AMD Zen 2 (Ryzen 3000) - * NO COMPETITOR HINTING *

Andromeda1983 · 8 Jul 2019 at 06:40

There seems to be problems with Nvidia drivers causing WHEA errors (Windows Hardware Error Architecture) which doesn't seem to have anything to do with PBO. It's likley an Nvidia issue with the new Ryzen CPU's, hopefully a hotfix will come out soon

Kashinoda · 8 Jul 2019 at 06:47

edit..

Vish Petrol · 8 Jul 2019 at 06:53

Nvidia drivers are just "Super"

SiDeards73 · 8 Jul 2019 at 07:02

Kashinoda said:
Looks like that boost issue was much ado about nothing.

Interesting, it kinda conflicts with what this thread on Reddit was saying https://www.reddit.com/r/Amd/commen...ing_gimped/?utm_source=share&utm_medium=web2x

"
So people don't have to read through the whole thing, this is the part about the boost frequencies. (seems like benchmarks might need to be redone)

The whole story…

During the first three hours of testing of the AMD Ryzen 9 3900X processor, using the X570 AORUS XTREME board, I noticed the problem when PCMark 8 did not pass the first test after 40 minutes (this is a total of ten tests). I noticed WHEA error (Windows Hardware Error Architecture) in HWInfo64 (se this software for PC telemetry, highly suggested).

From there I also decided to pay more attention to HWInfo64 and also checked that the BOOST frequencies of the processor had problems, since it didn’t get to “boost” all its cores to the maximum that it should, which is 4.6 GHz. It reached 4.5 GHz to 4.575 GHz in a pair of cores and the rest of cores to 4.3-4.4 GHz… We used manufacturers chipset driver, we have used press chipsets, as more current chipset driver version, same results.

It seemed strange to me, so I first decided to write to my contact with GIGABYTE USA (Matthew Hurwitz, I thank him for all the time he has put in to find a solution) and showed him the WHEA (PCI Express) errors, as well as the rare behavior of the 3900X boost frequencies.

Midnight (Wednesday) GBT HQ gives us news and according to their tests, the new AGESA code, including NPRP BIOS (BIOS for press) replicated our results in single-core frequencies, BUT, the original BIOS (AGESA 1002, without code introduced NPRP) turbo boost was working well.

With this information, I decided to flash BIOS, the first BIOS released for the X570 AORUS MASTER board and surprise, the boost frequencies were working as they should, even beyond the processor at 4.65 GHz. The WHEA error problem in the PCI Express was still going on, so I kept pressing and trying if the problem was maybe the chipset driver."

Andromeda1983 · 8 Jul 2019 at 07:06

It could honestly just likley be as simple as a driver fix for the WHEA errors, I expect it will be fixed by mid week, weekend latest

haszek · 8 Jul 2019 at 07:07

CAT-THE-FIFTH said:
Gamersnexus with their Ryzen 9 3900X sample gained some significant performance in certain games with SMT off:
https://www.youtube.com/watch?v=yqQ2X1y0jvw

PurePC (Polish) did their usual testing for this.
SMT ON/ OFF
https://www.purepc.pl/procesory/tes...7_3700x_premiera_architektury_zen_2?page=0,53 productivity

https://www.purepc.pl/procesory/tes...7_3700x_premiera_architektury_zen_2?page=0,54 games

Memory 2133 vs 3200
https://www.purepc.pl/procesory/tes...7_3700x_premiera_architektury_zen_2?page=0,55
https://www.purepc.pl/procesory/tes...7_3700x_premiera_architektury_zen_2?page=0,56

Kashinoda · 8 Jul 2019 at 07:07

SiDeards73 said:
Interesting, it kinda conflicts with what this thread on Reddit was saying https://www.reddit.com/r/Amd/commen...ing_gimped/?utm_source=share&utm_medium=web2x

"
So people don't have to read through the whole thing, this is the part about the boost frequencies. (seems like benchmarks might need to be redone)

The whole story…

During the first three hours of testing of the AMD Ryzen 9 3900X processor, using the X570 AORUS XTREME board, I noticed the problem when PCMark 8 did not pass the first test after 40 minutes (this is a total of ten tests). I noticed WHEA error (Windows Hardware Error Architecture) in HWInfo64 (se this software for PC telemetry, highly suggested).

From there I also decided to pay more attention to HWInfo64 and also checked that the BOOST frequencies of the processor had problems, since it didn’t get to “boost” all its cores to the maximum that it should, which is 4.6 GHz. It reached 4.5 GHz to 4.575 GHz in a pair of cores and the rest of cores to 4.3-4.4 GHz… We used manufacturers chipset driver, we have used press chipsets, as more current chipset driver version, same results.

It seemed strange to me, so I first decided to write to my contact with GIGABYTE USA (Matthew Hurwitz, I thank him for all the time he has put in to find a solution) and showed him the WHEA (PCI Express) errors, as well as the rare behavior of the 3900X boost frequencies.

Midnight (Wednesday) GBT HQ gives us news and according to their tests, the new AGESA code, including NPRP BIOS (BIOS for press) replicated our results in single-core frequencies, BUT, the original BIOS (AGESA 1002, without code introduced NPRP) turbo boost was working well.

With this information, I decided to flash BIOS, the first BIOS released for the X570 AORUS MASTER board and surprise, the boost frequencies were working as they should, even beyond the processor at 4.65 GHz. The WHEA error problem in the PCI Express was still going on, so I kept pressing and trying if the problem was maybe the chipset driver."

I think the BIOSs are just all over the place right now.

Some of the benchmarks on GamersNexus are pretty alarming too, disabling SMT on the 3900X can often give 20-30% performance uplift depending on the game.

Kashinoda · 8 Jul 2019 at 07:09

haszek said:
PurePC (Polish) did their usual testing for this.
SMT ON/ OFF
https://www.purepc.pl/procesory/tes...7_3700x_premiera_architektury_zen_2?page=0,53 productivity

https://www.purepc.pl/procesory/tes...7_3700x_premiera_architektury_zen_2?page=0,54 games

Memory 2133 vs 3200
https://www.purepc.pl/procesory/tes...7_3700x_premiera_architektury_zen_2?page=0,55
https://www.purepc.pl/procesory/tes...7_3700x_premiera_architektury_zen_2?page=0,56

Seems to not be an issue for the 3700X when disabling SMT, looks like the chiplet design is an issue.

razorpakk · 8 Jul 2019 at 07:11

Grim5 said:
3900x OC vs 9900k OC review for games, work and streaming

That's why I didn't feel bad buying a 9700k yesterday... For 350£, if all you do is gaming, is not that bad.

haszek · 8 Jul 2019 at 07:12

Also on one review they mentioned F clock on auto was 7% slower than manually set to 1:1 ( in AIDA latency but difference was visible in game testing results also).

DragonQ · 8 Jul 2019 at 07:13

easyrider said:
I think the extra 60 for the 3800x is worth it...even before reviews...I reckon buyers remorse will kick in for 3700x owners who just couldn’t wait

Der8auer had a little table of maximum clocks he could get for 12 CPUs, and for the R7 3800X and it was "typically 4.4 GHz". Basically AMD's advertised boost clocks are theoretical maximums that you'll never see in reality. Very shady stuff to be honest, and it means that an R7 3800X ain't gonna do much more than an R7 3700X (100 MHz difference in maximum boost clocks).

SiDeards73 · 8 Jul 2019 at 07:14

Kashinoda said:
Seems to not be an issue for the 3700X when disabling SMT, looks like the chiplet design is an issue.

I agree i think the BIOS's are in a total mess right now, i checked Asrocks page yesterday and there is already 5 or 6 bios releases for the x570 Taichi, 8Pack did say that there was almost daily bios releases being issued.

Hopefully it'll get sorted in a few days / weeks as it was plainly obvious a lot of the chips were not even hitting their stated boost clocks on the box, let alone OC above that.

sityeuropa · 8 Jul 2019 at 07:15

Is it correct in saying if you are not gaming then buying expensive memory is rather pointless for the difference it makes?

Potatowithearsontheside · 8 Jul 2019 at 07:15

Kashinoda said:
I think the BIOSs are just all over the place right now.

Some of the benchmarks on GamersNexus are pretty alarming too, disabling SMT on the 3900X can often give 20-30% performance uplift depending on the game.

I recall seeing one bench where the 3900X had shocking 0.1% and 1% lows when left untouched, and then vastly improved when affinity was forced...with greater overall performance too. This suggests that the latest chipset drivers and 1903 windows update aren't actually doing what they're supposed to do.

SiDeards73 · 8 Jul 2019 at 07:15

DragonQ said:
Der8auer had a little table of maximum clocks he could get for 12 CPUs, and for the R7 3800X and it was "typically 4.4 GHz". Basically AMD's advertised boost clocks are theoretical maximums that you'll never see in reality. Very shady stuff to be honest, and it means that an R7 3800X ain't gonna do much more than an R7 3700X (100 MHz difference in maximum boost clocks).

As stated above, i think its really too early to say much on boost clocks, early bios's were allowing boosts past the box rated speeds, current BIOS are gimping the boosts, something is not quite right. AMD and the board partners need to get this sorted ASAP though thats for sure.

Grim5 · 8 Jul 2019 at 07:16

Kashinoda said:
I think the BIOSs are just all over the place right now.

Some of the benchmarks on GamersNexus are pretty alarming too, disabling SMT on the 3900X can often give 20-30% performance uplift depending on the game.

Why alarming that’s good

Gives a legitimate option to run SMT off to maximise overclock and just run 12 cores for maximum gaming performance

DragonQ · 8 Jul 2019 at 07:19

SiDeards73 said:
As stated above, i think its really too early to say much on boost clocks, early bios's were allowing boosts past the box rated speeds, current BIOS are gimping the boosts, something is not quite right. AMD and the board partners need to get this sorted ASAP though thats for sure.

If that is the case, it really is a terrible blunder from AMD and the board partners because higher clocks obviously = higher performance in most cases and as we all know, the majority of people looking to buy these CPUs will make up their minds from the first round of reviews, not a second round in a month's time. They've had long enough to sort out BIOS problems etc. with a July launch and experience from Ryzen 1 and 2.

Could be nonsense though, I'm not buying expecting some unicorn performance boost down the line, although it'd be nice.

Andromeda1983 · 8 Jul 2019 at 07:21

I am going to keep my pre-order for the 3700X but i'll give it about 1-2 weeks before installing to allow for the bios and nvidia driver bugs to be ironed out, I think thats probably for the best

Kashinoda · 8 Jul 2019 at 07:22

Grim5 said:
Why alarming that’s good

Gives a legitimate option to run SMT off to maximise overclock and just run 12 cores for maximum gaming performance

Why would you want to turn SMT off on such a production heavy CPU? Not to mention the games that's a disadvantage on.

At that point you need to ask yourself why you haven't bought a 9900k lol.

Andromeda1983 · 8 Jul 2019 at 07:24

At that point you need to ask yourself why you haven't bought a 9900k lol.

Because at least when your AMD, a new CPU does not come with a required motherboard upgrade (at least way less than Intel)

Competitor rules

AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

AMD Zen 2 (Ryzen 3000) - * NO COMPETITOR HINTING *