With the new Ryzen CPUs around the corner, which gaming FPS benchmark specs are most important to you?

[dirtyferret]

Every time a new CPU launches PC gamers by the herd go to their favorite review sites and check out FPS benchmarks in their favorite games. Out of curiosity, which FPS benchmarks are most important to you (assuming resolution matches your specific monitor)?

You can vote for as many as you want.

 

[chaosmassive]

none of the above, because if the AMD can match intel in gaming performance, that is enough for me

 

[cucker tarlson]

choosing a testing place is much important than what metric you're gonna use anyway.

 

[dirtyferret]

none of the above, because if the AMD can match intel in gaming performance, that is enough for me

There are dozens of synthetic tests & games with various specs at various resolutions. Which match are you looking for?

 

[phanbuey]

Ryzen's issue was latency and high FPS consistency (frame time variance in game).

I like seeing frame pacing graphs to see how the system reacts in badly optimized game to cache misses, etc. Farcry 5 and FarCry New Dawn, Difficult areas in witcher are great at dropping frames at high FPS, so I would want to see how the new CPUs perform there.

 

[HUSKIE]

Not bother for FPS benchmarking. So non of the above.

 

[cucker tarlson]

Ryzen's issue was latency and high FPS consistency (frame time variance in game).

I like seeing frame pacing graphs to see how the system reacts in badly optimized game to cache misses, etc. Farcry 5 and FarCry New Dawn, Difficult areas in witcher are great at dropping frames at high FPS, so I would want to see how the new CPUs perform there.

hands on exprience is probably the only way you can really review a cpu thoroughly.play 20 games beginning to end and I'll tell you how good it is.like I said,what I value in a cpu review more than anything is choosing the testing places meticulously.you gotta play the game first to see where the cpu struggles.otherwise it's not much of a cpu review.
some of what you mentioned has been the achilles heel of ryzens

Test procesora Intel Core i9-9900K - Bezkompromisowa wydajność | PurePC.pl

Test procesora Intel Core i9-9900K - Bezkompromisowa wydajność (strona 45) Test wydajności procesora Intel Core i9-9900K, który ma osiem rdzeni i szesnaście wątków, będąc najszybszym modelem dla platformy LGA 1151. No i ma także lutowany IHS!

www.purepc.pl

Far Cry New Dawn - test wydajności kart graficznych i procesorów

Far Cry New Dawn stanowi kontynuację wydarzeń z poprzedniej odsłony gry z serii wydawanej przez Ubisoft. Sprawdziliśmy, jakie wymagania przed sprzętem komputerowym stawia nowa wersja Hope County. Oto test kart graficznych i procesorów w Far Cry New Dawn.

pclab.pl

 

[Vario]

The FPS minimum and other latency metrics are most important.

 

[Steevo]

Min FPS, math capability, compression times, render times, USB and connectivity performance metrics, memory performance and compatibility, storage connectivity and performance.

 

[dirtyferret]

choosing a testing place is much important than what metric you're gonna use anyway.

If I asked 50 different people what testing place/game/synthetic benchmark they prefer I would get close to 50 different answers. Then include resolutions, variable FPS benchmarks settings, even graphic settings...no way I could fit that into a poll that limits me to ten options. Luckily comment section allows you to post specifics

 

[Vayra86]

If I asked 50 different people what testing place/game/synthetic benchmark they prefer I would get close to 50 different answers. Then include resolutions, variable FPS benchmarks settings, even graphic settings...no way I could fit that into a poll that limits me to ten options. Luckily comment section allows you to post specifics

This is why I voted for:
AVG + Min FPS and 95th percentile.

It covers the worst case pretty well without being too picky. If you have these numbers over a broad variety of games, you know all you really need to know. Max FPS is pretty useless, nobody spends gaming time staring at the skybox.

 

[cucker tarlson]

most likely it'll be consistency between games,not any particular performance metrics.

If you compare i5 OC to i7 locked you realize the performance will vary,they'll go back and forth depending how the game manages resources.What comes on top on avg. is irrelevant.What is relevant is how cpu a is doing in the scenario which favors cpu b and vice versa.
85% of 9900k performance is still ridiculously good for most gamers,if the tests are really cpu bound.

Thing is,you don't wanna get your average from in 5 games where a cpu comes within 5% and then 5 others where it's at i3 performance or in the ballpark.
4790k stock performance wasn't good enough to drive my gtx 1080 at 1440p with a high refresh display.AMD has to take care of cases where the performance falls off a cliff.Like I said,even 85% of 9900K would be very good,if it were consistent.

 

[MrGRiMv25]

Frame times are pretty much my favourite way to see how something performs, consistency is much more important to me than peak FPS (especially in the Trials series, a laggy frame here and there can mess zero fault runs up very quickly).

Out of the options above I went for FPS lows.

 

[Ebo]

What test that makes my game slow, stutter and finally crash.......that my favorite.

 

[RealNeil]

I'm looking at the overall picture in reviews from multiple sites.
Differing system specs as well as differing tests.
Availability and pricing will factor in too.

I will attempt to wait a while before I buy, just to allow time for others to encounter the bleeding edge problems that new gear can have.

 

[windwhirl]

Eh, Min FPS, Max FPS, Average and 95th percentile.

 

[dirtyferret]

most likely it'll be consistency between games,not any particular performance metrics.

If you compare i5 OC to i7 locked you realize the performance will vary,they'll go back and forth depending how the game manages resources.What comes on top on avg. is irrelevant.What is relevant is how cpu a is doing in the scenario which favors cpu b and vice versa.
85% of 9900k performance is still ridiculously good for most gamers,if the tests are really cpu bound.

View attachment 124412
View attachment 124413

Thing is,you don't wanna get your average from in 5 games where a cpu comes within 5% and then 5 others where it's at i3 performance or in the ballpark.
4790k stock performance wasn't good enough to drive my gtx 1080 at 1440p with a high refresh display.AMD has to take care of cases where the performance falls off a cliff.Like I said,even 85% of 9900K would be very good,if it were consistent.

Not only is 85% of 9900k performance really good but all the CPUs are capable of very good performance in that game with 1% low all above 70.

On a slight tangent; I love Far Cry 5 performance tests especially the GN .1% that people use as the standard for a gaming CPU.

Here's another Far Cry 5 test from Tom's using a Ryzen 1600x clearing showing AMD's simultaneous multi-threading technology works when you have two physical cores but has the reverse effect once you reach four physical cores. Therefore, to future proof your Ryzen CPUs turn off SMT!...or a a cherry picked result from a single game outside the normal range of performance should not be a standard bearer for a CPU...I'm leaning to the former as the latter makes too much sense.

 

[Vayra86]

Not only is 85% of 9900k performance really good but all the CPUs are capable of very good performance in that game with 1% low all above 70.

On a slight tangent; I love Far Cry 5 performance tests especially the GN .1% that people use as the standard for a gaming CPU.

Here's another Far Cry 5 test from Tom's using a Ryzen 1600x clearing showing AMD's simultaneous multi-threading technology works when you have two physical cores but has the reverse effect once you reach four physical cores. Therefore, to future proof your Ryzen CPUs turn off SMT!...or a a cherry picked result from a single game outside the normal range of performance should not be a standard bearer for a CPU...I'm leaning to the former as the latter makes too much sense.

This is not really a good analysis. Using one game, on a very old engine, and going full detail on minute differences in performance, tells you exactly nothing about CPUs. It just tells you how that game works. Let's face it, after 40 odd hours you uninstall FC5 and never look back anyway.

To circle back to the subject at hand: the subject is 'choosing an optimal gaming CPU'. You don't do that based on a single game - and if you do, you're doing it wrong. You're saying you see a 'reverse effect' with SMT on, but honestly? The most I see is a few % longer dark green bar, which is actually the difference between a 2ms and 4ms frametime latency gap. This is margin of error territory. That CPU is smooth as butter in all cases, and the way different core counts respond wrt frametimes is also a thing of clockspeeds and latencies to RAM, which first gen Ryzen lacks. You say SMT is meh but that conclusion is flat out wrong, I'd say you put a CPU out of its comfort zone, so it loses consistency. The only real conclusion you can draw here is that this CPU is best for fixed 60~75hz monitors/FPS, and unfit for high refresh.

Its a good way to say alot without saying anything - which is of course great stuff for a YT channel like GN. Not saying the testing is irrelevant, but it needs more context. Its easy to get drowned in numbers and lose sight of reality.

 

[dirtyferret]

This is not really a good analysis. Using one game, on a very old engine, and going full detail on minute differences in performance, tells you exactly nothing about CPUs. It just tells you how that game works. Let's face it, after 40 odd hours you uninstall FC5 and never look back anyway.

To circle back to the subject at hand: the subject is 'choosing an optimal gaming CPU'. You don't do that based on a single game - and if you do, you're doing it wrong. You're saying you see a 'reverse effect' with SMT on, but honestly? The most I see is a few % longer dark green bar, which is actually the difference between a 2ms and 4ms frametime latency gap. This is margin of error territory. That CPU is smooth as butter in all cases, and the way different core counts respond wrt frametimes is also a thing of clockspeeds and latencies to RAM, which first gen Ryzen lacks. You say SMT is meh but that conclusion is flat out wrong, I'd say you put a CPU out of its comfort zone, so it loses consistency. The only real conclusion you can draw here is that this CPU is best for fixed 60~75hz monitors/FPS, and unfit for high refresh.

Its a good way to say alot without saying anything - which is of course great stuff for a YT channel like GN. Not saying the testing is irrelevant, but it needs more context. Its easy to get drowned in numbers and lose sight of reality.

Re-read my last line then read link

How Sarcasm Works

It's dreary outside, and you remark on the beautiful weather. An incompetent co-worker is promoted, and you say she'll do a great job. It's not Opposite Day -- it's sarcasm.

people.howstuffworks.com

 

[Vayra86]

How Sarcasm Works

It's dreary outside, and you remark on the beautiful weather. An incompetent co-worker is promoted, and you say she'll do a great job. It's not Opposite Day -- it's sarcasm.

people.howstuffworks.com

The penny dropped, thanks!

 

[vega22]

for gaming its highs and lows that matter most to me. the bigger the gap, the worse the exp.

but the cpu only really matters for gaming in a couple of the games i play. only really source games tbh.

@1440p or 2160p i am mostly gpu bound anyway :/

 

[Chomiq]

Personally I care only about frame time consistency. Spikes/dips are immediately noticeable to me.

 

[cucker tarlson]

I would really love to get some hands on experience with gaming on a ryzen 3000,just to see how that cpu handles various requirements.
testing a cpu is much more than just numbers,although numbers have to be there too of course.for example when comparing my own two systems - 4790k with gtx 1080 and 5775c with 1080ti - it's not just the numbers.when 4790k was hitting +90% utilization you'd see frame drops and hitching.on 5775c there's no perceptible difference how the game runs at 90% and 50% cpu utilization.At 97-100%,which happens in Odyssey,you'll see intermittent microstutter here and there but still nothing game breaking.

 

[TTU]

hands on exprience is probably the only way you can really review a cpu thoroughly.play 20 games beginning to end and I'll tell you how good it is.like I said,what I value in a cpu review more than anything is choosing the testing places meticulously.you gotta play the game first to see where the cpu struggles.otherwise it's not much of a cpu review.
some of what you mentioned has been the achilles heel of ryzens

Test procesora Intel Core i9-9900K - Bezkompromisowa wydajność | PurePC.pl

Test procesora Intel Core i9-9900K - Bezkompromisowa wydajność (strona 45) Test wydajności procesora Intel Core i9-9900K, który ma osiem rdzeni i szesnaście wątków, będąc najszybszym modelem dla platformy LGA 1151. No i ma także lutowany IHS!

www.purepc.pl

Far Cry New Dawn - test wydajności kart graficznych i procesorów

Far Cry New Dawn stanowi kontynuację wydarzeń z poprzedniej odsłony gry z serii wydawanej przez Ubisoft. Sprawdziliśmy, jakie wymagania przed sprzętem komputerowym stawia nowa wersja Hope County. Oto test kart graficznych i procesorów w Far Cry New Dawn.

pclab.pl

Yep, choosing the right testing spot is crucial ! Also how you gather and transalte the data is important.

I couldnt agree more on what you've said but the thing is, when CPU testing almost all reviewers just load up the in-game benchmark and calls it a day. If a game doesnt have it, they probably will use something from the start of the game. Thats why, in a lot of games we see little to no difference, even between CPUs in totally different classes. And I dont think the situation will change with Zen 2. Even if all the problems and ''achilles heel'' of Ryzen are fixed and the CPUs are more than competitive. That is at least from the usual review outlets.

For Example, the Far Cry 5 results from PurePC are pretty much inline with Techpowerup's. Im not sure, but I think FaCry 5s benchark is not good for CPU testing, it looks like pre rendered tour, without much interraction/AI, but the game is so single-threaded that it doesnt really matter. And in contrast The Witcher 3 results are way off. PurePC use Novigrad as a test run, which is infamously heavy on the CPU. Wonder what section of the game TPU have for their benchmark.

I usually read PurePC's reviews, and Ive always been curious about the second metric they use. Im pretty sure its not 1% low's (let alone 0.1%), because its super consistent with the main, and not too far off. Im also curious what tools they use.

I also like Gamers Nexus, but they use in-game benchmarks, and have heavy, unnecessary scientific approach that hurts their work (back to the thing you've said about playing the game). And the irony is their name is ''Gamers Nexus'' . I Dont wanna bash on them of course, I guess this is how it works out when you have to work with such a small team + their work is quite intensive to publish. If they change their ways, they will be one of the best, If not the best, especially since TechReport is no more.

Frame times are pretty much my favourite way to see how something performs, consistency is much more important to me than peak FPS (especially in the Trials series, a laggy frame here and there can mess zero fault runs up very quickly).

Out of the options above I went for FPS lows.

Given your expectations you should've chose the varous percentiles, not the FPS low ?

 

[Vayra86]

Yep, choosing the right testing spot is crucial ! Also how you gather and transalte the data is important.

I couldnt agree more on what you've said but the thing is, when CPU testing almost all reviewers just load up the in-game benchmark and calls it a day. If a game doesnt have it, they probably will use something from the start of the game. Thats why, in a lot of games we see little to no difference, even between CPUs in totally different classes. And I dont think the situation will change with Zen 2. Even if all the problems and ''achilles heel'' of Ryzen are fixed and the CPUs are more than competitive. That is at least from the usual review outlets.

For Example, the Far Cry 5 results from PurePC are pretty much inline with Techpowerup's. Im not sure, but I think FaCry 5s benchark is not good for CPU testing, it looks like pre rendered tour, without much interraction/AI, but the game is so single-threaded that it doesnt really matter. And in contrast The Witcher 3 results are way off. PurePC use Novigrad as a test run, which is infamously heavy on the CPU. Wonder what section of the game TPU have for their benchmark.

I usually read PurePC's reviews, and Ive always been curious about the second metric they use. Im pretty sure its not 1% low's (let alone 0.1%), because its super consistent with the main, and not too far off. Im also curious what tools they use.

I also like Gamers Nexus, but they use in-game benchmarks, and have heavy, unnecessary scientific approach that hurts their work (back to the thing you've said about playing the game). And the irony is their name is ''Gamers Nexus'' . I Dont wanna bash on them of course, I guess this is how it works out when you have to work with such a small team + their work is quite intensive to publish. If they change their ways, they will be one of the best, If not the best, especially since TechReport is no more.




Given your expectations you should've chose the varous percentiles, not the FPS low ?

This is really what I advocate with CPU testing as well, test games actually in-game, and not canned benchmark, and test them in the worst-case scenarios - that data is far more valuable than the 'average' run, really. Even better is when the average run or canned benchmark and the worst-case one are benched side by side. Thát is providing some real insights.

As it stands, you really need to be reading several different review sites AND you have to know where to look.