?page=article&item=2990wx-linux-windows&num=1

 

I think you will find with Epyc or TR1 , 2950x or even TR3 with just 2 CCX complex's that there is not an issue. I am pretty sure this has to do with the scheduler and the slower non direct memory access cores and their slower memory access etc.. So any fix will not be needed on the TR3 and since Intel doe not have non uniform cores they may not be effected either.

Thing is AMD should have been all over M$ from day one of hardware testing, not waiting for the world to find out the issue for them. Maybe now that AMD is becoming hardware popular they just need to stop recommending Windows 10!

 

Look again, some of the scores on those other processors are higher than on Windows 10. Geekbench has been known to be better optimized on Linux for a long time. You are wrong on this being ONLY the 2990WX. To be fair, on some tasks, same is true with Intel's lineup.

Edit: Just as one example, granted on blender, but few programs are on both systems (and the version issue, with this being 2.79a and windows being 2.79b, but likely closer than 2.78)


https://www.pcper.com/image/view/92878?return=node/70046 (In line with other reviews I've seen)

https://www.phoronix.com/scan.php?page=article&item=amd-tr2950x-linux&num=6

Edit 2: I also want to draw your attention to the 8700K, which with a custom compiled Linux Kernel actually had a significant percentage uplift like the 1950X and 2950X. This is saying, like I said, that the higher core count chips on both sides can benefit from M$ working on their scheduler (among other things that need addressed).

The benefit to the 2950X and 1950X is almost 20%. The benefit to the 8700K is also 20%. Shows that both can benefit from certain workloads on Linux over M$. But Phoronix reviewer also showed that on certain programs, Windows was optimized as well or slightly better than Linux. So, bit of a mixed bag there (which mostly benefits Linux on performance, but Windows wins on compatibility overall and takes the day on what people use).

 

No, I don't think it is a 2990wx only issue. Let me reiterate, The W10 TS has a lot of general issues but I think there are some specific 2990wx issues that need addressing. Of those specific fix's I do not that the TS will benefit Intel or say even the 2950x.

 

Added in...

• AMD Ryzen Threadripper 2950X Hexus.net
• AMD Ryzen Threadripper 2950X Tech Report
• AMD Ryzen Threadripper 2990WX KitGuru

---------------------------------------------

Full Intel 9th Gen Core Processor Lineup Specs Confirmed – Core i9-9900K 8 Core CPU Rocks A Soldered IHS, Aimed at High Performance Gaming and Overclocking PCs-Wccftech.com

 

http://cpu.zol.com.cn/696/6966410.html

That 9700K (@5.5Ghz) is some kind of beast with no HT. Imagine the 8/16 9900K is going to be a beast.

Multi: 1827cb

Single: 250cb

 

This means coming 8c/8t has about equal performance as stock clocked 6 cores 8700K. Stock 9900K more like +-2100cb

 

That is a good score. Also, that should avoid the problem of HT exposing memory as a security vulnerability. Too bad it is overclocked so high and we do not have an idea on cooling needed. But, either way, very impressive.

That may or may not be true. If you look below, you see almost all stock scores being around 1230. The outlier is a function of the setting where it boosted all cores to the max boost clock, which resulted in the 1430 score from OC3D and the resulting hoopla surrounding that.

Here, if we estimate the 5.5GHz all core boost is 4.6GHz, this is almost a 1GHz overclock. If assuming perfectly linear scaling, we could see 1528 at the stock all core boost, which is a 300 point increase over the stock 8700K. If we assume 100MHz higher and a 30% increase due to HT on the 9900K, we come to around 2029 (which is a rough estimate as the HT advantage is better leveraged in some programs and use cases, which explains leaks being closer to 2100, with this estimate still being in a reasonable range).

Just wanted to give a little more context and information on the expected performance, which is quite impressive and spanks the current offerings on AMD's side on mainstream to varying degrees (with the higher single core performance, and with the no-HT score at stock being close to a stock 1800X using SMT (1643 - https://www.techspot.com/review/1613-amd-ryzen-2700x-2600x/), and with the 9900K expected score spanking the stock 2700X 1771 score (same source)).

https://www.guru3d.com/news-story/intel-core-i7-8700k-benchmarks.html

https://www.techpowerup.com/236962/intel-core-i7-8700k-put-through-cinebench-r15#g236962-1

https://www.overclock3d.net/reviews/cpu_mainboard/intel_coffee_lake_i7_8700k_review/8

 

Hey man that just isn't correct. I'm sorry that is a 100% false statement and those BS benchmarks. A 100% stock 8700k running it's STOCK 4.3Ghz all core boost is around 1410-1430CB. The early reports of higher scores were in fact with MCE enabled by default in many manufacturer bios options. Mine was also on by default and I was like WTF are these voltages when I first booted up. That said 4.3Ghz is the all core turbo speed under 100% load on a stock 8700k without MCE and it will result in the scores I listed. Ask any 8700K owner.

https://imgur.com/a/G7KBeVh -- I managed an easy 1Ghz overclock on my all core (yes I know I'm showing 900Mhz here) 8700K with LM. I was able to post a 1.1Ghz overclock and CPU-Z validate it as well. A soldered 9700K should be able to get 900Mhz especially without HT. The problem I saw at least with my motherboard is that with clockspeed performance didn't scale well at the top end. So what we are seeing here at 5.55Ghz is probably top end and performance gave up around 100-200Mhz ago. That means a nice 5.3-5.4Ghz overclock would show very similar results at a much safer voltage.

 

You as me know many of the scores out there from review sites is crippled. And I don't even have fast ram in my laptop.

Stock 8700K. But less tuned means still +1430cb and more with faster ram.


With perfect scaling then add in 33% on top for stock 9900K = 2113cb

 

@Talon

And, you are now talking mods and overclocks. I was talking stock performance here. You seem to miss the point.

 

Come on... You can't trust reviews, man See... Guru3d.com posted even two different stock CBR-15 scores for 8700K in same review. https://www.guru3d.com/articles_pages/intel_core_i7_8700k_processor_review,27.html

On

 

HDTecnologia and PCEva have leaked the slides of AMD’s upcoming second-generation Ryzen PRO
AMD crippling one of their coming Ryzen PRO SKUs: The Pro 2700X will come with lower TDP and hence lower clock speed by 100-200 MHz compared to the non-Pro series. This most likely due they will offer 18-month stability guarantee.

AMD backs the processors with an 18-month stability guarantee (thanks to the company picking the chips from the highest-yield wafers), and promises that it will make the processors available for at least 24 months. The Ryzen PRO processors are backed by AMD Secure technologies and include a built-in AES 128-bit encryption engine.

Maybe pick some chips of their highest-yield wafers to push out 8 core 2800X to compete with Intel's coming i9-9900K? Or is 4.3GHz max boost all they is able to get?

 

Are you familiar with what market segment is being targeted? These aren't consumer chips! These are chips that go into large amounts of client systems deployed at large companies. For comparison, most companies buy client systems for deployment containing locked i5 chips. So, this is going against the "T" suffix, the "H" suffix, etc., like the i5+8500 (6C/6T; 4.1GHz), the i5-7600T (4C/4T; 3.7GHz), the i5-7600 (4C/4T; 4.1GHz), the i5-7500 (4C/4T; 4.1GHz), and the i5-7440HQ (4C/4T; 3.8GHz). So I think you are missing what the line is competing with entirely when discussing where this falls for use cases.

Edit: Here is the vPro lineup (most companies do not buy the "K" chips, and if they do, they do not overclock them). https://www.intel.com/content/www/us/en/products/processors/core/core-vpro.html

 

Market segment or not... I know the PRO processors stand out from regular Ryzen chips and is target for other market than gaming and consumers. The focus is on integrated security and encryption features. But If they offer same TDP and clock speed for all the other Pro models. They could do the same for the top dog as well. My meaning about tech doesn't have to be what other expect or want. Put TDP and clock speed down. It's called deterioration, decline, retraction, decay, falling off, stoppage, turnaround or whatever you will call it. As I pointed out... They will be sure that the chips can hold 18 months(I mentioned the 18-month stability guarantee in the post). And I don't talk about overclocking. As always... I can have my own meaning about tech in the same way as others.
And as I stated in same post (not all is about PRO)... They could use some of their best chips for a possible faster 8 core Ryzen 2800x. If it's possible.

 

You are directly comparing it to unreleased hardware that NO company would buy for mass deployment as regular machines for their employees to use, thereby creating an unrealistic benchmark for appearance that AMD is offering something not competitive for the market segment and use case. That is problematic. So, I gave you Intel's vPro site so that you could properly compare how those chips stack up against Intel's offerings, with the note on "k" series chips. When you compare the Ryzen Pro line to the vPro line, it actually is quite compelling for some use cases, although not for others. Also, the 9900K is a $450 chip. AMD has nothing at that price point. The 2700X Pro is running against the current 8700K chip, or 8700 chip, in the vPro Intel lineup. The 2600 Pro is running up against Intel's i5 offerings, except it has 6 cores and 12 threads, whereas Intel's offerings are lower than that for that price point, like no HT and usually 4C/4T with competitive speeds. There is more that goes into performance than that, and we both know it, but you get what I'm also saying, which is compare the products with what they are meant to compete against, on performance and price point in a professional/corporate setting.

 

How those Pro chips stack up against Intel's offerings shouldn't be a reason enough for AMD to have to lower TDP and clocks for mentioned chips. If the consumer chips is able to run reliable on 105w TDP I expect their best Pro chips from the highest-yield wafers should be able as well.

See... For example, the older Ryzen 7 PRO 1700x features the same base specs as the Ryzen 7 1700x, but with added security and management features for enterprise deployment.

The same should be for their newest 2nd gen Pro top dog chips also. Oh' well

 

Then you don't understand the market being targeted. Look at server chips, as an example. They use the chips from the center of the wafer, bin those chips, then have them clocked lower to meet TDP, etc. In the same way, except for the "k" chips, the no suffix, h, and t suffix chips are clocked significantly lower than the chips consumers normally buy. Why? Because the electric costs for office buildings are higher, so you want it to run at lower TDPs to draw fewer watts, which makes a huge difference on deployment for commercial/corporate systems.

You are conflating use cases and saying that it should be higher clocked, blah, blah, blah. That isn't what these chips are for. Review Intel's offerings at the link I posted, with their specs. You will see similar behavior. It is normal in the industry for this. You are making a mountain out of a mole hill due to not understanding who this is marketed to.

Also, as an aside, for the screenshots of the 9900K on air, I present to you a throwback article. https://www.guru3d.com/news-story/intel-core-i7-5775c-broadwel-reach-5-ghz-oc-on-air.html

 

Just to see Windows vs Linux performance on the 1950X with Linux using CFLAGS="-O3 -march=znver1" on a custom compiled 4.19-rc2 Kernel for Ubuntu cosmic cuttlefish:

Spoiler: GB3&4 Win vs Linux

GB3:

GB4


Also, this is with the CPU being clocked to 4GHz on Linux, instead of 4.05GHz to 4.1GHz on Windows.

 

I don’t talk about what Intel offer. And no need to explain how binning works.

You mean AMD has finally found the sweet spot this time bethween performance vs. TDP to fulfill what is acceptable vs. electricity costs? Aka 95W same as previous gen? Maybe they have. And/or maybe because AMD don't want stretch the limit/strap longer than this.

Did you see the link and picture I posted above? Last years maxed out Ryzen Pro chips had exactly same base specs. And for 2nd gen all the others outside the top dog is clocked equal as the consumer chips. I don't buy it. What would AMD do if Intel already had 105w chips? Of course they would try to max out what they had. This is more like how the market works Seems we have two different/opposite opinions on this and wont come any further.

 

No, for professional corporate machines, they design cooling around TDP. This means, when preparing drop in replacements, keeping TDPs for the pro series is more important than getting the highest clocks. So, keeping 95W is more important than allowing the extra 10W given to consumers, while also having good boost within the lower TDP. You are missing how the industry does these solutions.

Also, you compared it to the unreleased 9900K, then change to not wanting to talk about Intel's offerings after I have the vPro context?

So, I'm not saying that and you are thinking wrong on the topic for the pro series.

Sent from my SM-G900P using Tapatalk

 

And you don’t see my point... It’s a change vs. last year. Or better say they allowed max TDP and equal clocks for last gen Chips (1700x and 1700X Pro was equal). This means AMD think or better say mean 95W is enough although they offer cooling intended for higher TDP (if they offer same cooler as for the consumer chips). They have found the sweet spot or meant 95w was enough.

Regarding you talk in 9900K chips into this. This has nothing with the Pro chips. I put in 9900K into my post due a rhetoric question that AMD could use some better binned chips for a possible 2800x. Nothing more than that.

 

Last year, the TDP was 95W. These are usually within OEM builds from Lenovo, Dell, HP, etc., which don't use wraith but their own solution. So keeping TDP between last year and this year is more important than keeping TDP between the commercial and consumer product lines, especially since AMD allows upgrading to newer chips, whereas Intel requires buying the same chip or a new system. Since these chips can drop into last year's MB, ignoring Intel's quirkiness since skylake of locking features on z170 to z270 for optane, then coffee with z270 and z370, and then coffee refresh, which is not their standard practice, keeping TDP to fit OEM and ODM spec matters more.

So, I get your point, and I reject it based on how the industry works.

As to the rhetorical reference, they don't need to have a 2800X. The 1800X was overpriced and didn't sell well. The Intel chip, still unreleased, comes out 6-7 months after their current gen chips, and those binned chips do better going to the TR2 series which have better sales. Their new chips for mainstream are likely 6 months after Intel's release, give or take a couple months. So there is no need to introduce that product at all.

Sent from my SM-G900P using Tapatalk