Discussion AMD Ryzen MegaThread! FAQ and Resources

Page 42 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar


Ryzen is quite new (I don't have one, yet) so I have no hands on experience with them. That said, max temps of circa 76C don't sound too bad- usually silicone is good up into the 90 range before it gets dangerous, so personally I'd say your overclock is safe enough, though I don't think yo are going to be able to push it higher.
 


Ok, but from the same post:
"Simultaneous Multi-threading (SMT)
Finally, we have investigated reports of instances where SMT is producing reduced performance in a handful of games. Based on our characterization of game workloads, it is our expectation that gaming applications should generally see a neutral/positive benefit from SMT. We see this neutral/positive behavior in a wide range of titles, including: Arma® 3, Battlefield™ 1, Mafia™ III, Watch Dogs™ 2, Sid Meier’s Civilization® VI, For Honor™, Hitman™, Mirror’s Edge™ Catalyst and The Division™. Independent 3rd-party analyses have corroborated these findings.

For the remaining outliers, AMD again sees multiple opportunities within the codebases of specific applications to improve how this software addresses the “Zen” architecture. We have already identified some simple changes that can improve a game’s understanding of the "Zen" core/cache topology, and we intend to provide a status update to the community when they are ready."

So that does suggest *there is a problem with how some games handle SMT with Zen* (which *isn't* a problem on Intel- most likely due to the relative ages of the two implementations). What this means is:
- The issue isn't a fundamental Windows problem (a shame really- as a Windows fix would have worked for multiple titles).
- There is still an issue to be worked on, although it looks like the 'fix' is on a per application basis and down to the devs. That suggests that it's something that will get sorted out for future titles, but is unlikely to get much support from existing titles imo.

It's interesting all the same, although I maintain that the gaming performance on Ryzen is actually quite impressive when you take into consideration it's starting point. There's also quite a bit of evidence to support the notion that whilst not *quite* as fast at 1080p lowered settings, the higher thread count does result in a smoother experience with very high end gpu's at higher settings (I think it's worth keeping in mind that the 1080p 'cpu' tests merely show the maximum draw call performance of the cpu and there are other factors required to maintain smooth frame rates).
 


Funny evidence says otherwise though... Still, if AMD says so, then that's the end of it?

LOL, no it isn't. Even AMD engies can be wrong about stuff like that.

I still see The Stilt being serious about it:

"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."

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

Well, I guess it's old by now, but there hasn't been any patches for TW:W that I know of to address that performance delta.

EDIT: I'd put some links to a French review/article investigating a bit more about core distribution (AKA core parking) and cache movement, but it seems it's taboo to do it or Toms thinks people has issues with google translate.

But long story short: heavy context switches hurt Ryzen badly. That is an scheduler problem. Whether AMD wants to admit/pass blame on that or not, is something else.

Cheers!
 


So that means performance is final? Then yeah if so even when games use 8 cores Ryzen loses to a 4 core Kabylake and in those same games the 8 core broadwell part beats a 4 core kaby lake.

Yeah the R7 Ryzen should not be considered gamer first CPUs

R3 quad cores should be decent for budget gamers however.
 


Effectively explained by some of the users.
"Win10 scheduler may be operating properly but that does not mean it is fully utilizing the Ryzen architecture in an efficient manner. As things are now, you can have a simple 2-thread application in which, randomly, one thread is assigned to a different CCX than the other. This visibly reduces performance. You can run the same application again and, if you're lucky, have that same application's threads assigned to the same CCX--resulting in improved performance. This has been demonstrated. All an OS would have to do is prefer to put application threads on the same CCX when possible.
Same with SMT, the performance losses in some games with SMT enabled is NOT because the scheduler does not see the difference in them, but because it is more likely to switch threads from one CCX to another when SMT is enabled sometimes."
 
http://www.digitaltrends.com/processor-reviews/amd-ryzen-7-1700-review/

Very good article. I agree with this one. AMD's Ryzen 1700 is consider better value to use for heavy task projects or gaming awhile streaming via twitch is better performance. Use Ryzen 1700 for now... Until Ryzen 2 or Ryzen 3, we able to upgrade it from Ryzen 1700 in future. because Ryzen 2/3 will use same AM4 socket.

-DG
 


Hi,
I've got the same build as you but instead of the 1070 i've got the RX480.
I think i'll have some good room to play and do fun stuff.
 


I said there is no scheduler issue. Evidence from third party reviews demonstrated there is no scheduler issue. Finally, AMD is officially admitting there is no scheduler issue.

The AT forum is completely hilarious. The same people has been for months posting wrong information is now rejecting the official explanation and inventing conspiracy theories about Wintel and so. :lol:
 


It's fine that they do it. Intel and AMD have been challenged before and people has been right.

Well, as long as the conspiracy theories don't go too far, that is.

I'm also surprised you would laugh at them, when you fight all the time stuff that is given out to the public even by AMD itself. I believe you're being a bit unfair there.

Cheers!
 


Reviews have demonstrated that SMT and CCX configurations don't affect gaming significantly. SMT on/off changes average gaming performance by about 1% and 3%, min-FPS and avg-FPS respectively. On the other hand (2+2) vs (4+0) CCX configurations change average performance by about 2%. The W10 scheduler is working fine. Some concrete games can be seeing higher performance penalties and AMD can collaborate with game developers to improve those concrete titles, but overall, the game performance will not change above the very low single digit percents.
 


The comment from an anonymous poster that is giving cherry picked information? LOL He only mentions Battlefield 1 result from Hardware.fr and ignores anything else.

Testing by PcPer shows that the scheduler is aware of the CCX structure and assigns threads to cores in the same CCX when it is possible like in lightly threaded workloads.

Hardware.fr tested several games and applications and found that the average performance gap between 2+2 and 4+0 configurations was so small like 3%

getgraphimg.php


Battlefield 1 looks as an outlier datapoint. More about this below.

Testing made by PcGamesHardware finds the same. For Honor, Rise of the Tomb Raider, and Watch Dogs 2 showed minor performance changes between 4+0 and 2+2 configurations. Only Battlefield 1 showed a 10% variation in performance, and the test was made at 720p

Ryzen-R7-1800X-Test-CPU-Core-Scaling-Battlefield-1-pcgh.png


Battlefield 1 is the exception that confirms the general rule. And the huge 20% gap measured by Hardware.fr could even be an outlier

http://www.itl.nist.gov/div898/handbook/eda/section3/eda35h.htm
 


That is fine and all, but when you're within strike distance of your competition (20% or less) that 3% is very significant and the outliers are actually important when they are popular titles. So, caveats all around.

I've been reading that MS will indeed put a patch for Ryzen eventually addressing the power and scheduling "discrepancy", so there's that. Still rumor and I haven't seen any confirmation from MS.

Cheers!
 


Hummm! A FX-8350 beating an i5-2500k in games? Well, I suppose that is possible if one cherry pick games and select only those that scale to 8 threads. Here we have the result of extensive gaming testing

1080_All.png


FX-8370 is beated by i3s and Pentiums. There is no i5-2500k tested, but it seems evident that it will be not 50% slower than the i5-6600k

Nice article explaining why avoiding a GPU bottleneck is so important for testing and understanding CPU performance

http://www.techspot.com/news/68407-tackling-subject-gpu-bottlenecking-cpu-gaming-benchmarks-using.html
 


As said before, beware of using average, because you can't put into the same calculation the games that are not affected by the issues. For example, if 9 games are single threaded and one is multi threaded, average will be 1/10th of the increase in that game.

And, as I posted earlier, performance can increase by as much as 15% both in average and minimum FPS, which is substantial.
 


There is a huge issue with the 'low resolution / settings' approach though- and that is that many of the higher graphics settings also impact cpu (quite a few effects load the CPU for example, whilst higher draw distance increases the number of AI / physics objects on screen).

I think what people keep missing over and over is exactly what a low resolution / settings 'cpu' gaming test actually shows- it shows 1 specific facet of performance: CPU bound *draw call* performance. It imo however ignores many other sources of CPU load, meaning a high clocked low core count CPU has an unrealistic advantage. There are a few cases where Ryzen bests Intel at higher resolution but not at 1080p lowered settings- the reason being all those extra threads / cache can be put to use when you load everything up. I mean these 'cpu' tests regularly show the i5 matching the i7 in games- yet when you get into a multiplayer scenario the i5 really starts to struggle (whereas Ryzen or an i7 show real value). These cpu game tests are an interesting data point- but I think far too much emphasis is put on them. I'd like to see more high load multi player tests (a number of vids on YouTube suggest an R7 is a much better processor than an i5 for example in such cases- as the workload isn't a flat out fps race).

The final point is that there are a few cases where Ryzen isn't performing anywhere near where it should (AOTS for example)- and based on early comments from the developers that is something that is likely to change. My guess is AOTS post patch will put the R7 parts above the 7700k but probably below the 6900k.
 


If there's nothing wrong with how Win10 handles Ryzen why did Microsoft say they were releasing an update for it in their batch Windows Updates due for release in April ?
 
^ low resolution and AA don't use cpu so they can be turned down without affecting cpu's useage.
some settings like draw distance use both so they needs to be turned up
some use mostly cpu like number of cpu bots/players so they should be maxed out.

in short there is nothing wrong with testing cpu at low resolution, if we don't do that then 1090t will be as fast as r7-1800x @4k ( someone please test an overclocked 1090t ( with atleast 2400mhz cl10 ram and cpu-nb to ~3ghz ) )
 


That's a nice point. Why don't MP game developers make a benchmark option, in which they start a match with only pre-programmed bots and camera movement? That wouldn't be hard to make, I think.

But about low-res testing, if they test at maximum settings on 1080p, we must agree that all possible CPU settings are on, so as long as the GPU is powerful enough, there's no problem.



Maybe because it works, but could work better. Like the power profile thing, AMD said they are fixing but Microsoft owns the system, so a fix on their side will help too. Also, I believe they said "nothing is wrong with how W10 handles threads for Ryzen", but not other things.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom