AMD's Ryzen CPUs (Ryzen/TR/Epyc) & Vega/Polaris/Navi GPUs

I'm happy with pcie 3 for another 2-3 years.

 

two sides of the same coin. some people were saying stuff like win 1903 build is amazing etc and what not, it barely did anything to improve window as a whole and it is sitll a junk OS that doesnt take full advantage of the hardware. there are fanboys everywhere just need to know when to take salt from them.

we enthusiasts especially in laptop are wanting to keep using same mobo if it worked. look at intel's chip for 4 generation people still try to get 9900k and 8700k to work with z170 chipset.

power delivery maybe one thing, PCIE might be another. i'd be happy to take PCIe 4 on the cloest PCIE slot to CPU. everything else for pcie 3.0 no problem.

 

Windows 7 FTW!! - It looks like the Ryzen 3 idle voltage due to being rapidly over polled in Windows 10 works fine in Windows 7:

Ryzen 3000 Idle Voltage is NORMAL on Windows 7!
Submitted 2 hours ago by click4dylan 50 comments
https://www.reddit.com/r/Amd/comments/cfbje2/ryzen_3000_idle_voltage_is_normal_on_windows_7/

click4dylan S 4 points an hour ago
EDIT: the 4 cores 8 threads is fixed. It was a setting in MSCONFIG that was setting 8 cpus max. It is now showing the full 8 cores 16 threads
https://i.imgur.com/at8eIqZ.png permalink

20150614 3 points 2 hours ago
"Hopefully this is right. It would mean the voltage kerfuffle is just software based and could be solved by a simple update."

Admixues 3 points 2 hours ago
"probably something to do with 1903 pooling zen every 1-2ms for their boost algorithm, AMD will work it out."

 

Unless you're doing Quad SLI you are fine with PCIe 3.0 x16. I'm thinking what might Apple be charging for PCIe 4 mobo in their Money grater?

OLD is gold!

 

Techpowerup has a survey on which Ryzen CPU you are exited about. The 3700x is on top, with the next 3 a close 2nd place: 3900x, 3950x, 3600:
https://www.techpowerup.com/forums/...ryzen-chip-are-you-most-excited-about.256444/

Vote for this poll on the frontpage Total voters 27,853

 

As I suspected not a useful feature for high fps competitive gaming. Want better input lag? Get a faster GPU/CPU combo that deliver high fps.

 

It's not like the Intel socket upgrades for more power thing, although to be fair the higher OC'd follow on CPU's did draw a lot more power and there could be old motherboards that don't fully support 100% performance - while some might.

AMD is trying to save you and the rest of it's customers from disappointment wasting time and money while trying to get high performance applications of PCIE 4.0 devices to work reliably on PCIE 3.0 designs uncertified for PCIE 4.0 applications.

https://www.techpowerup.com/257515/...4-on-its-amd-400-series-motherboards#comments

TheLostSwede
"AMD never promised it would work on older boards.
Besides, this could cause data loss if the board is a borderline case.
I think this was a responsible thing to do, as devices have to pass certification for PCIe. If you don't meet the strict standards, then you don't pass, there are no borderline cases. Posted on Jul 19th 2019, 2:52"

tony359
"I have to say all of this (BIOS ROM limitations, new MAX motherboards with larger chips, CPU's not being supported anymore, RAID support going and then coming back, PCIEx-4 only on X570 then also on older mobos) is very confusion and I feel it shows a lack of communication between AMD and their partners - or even worse, some incompetence in planning all of this.

THAT BEING SAID, it is nice to see AMD and partners trying to give existing users the best they can. For so many years Intel could not be bothered and just changed the chipset at every generation - that's easier, no need to plan, no need to communicate. Posted on Jul 19th 2019, 3:25"

Deathy
"i dont think amd is to happy about this as they stated pcie gen 4 cut not work on 400 series chipsæt/boards"

They never said that. They said they couldn't guarantuee PCIe 4.0 running on every older chipset out there and didn't want the confusion of some motherboard supporting it on B450 while another B450 or even X470 not supporting it. These half-truths are how fake news get spawned.

AMD as quoted by Anandtech:
"Pre-X570 boards will not support PCIe Gen 4. There's no guarantee that older motherboards can reliably run the more stringent signaling requirements of Gen4, and we simply cannot have a mix of "yes, no, maybe" in the market for all the older motherboards. The potential for confusion is too high. When final BIOSes are released for 3rd Gen Ryzen (AGESA 1000+), Gen4 will not be an option anymore. We wish we could've enabled this backwards, but the risk is too great."

Edit: And AMDs response so far has been (roughly) "we will remove this feature from future AGESA versions, so unless you want to stick with the outdated AGESA, you won't get PCIe 4.0 on old generation chipsets". We'll see who ultimately wins this battle. Posted on Jul 19th 2019, 4:13"

Except for some high speed PCIE4.0 data transfers to/from the GPU when processing rendering, and high speed PCIE4.0 SATA controllers - so far, there might be more PCIE 4.0 devices that will benefit from the 2x higher bandwidth, PCIE 3.0 will be enough for now.

You could wait for AMD's follow on AM5 socket to get full PCI4.0 bandwidth and while on PCIE 3.0 set everything to PCIE 3.0 in the BIOS, - even if the BIOS defaults to 4.0 or allows 4.0 operation.

If you notice less than expected maximum PCIE 4.0 performance from your PCIE 4.0 M.2 SSD on your PCIE 3.0 motherboard, you might be experiencing uncorrected data errors. So even if it looks like it's working, check your Windows logs for PCIE errors - and compare a good copy vs a transferred copy on a few different large files to make sure you aren't having unnoticed data corruption.

 

An AMD Radeon 5700 @ $350 GPU outperforming and keeping up with the Nvidia 2080 @ $700 GPU in various rendering tasks, more extensive tests than previously shown, with details on AMD involvement in improving performance and encoder support.

Resolve, Premiere & Handbrake TESTED: AMD RX 5700 vs Nvidia RTX 2080 - WHY IS THIS SO FAST?!
EposVox
Published on Jul 17, 2019
I wasn't satisfied with my original test runs of the AMD RX5700 and RX 5700XT and wanted to dig deeper and see how it really stacks up to the Nvidia RTX 2080. And wow... the results are impressive. In this video I benchmark and test AMD's new Navi graphics card, the RX 5700 and compare it to Nvidia's RTX 2080 for projects in Adobe Premiere Pro CC, BlackMagic DaVinci Resolve (16 beta 5) including BlackMagic RAW tests, playback tests, render time tests. THIS IS A VIDEO PRODUCTION SHOWDOWN: Video editing, video encoding and NLEs tested for the AMD RX 5700 VS NVIDIA RTX 2080 Puget Systems Radeon VII benchmarking & review: https://www.pugetsystems.com/labs/art... Check out this playlist for more AMD Navi GPU coverage: https://www.youtube.com/playlist?list...


According to leaks so far it looks like the 2080S isn't going to be much more than "a tad faster" than the 2080. So the 5700/5700XT AIB boards could press the advantage back to AMD even further in price / performance in productivity tasks as above:

The NVIDIA GeForce RTX 2080 SUPER falls well short of the RTX 2080 Ti in gaming and synthetic benchmarks; only just beats the RTX 2080
NVIDIA has discontinued the GeForce RTX 2080, having replaced it with its SUPER successor. However, some synthetic and gaming benchmarks show that the RTX 2080 SUPER offers hardly any performance improvements over its predecessor, with it significantly trailing the RTX 2080 Ti too.
https://www.notebookcheck.net/The-N...ks-only-just-beats-the-RTX-2080.427795.0.html

 

i know what AMD is trying to do frankly i dont really care all that much. we laptop enthusiasts been scewed around for way too long, anything on desktop is like what we dreamed of happening on laptop and more lol. which was why i was okay as long as I get to use PCIe 4 SSD with older mobo, i do not mind GPU on PCIe 3.0 not like its gonna max out anyway.

and yes i am definitely waiting to see zen3 though imo i think the overclock will probably be another disappointment. TSMC isnt like intel to stay on a node for so long they are keep pushing forward, meaning new node means we likely won't see much optimization. intel's 14nm lasted 5 yrs, into 6th yr next hence the high clocks. TSMC on the other hand will have 7nm+ on next then move onto 5nm.

if we want high overclock + great IPC, the only real option here is intel when it hits icelake desktop or tiger lake.

thanks to AMD i can now wait while knowing intel has to give their best for next 3-4 yrs instead of back when we were at haswell, stuck on 4 cores for next 4-5 yrs.

 

Maybe it can still happen, but I wouldn't count on it. AMD said future AGESA updates won't have PCIE 4.0 support for PCIE 3.0 motherboards.

Maybe shop around and find a nice x570 motherboard and sell your x470 motherboard to someone that needs it? Then 2 people have what they need.

It might not be a good idea to wait for the A520 / B550 motherboards as I haven't heard whether they will have PCIE 4.0 or not. Those motherboards are usually budget and less cost implementations and might skip PCIE 4.0 to save $.

Overclocking isn't really a thing anymore. Overclocking is an anomaly of the past, before AI / Auto-Tune features make sure everyone gets the full performance of their CPU / GPU.

Any smidgen of performance left will be at the edge of stability and thermal shutdown, well outside the range of most people to take advantage of. That's why you are seeing "mainstream" YT's doing LN2 OCing features. That's all that is left to explore.

AMD Ryzen leaves left very little on the table at launch for OC'ing and the automation improves over time such that even that little sliver of OC range is eventually given to every user.

I would expect Ryzen 4 / Zen 3 to have even less OC potential at launch, giving even more performance to everyone instead of wasting it like Intel.

Intel is a slow dinosaur, and likely to stay so for quite some time to come. Why follow the dinosaur lumbering down it's slow path to irrelevance?

What good is it to sit around doing nothing waiting for Intel to get it's act together and offer something compelling?

 

yea some bios mod with maybe older AGESA probably needed. i guess the person would have to weigh benefits of each and see what they'd like more. performance on old AGESA isnt exactly bad.

well thats another thing to look at. with all these auto tune OC such as PBO/maximizer, if these software does give VERY good performance balance with VERY little work involved i'd probably use that.

but right now ASUS got that CCX OC utility which is very attractive, also intel chips with the right mobo support has capability of adjusting individual core voltage, all of those are good. if these new auto tune software can do those then i'd imagine replaces most of OC.

also sometimes these OC tool use like 1.4v man i just can't have that, 1.33 is prob most i'll go.

we just have to see how intel responds to AMD. if they'll accept defeat for next 2 yrs and have a junk of chip with almost no IPC improvement then AMD is the choice. you also got to know I am looking out for what benefits my use case the most, which is like 80% ST performance. of course in due time i'll find replacement software but thats also a lot of effort when a lot of my software are outdated.

I really want TSMC to tune their 7nm+ so we can see a 4.7ghz OC with decent voltage. seriously doubt that will happen but i still hope for it. imagine 4.7ghz on zen3, with higher IPC, 16-24 cores CPU, lower memory & inter-core latency, it'll destroy comet lake 10cores on 14nm+++

now if the above doesnt happen, then intel might come up with 12 cores on 10nm+ using as much power as their current 8 cores along with 5-10% IPC boost while allowing high OC, just have to wait man

 

Yup, Asus's first automatic tuning as well as others, and even today, overvolt, which isn't good for temps, but so far AMD seems to be making it work.

The voltage readings on AMD Ryzen 3 seems to be a victim of being elevated from too fast sampling causing the CPU to think it needs to boost - raising CPU voltage - then dropping down - up / down unnecessarily. AMD should get that fixed as well as getting Microsoft and other tools makers to change their polling / sampling methods.

The best AI / autotune software will allow us to tweak the weighting of voltage, current, frequency boost / deboost curves - along with more custom assignment of values per application.

AMD is serious about making their software reflect users needs and desires, so it's important to participate by reporting problems, fixes, and suggestions for improvements. AMD is listening, and acting on those requests.

AMD is choosing their silicon properties carefully, and I think they know breaking 5ghz is a "thing" that needs to happen at some point. It's hard to argue with the lower clocks on AMD CPU's outperforming the higher clocks on the Intel parts in multi-threading and matching performance in single threaded performance.

AMD's performance against Intel and Nvidia being so close yet not surpassing it makes me sometimes wonder if AMD is holding back to "help out their buddies" at Intel and Nvidia.

 

You guys may be thinking of this all wrong. A CPU at 4.0 GHz has a rated IPC, meaning at that speed their is a certain amount of active and idle clock cycles. at a raised IPC at the same frequency you then have to raise the number of active clock cycles. Since activity, and not idle, generate heat etc.....

Edit; so single thread 5 GHz may be reachable but full load will be much harder to get a boost.

 

Interesting point of view and rebuttal from De8auer on AMD/Intel fanboys.

I never thought De8auer was anti AMD or pro Intel. He always came across as strictly, scientifically speaking "It is what it is."

 

I do not think of his video's as anti AMD. for the OC hobby zen 3 is horrible.

 

Unless you want to see AMD staying on a specific node for longer periods of time, very high clocks will likely take a back seat with priority being placed on higher IPC and uArch optimizations.
Given that Zen 2 is made on first iteration of 7nm node, I think the clocks are pretty good (in regards to boost) - especially when you factor in how much the clock projections fell (from the initial 'we project over 5GHz on 7nm').

The higher clocks would come from a better version of the node, which would be 7nm+ (EUV)... but this is bound to be a bump up at best with some further uArch optimizations, after which, they will probably move onto 6nm or 5nm (depending on how AMD wants to play it post Zen 3).

One of the primary reasons Intel managed to achieve 5GhZ was due to staying on 14nm for so long. However, their upcoming 7nm (when it finally hits production) might not achieve such high clocks (and they had severe problems with 10nm already).

AMD managed to close the gap and surpass Intel in many (non-gaming) tasks with lower clocks and higher IPC.
This is clearly evident from third party reviews.
The reviewers also need to adjust the benchmarks to take into account the fact optimisations for Intel itself (which would mean there's still performance to be gained on AMD end if software devs properly optimized for them).

 

currently at how AMD is pacing their improvement on CPU, we'd need at least until zen4 or possibly 5 IPC/Arch combined to overcome intel's IPC + high clocks combined.

15% up in density and 10% up in performance only in optimized frequency range, this resulted in only 100-200mhz at most when CPU is OC to the limit and thats improvement from zen to zen+. TSMC MAY have a better node, going by what ANANDTECH suggest here:

https://www.anandtech.com/show/1422...echnology-7-nm-with-higher-transistor-density

7nm+ or 6nm we really got nothing much. not even a 10% performance improvement. we likely will see zen 3 stuck on 4.3ghz, pretty much what we have now, hence waiting for zen4/5 on 5nm. 7nm+ -> 5nm ~15% performance improvement with 10% power efficiency, similar to that of GF node improvement so we might see another 200mhz squeeze.

 

Right now there are 3900x's boosting to 4.6ghz and sustaining 4.4ghz in reliable testing, so another 200mhz would be 4.8ghz single core and 4.6ghz in reliable sustained use.

Not quite 5.0ghz, but edging closer. Perhaps there is a process node that can give 5.0ghz for fewer cores/threads, up to 6c/12t only, with thermals running too high beyond that to fit in the AM socket, but the TR socket could be made able to run 8c/16t or higher core counts at 5.0ghz+.

And, that might be the trade off. Better power / thermals for high core count "lower frequency" CPU's vs high power / high thermals + high frequency with 1/2 the core count.

Once AMD hits "enough cores" the next goal could be to increase frequency. One way would be to cut down the core count and tilt the process node options toward performance and away from the power savings needed for high core count CPU's.

It will be interesting to see if AMD comes out with a line of lower core count higher top frequency CPU's. It depends on how high they can reach as to whether they would be "special" enough to sustain them as an independent product line.

Or we may be "stuck" with high core count sub-5ghz CPU's for a long time from both AMD and Intel. Intel's 10nm CPU's aren't reaching very high clocks or high core counts.

 

4.4ghz isnt good enough if it requires over 1.35v

my bet is on zen 4 that we maybe able to get a 4.6-4.7ghz all core boost at below 1.35v. this also means i'll need to purchase from silicon lottery of the best zen4 available to hit that goal. this pair up with any IPC bonuses going from zen3 --> zen4 would be excellent.

if the above comes true, i could ditch a 5.2-5.3 OC from intel for zen because of stronger IPC and better power efficiency. all in due time tho

 

TR3 leaks;

 

Finally getting the news about ThreadRipper 3

AMD Threadripper 3: Everything we know so far
AMD Threadripper 3000 CPUs are nearly here, and they could be amazingly powerful
By Jon Martindale — Posted on July 22, 2019 9:45AM PST
https://www.digitaltrends.com/computing/amd-threadripper-3-news-rumors-price-release-date-specs/

 

I think Paul is correct and quad channel memory will be the limit. Speculation but I think the 3950x may be limited, or not scale up too 100%, with dual channel. On the same note this may be true of the 32 through 64 core Threadripper too. Especially this could be true in full load of memory speed limited older boards.

Edit; I am interested in the 32 core if the ccx is a 2+2 configuration direct to the memory channels. If it is a 1+1+1+1 it may have limitations like the current 2990wx, not interested at all.

 

This. Yes.

Maybe I'm falling for AMD marketing, but it really looks to me like AMD's attitude here is that doing right by the customer is doing right by the company. Ryzen processors aren't locked; they don't overclock well because they're designed (by a combination of the chipset, the UEFI, and the chip) to always run as fast as possible given the thermal and power constraints. To me, that's a very customer-friendly approach.

 

This is an interesting approach to power control reducing thermals, with an eye toward keeping performance up this could be useful - and actually if you don't care about giving up some performance this could help reduce temps / fans when Ryzen 3 gets to laptops (if they all such levels of control):

AMD Ryzen 3000 Undervolting Offset vs. Override | Vcore Voltage
Gamers Nexus
Published on Jul 22, 2019
In this content, we benchmark the Gigabyte X570 Master and MSI X570 Godlike to demonstrate undervolting behavior on the AMD R9 3900X & R5 3600. Undervolting is a valid concept that can net real performance improvements in GPUs and CPUs alike. Those improvements are typically in the form of reduced power consumption (watts * amps = power), reduced heat load as a result, and increased frequency boosting headroom under Precision Boost 2 parameters. Today's video looks at valid and invalid undervolting results between two boards and two CPUs, showing how to know when your undervolt is working (or not). We also demonstrate offset vs. override voltages for this test.


The original video is pulled, but here's his update / redo / restatement of his undervolting:

Ryzen 3900X Performance @ 1.00V - I Was Wrong...
Optimum Tech
Published on Jul 16, 2019


@ole!!! - check out his low voltage results, little to no FPS loss with extremely low voltage.

AMD Ryzen 3 seems to be adjusting to compensate for too low vcore - allowing it to continue running with reduced performance - somewhere in between normal vs too low might be an interesting range to experiment with now and later after AMD updates the firmware.

 

Th clocks for this gen chips is A joke! I mean OC this chips tragic

 

OC is not tragic, it does so differently than Intel. With a higher IPC at the same clock as Intel the fastest clocks just do not go as high. percentage wise they go well beyond base clock but do so automatically. If the 9900k were to do as well and hit 5.3 GHz automatically you would complain about Intel?

Remember too 7nm is just out, give it some time to mature.

 

agreed. people tend to focus too much on the naked Ghz numbers. sure, having 8.72 Ghz set in an AMD FX-8370 sounds TOTALLY AWESOME! but is the naked number worth the crappy IPC of that sku?

 

I want both​

 

We are all greedy but because of physics we can only have so much.

 

Yes, we are greedy and for right now, we can only have so much.

But physics is just an idea and ideas are malleable.

Remember, there is no spoon.

What is impossible today will be possible tomorrow.

 

Why? If you get the performance, why does the clock matter?

 

One thing I'm wondering about -- time will tell -- is whether process maturity and tweaks will result in future Ryzen 2 chips running faster in practice than the earliest ones. Since the chip dynamically adjusts its clock based on thermals and power, process improvements that result in better heat dissipation or less power consumption might result in chips running faster.

Of course, if that happens it's possible AMD will release new SKUs and just put the better chips in the new bins.

 

Because. Higher clocks = higher performance.

 

That is only true as a comparison within the same process.

 

cuz then u get MOAR performance haha

i guess its about the principle of being able squeeze out more performance vs. stock, especially when its free and requires some knowledge and/or tweaking. its just satifying, i get it.

now with ryzen the average joe basically gets the maximum performance automatically out of the box, exactly what he/she paid for. but any enthusiasts willing to go the extra mile dont really get the satisfaction of significant performance improvements.

so yeah, the actual performance is objectively the same, but the satisfaction of getting MOAR due to one's own tweaks is veeeery subjective!

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

B3cuz th3n u g1t t0 b 1337 h4x0rz sh0w1ng ur m4d sk1llz!

It really sucks when anyone who can turn on a computer get just as much performance as you.

 

It's actually awesome that everyone can get full performance out of the box - it's what I have been hoping to see implemented for 30+ years since I discovered OC'ing and how much performance was left unavailable for everyone.

OC'ing and tuning has been a constant distraction and a huge waste of time that could have been spent more productively / enjoyably. It's such a basic automatic tuning that we should have had as standard many years ago.

I blame Intel for the most part as being the industry leader. Intel should have captured that "lost" performance and made it available to everyone long ago.

 

Exactly. Overclocking (sans exotics like LN2, or risking failures/data corruption, or putting your chip at risk from thermals) means that the vendor left performance on the table. AMD has (correctly, in your and my view) decided not to do that.

 

What I've been waiting for, the HU OC Navi 5700 with waterblock, here's the EK + 5700K @ 18% OC, hopefully the AIB boards will stretch out the lead even more:

Radeon RX 5700 XT Overclocked, RTX 2080-Like Performance?
Hardware Unboxed
Published on Jul 24, 2019
Radeon RX 5700 XT, Overclocked to The Limits


TimmyJoe's new Anniversary 50th Edition 5700XT is running too hot stock - low fan curve out of the box - and it's started artifacting out of the box. Needs a better cooler at those frequencies / tuning... but it was running as fast as his Radeon VII... still a fun video, especially the beginning.

My 5700 XT Anniversary died after 3 hours of gaming...
Timmy Joe PC Tech
Published on Jul 19, 2019
I bought the AMD Radeon RX 5700 XT to be the first to review it and it started to artifact at stock clocks after only 3 hours of use...

 

It is indeed tragic.

Both Clocks and OC, latter is dead. Basically these are so maxed out just like a Tesla electric car or any "it just works" type of stuff, ak.a Apple. You can't tune them or modify them. Except to pray that all planets are aligned for that precious ST boost for the little bit of juice that occurs in the sea of confusion (more later) with their retarded terminologies - PPT, PB, PBO (No significant improvement), Auto OCand what not bu removing all OC and only expect OOTB max, and a Voltage curve type autoboosting, as GN Steve says it's more of a like the Ngreedia's stupid Pascal Temp scaling with less control. No fun or enthusiasm anymore as AMD is eating away all it's OC headroom and setting a CPU SKU at every pricepoint for the bins to beat Intel. And also to note how the Intel CPU clocks stay rock solid until it hits the CPU core TJmax with Power limitations removed, no drama.



If you tinker too much, It'll just end up nerfing the CPU since the PBO takes care well as per the reviews. To add salt, you can't even use HWInfo or CPU-Z or any tool to monitor the Ryzen 3000 chips because their Ryzen Master software only is capable of showing the proper appropriate information, forget TS sadly, TS helped me achieve more clocks/voltage and total power and optimize my CPU, It's just brilliant engineering by Unclewebb. There is high dependency of Temp (Alr posted but again) and then Voltage dependent scaling on these, yes ofc you can tweak them but by how much ? also there is high voltage going on with these chips. I think it's more of a reason (My guess) due to the TSMC process not being optimized for these beastly x86 desktop parts, vs that stupid A series processors (stupid because they are relegated to be used only for a Social media withdrawl fixation toy vs Android / powerful OS) also they pay a tonne for R&D to TSMC.



Add the continuous OS updates (This is an unfortunate byproduct, Intel optimized SW until now and all but at the expense of the latest Win10 1903 ? that's like choosing the worse of cancer varieties) then we have the upcoming updates and expect another OS optimization maybe (?) when 4000 hits. That means LTSC won't cut it for X570 and Ryzen 3000. This is more of a niche issue but it is an issue nonetheless due to the perpetual alpha OS of Win10, 6Mo EOL and 18Mo support max for the SAC releases so it will never be stable. Period. However the software scaling across the SMT/HT will be interesting cadence...The performance of the Ryzen 3000 is undeniably excellent and the value as well ofc still not a king in gaming, 9900K still is the king followed by 9700K (3600 vs 3600x vs 3700 vs 3700x vs 3800x all are ~6% off lol). They improved the IMC this round but as with the Fclk and Uclk this time and the IF all go in hands, there is automatic override of the IF frequency if DRAM speed goes over 3600/3733MHz. IIRC there is one step there instead of leaving it auto, one can manually adjust it to maximize further perhaps (Need GN video on it) but that 4000MHz "Plug it, just works" is a hogwash as it impacts the Perf severely, we might not see it in some loads just because of the eDRAM/L3/Gamecache on the Zen 2 parts. SKL Mem scaling & Z390 Memory scaling.



Then we have this X570 thing, the plastic shrouds all over the place for the active cooling chipset for the Matisse repurposed GF 14nm dies from EPYC/Rome (?) except one chipset, X570 Aorus Extreme, shame that ASUS Crosshair VIII (Even Formula) didn't went for this approach. And ofc we miss the god-tier EVGA in action due to their business partnerships.



Finally, 2 more things that I wanted to speak about..

One - No damn datasheet on the Ryzen processors or the chipsets or any. The AMD site has all old junk documentation. I wonder these PPT and PBO and PB all are mentioned from the reviewers guide not any datasheet, With Intel you can have a datasheet dating all the way back and gain a TON of knowledge on the excellence of technology of the CPUs to chipsets and even gimp tech (TVB, cTDP) just off them. It's a shame. AMD as a company which is not divested into failures like Intel's McAfee / smartphone Atom division didn't get a documentation done (Except for 2017's Ryzen Master and OC guide, perhaps it's a v1 of the reviewers guide) as a company which is more into core processor business than Intel, esp given the size. Inexcusable.

Tw0 - Intel is sad. They had this 22nm technology showcased with FinFet Ivy Bridge where all rivals were scrambling and ultimately got at 14nm and then lost their total leadership totally and fell off grace due to zero competition, this is a major issue with these gigantic corporations. Add the political drama angle then we have the 10nm disaster, all line up with their ambitious flops at other industries instead of the bread and butter, prestigious core business. The long term path whoever laid the Core uArch was excellent unlike the investor driven greed, both Apple and Intel have billions in bank. FF to now, no Ring bus anymore past 10C and the Process depedency, ultimately clocks. I think the age of 5GHz will be soon over, once Intel's 14nm++ the mega optimized node is done building over Foveros and "Wider" uarch won't allow them to scale clocks like this era. Unless Intel innovates again like Sandy Bridge era (Jim Keller magic ? like the Israel Core team..) And I hope Intel keeps this OC game going forth, the tuning and optimizing, tinkering is just fantastic, esp they still mention Moore's law in place.. just like the sonic bliss of the old Naturally Aspirated engines with stick shifts vs the latest era of dumpyard automatic/autopilot/hybrid joke small engines with turbos what not.

See below for the downgrade on F1, from Driver skill to automation and ofc the Symphony of sound from the golden era.



Still waiting on the news of the Cometlake / Rocketlake and Z390, I want a beast of the CPU which runs old OSes (Win7 / 8.1) and games (New political agenda driven games are ruining franchises add the MTX disaster with the Always OL games.. multiply with the stupid Cloud BS, they don't want us to own anything) without any hassles and games at highest FPS but also a bit of Platform life, the X570 refresh in 2020 might not happen, If it happens then it's really bad, I think these overbuilt VRMs and Power connectors are the reason for that another cycle of Ryzen, the chipset refresh for Zen 4000 will be horrible in market, since we are alr seeing the X470 re-launch with extra BIOS memory chips (Only reason being the sales the companies are expecting off the old gen SKUs vs the X570) So X670 / or whatever refresh doesn't make sense to me at all, buying a top end SKU of Mobo now makes sense for a new buyer.

One last thing, TR3000, I do not think it will have high boost frequencies, the Mainstream CPUs are themselves hitting hard walls at the current temps/voltage. That CPU will be a beast but expecting uber high clocks is not possible, and also factor the power consumption/Heat too.

 

Aye, there's the rub, no? Is it about competition/bragging rights or getting good performance?

 

For all you "fans" of active cooling on motherboards, here's a review of x570 motherboards "noise" @ idle / load:

X570 Chipset Fan Noise Comparison - Who makes the quietest X570 motherboards?
Techtesters
Published on Jul 27, 2019

Spoiler: Notes / Clarification:

- To keep this video short and to the point I didn't go into every test detail, so just ask if you're interested!
- Note that during this test all chipsets had mild airflow from the GPU and CPU. It is very much possible to heat up the chipset beyond the ~50-55 we typically see if you use a poor airflow case.
- I didn't mention it later in the video, but the Aorus ITX boards results are correct and differ from it's ATX counterparts. It runs a few degrees warmer, but still plenty cool it's basically quiet all the time. Just remember this board doesn't have a Gen 4 expansion slot on the chipset, so the only load there is the Gen4 SSD.
- If you own a Gigabyte board and don't see the fan control, update your bios to the latest version.
- If you want me to test more boards, let me know! It takes a long time to test this, so I'll only do it if you guys show me you care about it!
- Questions? Just ask

@ idle a number of boards have a fan stop / off setting for noiseless operation:


With only about a 5c difference at maximum in idle vs load there seems to be not much need for an active fan cooling with current PCIE 4.0 usage...

 

AMD Can't Satisfy Red Hot Ryzen 9 3900X Demand Sending Prices Skyrocketing
Read more at https://hothardware.com/news/amd-ryzen-9-3900x-zen-2-out-of-stock

“The Ryzen 9 3900X supply woes, however, have us wondering if we'll see a repeat when the 16-core, 32-thread Ryzen 9 3950X arrives in a couple of months”

 

I saw price gouging early on, this is not news to me. yes it would be nice if supply was better.

 

Yeah I know, sir. We saw same happened with Intel. Many complained this was only a Intel problem from day 1 when they released 8700K. Not as easy. But this is not good for customers regardless what manufacturers products they prefer/want.

 

Well, you have the overclocking/hardware enthusiast market that want to overclock over spec and have come to expect that given margin of headroom Intel tends to bake into their chips for later binning/selling and then you
have Joe Basic User (JBU) who wants the CPU to work out of the box as is and doesn't care about tweaking.

AMD is shooting for max performance right out of the box. For JBU, this is a godsend as it means they won't have to worry about tweaking for optimal performance for the most part and will get effectively all the CPU can deliver from
box to system. This isn't a bad thing.

For overclocking enthusiasts, this really diminishes the overclocking margins to work with to get a major performance bump out of the chip and that's a bummer. But I tend to think the overclocking segment of the market is not
AMDs target or concern. To a smaller extent, I don't think It is Intel's either. It just happens to be a byproduct of how Intel develops and delivers their CPUs (for now).

 

The difference is that the regular retail outlets aren't raising the prices of the AMD 3900x even though they are bouncing in and out of stock - the $499 price is holding nicely.

The difference is that the Intel 9900K price was / is inflated at regular retail outlets too - even when they are in stock on a regular basis.

Now of course due to AMD stomping on Intel's value proposition + beating Intel's straight-line performance the 9900k isn't selling well so the price has dropped.

It's funny - Microcenter still shows the Intel 9900k normal "inflated" non-discounted price as $699! - "on sale for $449" - Save $250!! And still people are waiting for the 3900x to come back in stock...:


While the AMD 3900x still shows the MSRP price of $499 + you can get $50 off on a motherboard+CPU bundle:


The same 3900x retail MSRP of $499 is maintained at the other retail outlets mentioned in the article, they all show the retail msrp price even though they show out of stock.

The only price gouging for the AMD 3900x is in the grey market, which happens with most all popular items that are in demand in not in stock at retail. Intel's price gouging was universal, at all retail outlets.

AMD + Retail partners have held the MSRP pricing, so far. I checked all of the retailers that article listed and they all still have the MSRP pricing even if out of stock.

 

Intel had problems deliver enough chips. And if AMD was/or will come in same situation... The AMD buyers will be screwed as well. The market is as this. Want one when there isn’t many options... You have to pay. That’s the sad truth.

Edit. Of course included the 25% forced Norwegian tax

 

That "Userbenchmark" site has really gone off the rails - panicking due to Intel CPU's looking bad across the board using their previous weighting of single vs multi-threaded performance based scores they are fudging their scores to make higher core count CPU's look worse than higher frequency fewer core CPU's - making for some funny i3 win's over i7 Intel comparisons along with the intended fudging of the new AMD CPU scores - trying to induce a reduction of AMD CPU's lead over Intel:

Discussing UserBenchmark's Dodgy CPU Weighting Changes
Hardware Unboxed
Published on Jul 28, 2019


https://www.reddit.com/r/Amd/comments/ciw8ui/discussing_userbenchmarks_cpu_speed_index/

Userbenchmark says that it changed its scoring mechanism because Ryzen 3000 chips scored too high (tomshardware).
https://www.reddit.com/r/Amd/comments/cil1fi/userbenchmark_says_that_it_changed_its_scoring/

UserBenchmark changed the way they calculate average score for ranking CPUs to make Ryzen look worse than it is. Since this list is for newbies, this needs to change soon.
https://www.reddit.com/r/Amd/comments/chuu8h/userbenchmark_changed_the_way_they_calculate/

PSA: Use Benchmark.com have updated their CPU ranking algorithm and it majorly disadvantages AMD Ryzen CPUs (cpu.userbenchmark.com)
https://www.reddit.com/r/Amd/comments/chal0r/psa_use_benchmarkcom_have_updated_their_cpu/

Speaking of Intel price cuts... the AMD 3600 vs Intel 9400F shows a large performance gap vs. a small price vs performance difference:

Ryzen 5 3600 vs. Core i5 9400F, Does Intel Offer More Value @ $150?
Hardware Unboxed
Published on Jul 27, 2019

 

It's good to see them coming under fire. I've always viewed them as being sketchy and worthless, pretty much the same way that CPUBoss.com and GPUBoss.com are worthless. Their benchmark is more or less garbage as well. Always has been... even less useful than Passmark, which is pretty sad.

Blind leading the blind and stupid counseling the stupid has never been a good model. Might as well ask noobs on Facebook for technical advice.