Zen sporadic Performance in reviews due to Motherboards

SniperDude · 3 Mar 2017 at 11:51

Just cancelled order im not taking the risk, it might or might not be fixed/get better... its too much money to waste on 1800x and a k7 board.

so far I have heard its the BIOS's/new motherboards turn this and that off and it makes it better ..use 4k and all will be good

Firestar_3x · 3 Mar 2017 at 11:58

amigafan2003 said:
Yeah, 'cos AMD build/design the mobos.

A Toyota car crashes because the accelerator pedal sticks, the fault is actually with Siemens who made a sensor, however the public blames?....

It was AMD's choice to release the processor now, they should have waited, it is AMD's launch to control not anyone elses, they have rushed it and are yet again falling foul of some bad press.

xs2man · 3 Mar 2017 at 12:00

amigafan2003 said:
It'll work better if you switch 8 of them off though

Only if your a gamer. The other 8 threads are awesome sauce...

Spyhop · 3 Mar 2017 at 12:17

On the plus side, if this is fixable at the software level(s), then it definitely will end up getting fixed across the board.

AMD must be on red alert due to the bad PR it's causing. I bet there's some kind if internal taskforce been formed.

SniperDude · 3 Mar 2017 at 12:28

Spyhop said:
On the plus side, if this is fixable at the software level(s), then it definitely will end up getting fixed across the board.

AMD must be on red alert due to the bad PR it's causing. I bet there's some kind if internal taskforce been formed.

oh they know .....stock price dropped right on release

http://www.cnbc.com/2017/03/02/shar...cpu-disappoints-with-gaming-performance-.html

SiDeards73 · 3 Mar 2017 at 12:29

Spyhop said:
On the plus side, if this is fixable at the software level(s), then it definitely will end up getting fixed across the board.

AMD must be on red alert due to the bad PR it's causing. I bet there's some kind if internal taskforce been formed.

Well this is arguably make or break for them, this is last chance saloon for AMD with many users, luckily the architecture is good, i can imagine they will toss a lot of resources at getting this resolved asap with motherboard vendors, pulling strings with Microsoft to get their patches implemented and speaking to game devs to get code updated where possible.

THe whole buddying with Bethesda makes even more sense now.

N19h7m4r3 · 3 Mar 2017 at 13:50

SniperDude said:
oh they know .....stock price dropped right on release

http://www.cnbc.com/2017/03/02/shar...cpu-disappoints-with-gaming-performance-.html

The Entire Market took a hit, are you telling my Ryzen reviews caused NVIDIA to drop 3.66% as well? AMD took a hit due to gaming yes, but the rest is the entire market. Short term traders and people hopping on the AMD Hype train have jumped ship, causing the rest of the % drops.

AMD is still a buy long term, as they're set to get Naples into enterprise and HPC; where the real big money is. Even so they should have waited at least another month to launch without all these issues. It would have avoided so much trouble for them.

Since it's March there's most likely going to be an Interest Hike, which has caused all markets to drop.
http://www.cnbc.com/2017/03/01/chance-of-a-march-rate-hike-doubled-cme.html

Troezar · 3 Mar 2017 at 14:01

N19h7m4r3 said:
The Entire Market took a hit, are you telling my Ryzen reviews caused NVIDIA to drop 3.66% as well? AMD took a hit due to gaming yes, but the rest is the entire market. Short term traders and people hopping on the AMD Hype train have jumped ship, causing the rest of the % drops.

AMD is still a buy long term, as they're set to get Naples into enterprise and HPC; where the real big money is. Even so they should have waited at least another month to launch without all these issues. It would have avoided so much trouble for them.

Since it's March there's most likely going to be an Interest Hike, which has caused all markets to drop.
http://www.cnbc.com/2017/03/01/chance-of-a-march-rate-hike-doubled-cme.html

Yes it surprises me now many people don't like competition and would like to see AMD fail.

Sargatanas2511 · 3 Mar 2017 at 15:02

Well for me they have until May or June to get this fixed. As long as it's fixed by then, that's where my money will be going.

mmj_uk · 3 Mar 2017 at 15:35

Mauller said:
It appears that motherboards and the bios revisions seem to be a big culprit with the large variance observed between different reviews.

https://translate.google.co.uk/tran...ndlich-zurueck-1703-125996-4.html&prev=search

Even with the updated BIOS they still found game performance to be subpar:
https://translate.google.co.uk/translate?hl=en&sl=de&tl=en&u=https://www.golem.de/news/ryzen-7-1800x-im-test-amd-ist-endlich-zurueck-1703-125996-13.html

Games work differently than many applications
However, this will not change the basic weakness of the current Zen implementation at Ryzen: In games large amounts of data are constantly being moved between RAM, caches, cores and bus system. A fast memory controller is essential, and Ryzen does not have exactly that, only by overclocking it accelerates itself. The fact that the new AMD chips are fast in rendering applications or video transcoders is because the data from input and output material does not change at runtime and the algorithms are so small that they usually fit completely into the caches. This is different for games and packers, they always produce changing data at runtime. But: Games do not benefit from Intel's Quadchannel.

So according to them it's an architectural weakness with the memory controller/subsystem?

Curlyriff · 3 Mar 2017 at 15:54

DragonQ said:
This is why the issues need to be fixed before the 6 and 4 core versions are released. Those will garner a whole new wave of reviews, particularly with gaming benchmarks, and they need to have Ryzen performing at its best rather than without proper drivers and dodgy early BIOSs. It's the only chance they have to stop the entire Zen line-up being tainted by the initial batch of reviews.

This to me is the biggest reason why I think they released the CPU's this way around. I also feel it has already been pushed by AMD that more core/thread is the way to go as my understanding is that all engineering samples are the 8C/16T for the whole year and they were talking about 1k systems for developers to use so that is clearly their aim.

fs123 · 3 Mar 2017 at 16:14

mmj_uk said:
Even with the updated BIOS they still found game performance to be subpar:
https://translate.google.co.uk/translate?hl=en&sl=de&tl=en&u=https://www.golem.de/news/ryzen-7-1800x-im-test-amd-ist-endlich-zurueck-1703-125996-13.html

So according to them it's an architectural weakness with the memory controller/subsystem?

Which makes me wonder what was all the BS about Infinity fabric and other marketing fluff about if the communication between cores and memory systems is sub-par. I thought they designed it to be much faster than anything before it.

AMD needed to make sure everything was optimal before launch but yet again failed to make good first impressions. Hopefully they will sort out all the problems in time for the mainstream market X1600 and lower Ryzens.

Crockett · 3 Mar 2017 at 18:07

It's frustrating they seem to have done 95% of the work and the last little bit has not been ready for launch which paints the chips in a bad light. I hope they can sort/improve all the SMT, bios and memory issues out for the R5 launche and they will then be a in a great position.

Sayso · 3 Mar 2017 at 21:54

Martini1991 said:
It's like I'm witnessing bulldozers launch again.
The exact same things are being used as excuses again.

I put my blame on thread priority (which was an issue for BD). The performance is there this time unlike BD.

For me when I get my 1700, it'll either be fixed or I'll just have to disable SMT. Or just stick it till next iteration.

Except bulldozer had worse IPC than my 1090T...this is 40%+ better than last gen and not to far behind Intel, its nothing like the Bulldozer launch.

pete910 · 4 Mar 2017 at 00:31

Rossi~ said:
How many Ryzen threads do we need? Honestly.

Well, it is multi threaded

We need to be careful though as we may need to disable some...

joeyjojo · 5 Mar 2017 at 11:11

pete910 said:
Well, it is multi threaded

We need to be careful though as we may need to disable some...

See posts #12-14.

Rroff · 5 Mar 2017 at 11:21

mmj_uk said:
Even with the updated BIOS they still found game performance to be subpar:
https://translate.google.co.uk/translate?hl=en&sl=de&tl=en&u=https://www.golem.de/news/ryzen-7-1800x-im-test-amd-ist-endlich-zurueck-1703-125996-13.html

So according to them it's an architectural weakness with the memory controller/subsystem?

Quad channel isn't without its benefits for gaming - but its marginal. Games largely require very fast (tick rate), low latency memory ops and things like wide buses matter less - whatever way you shake it the main game loop that games run over and over again to build each new frame has a highly serial nature and much of it can only ever be processed serially due to dependency on data from the previous step in the chain and for most games will be the main performance bottleneck at the end of the day despite increased ability to utilise extra threads with things like AI and physics.

AMD was really showing signs they had learnt to balance this better architecturally so I'm quite disappointed to see these issues.

CAT-THE-FIFTH · 5 Mar 2017 at 11:36

Martini1991 said:
I feel like I'm in October 2011.

Yeah,it sounds like that too - remember all the Gigabyte motherboards which had throttling issues and could not overclock due to the lack of things like LLC,etc.

Martini1991 · 5 Mar 2017 at 11:40

CAT-THE-FIFTH said:
Yeah,it sounds like that too - remember all the Gigabyte motherboards which had throttling issues and could not overclock due to the lack of things like LLC,etc.

At least this time the underlying CPU's solid though.

If the issues are fixed for the hex and quad core, then unless they redo the octo core results, it'll probably look weird.
Either way, besides the AMD launch issues (Lets face it, there's a few) this is a cracking chip.

CAT-THE-FIFTH · 5 Mar 2017 at 11:40

Rroff said:
Quad channel isn't without its benefits for gaming - but its marginal. Games largely require very fast (tick rate), low latency memory ops and things like wide buses matter less - whatever way you shake it the main game loop that games run over and over again to build each new frame has a highly serial nature and much of it can only ever be processed serially due to dependency on data from the previous step in the chain and for most games will be the main performance bottleneck at the end of the day despite increased ability to utilise extra threads with things like AI and physics.

AMD was really showing signs they had learnt to balance this better architecturally so I'm quite disappointed to see these issues.

Its also very weird to see Windows 7 NOT have the SMT issues!!

It looks like The Stilt has done even more testing:

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-8#post-38775732

I did some 3D testing and eventhou there is not nearly enough data to confirm it, I'd say the SMT regression is infact a Windows 10 related issue.
In 3D testing I did recently on Windows 10, the title which illustrated the biggest SMT regression was Total War: Warhammer.

All of these were recorded at 3.5GHz, 2133MHz MEMCLK with R9 Nano:

Windows 10 - 1080 Ultra DX11:

8C/16T - 49.39fps (Min), 72.36fps (Avg)
8C/8T - 57.16fps (Min), 72.46fps (Avg)

Windows 7 - 1080 Ultra DX11:

8C/16T - 62.33fps (Min), 78.18fps (Avg)
8C/8T - 62.00fps (Min), 73.22fps (Avg)

At the moment this is just pure speculation as there were variables, which could not be isolated.
Windows 10 figures were recorded using PresentMon (OCAT), however with Windows 7 it was necessary to use Fraps.