Ashes of the Singularity Receives Ryzen Performance Update - Up to 31% Improvement

[Bouzoo]

52 minutes ago, Kloaked said:

I'm actually surprised if this is true.

 

 

Been saying it for a while, that faster memory can help relieve CPU bottlenecks. This wasn't only true for Ryzen - this was also true with previous Intel architectures. I experimented with this idea myself with my 2500k and going from 1600Mhz to 2133Mhz memory. My minimum frame rate increased in games like Witcher 3. I also ended up getting a good job sometime after I did this but haters will say there's no connection.

 

Yet sometimes when I mention memory speeds potentially holding a CPU back in certain applications, someone almost always links that one video Linus did like 6 years ago or something.

I perfectly know which video you're thinking of. It is not hard to demonstrate even with a non K chip by downclocking your memory. 

[Thony]

3 hours ago, Bouzoo said:

Please, someone tell me how RAM speed doesn't affect performance.

 

Surely does for Ryzen. Im pretty sure Intel has different story with their chips. But they already perform great anyhow.

[Drak3]

3 hours ago, Bouzoo said:

Please, someone tell me how RAM speed doesn't affect performance.

 

RAM speed doesn't affect performance. The hamsters inside the CPU does.

/S

 

Ya happy now?

[samcool55]

Nice to see devs actively optimizing games for ryzen, let's hope more devs will show their optimizations in the future

[DocSwag]

46 minutes ago, Valentyn said:

 

Well it already is if you consider the R5 upcoming ones.
Considering the FPS to price difference the 1400 and 1500X 4c/8t are going to be fantastic processors, even if they're that 7% slower than the i5 7400K-7600

Could be the Chips Raja meant when he said his goal for 2017 is $1000 4K gaming for consumers.

I was mainly referring to the 8 core CPUs, though yeah. Thing is, Ryzen is still a bit behind Kaby lake in gaming so perhaps optimizations really could catch up Ryzen to Kaby lake...

[Citadelen]

3 hours ago, Misanthrope said:

-snip-

The point is that gains of these magnitude are possible. They got this out in a few weeks, I don't see it being to hard for future games to be patched for AMD CPUs. Of course it remains to be seen if they do, but I do not think it's as bad as you make it out to be.

[Valentyn]

20 minutes ago, DocSwag said:

I was mainly referring to the 8 core CPUs, though yeah. Thing is, Ryzen is still a bit behind Kaby lake in gaming so perhaps optimizations really could catch up Ryzen to Kaby lake...

IPC wise would be extremely hard to ever match Kabylake. Kabylake has a 7% advantage there, and can clock higher. 

 

The difference comes from price to performance. This reminds me perfectly of the Althon XP days. Intel was slightly ahead, but lost when it came to price to performance over all.

Despite that though, Ryzen is a stellar processor and platform.

Let's hope Ryzen 2 is AMD's next Athlon 64.

[Misanthrope]

17 minutes ago, Citadelen said:

The point is that gains of these magnitude are possible. They got this out in a few weeks, I don't see it being to hard for future games to be patched for AMD CPUs. Of course it remains to be seen if they do, but I do not think it's as bad as you make it out to be.

You honestly don't see why they are able to and others won't be?

 

Let me recap if anything, for the cheap sits:

 

1) AMD is probably working closely with them. Most companies will not have that level of support from AMD simply because they don't have the resources and because AMD themselves probably do not have the resources to optimize for many other games

 

2) The game was a CPU bound game from it's inception so it's obviously much easier to see gains by optimizing for a CPU. So unless you want to specifically release an RTS game you're shit out of luck and optimizing that much for your game will not be feasible and maybe not even possible since your game prioritizes better and more detailed graphics, etc.

 

So no, most games will not be patched because it would be a hell of a lot more difficult often implying rebuilding the game engine from scratch.

[Morgan MLGman]

19 minutes ago, samcool55 said:

Nice to see devs actively optimizing games for ryzen, let's hope more devs will show their optimizations in the future

Well, while I'd like it to be true, I wouldn't go as far to say that "devs are actively optimizing games for ryzen" yet, because Oxide is in a close partnership with AMD...

[Valentyn]

1 minute ago, Morgan MLGman said:

Well, while I'd like it to be true, I wouldn't go as far to say that "devs are actively optimizing games for ryzen" yet, because Oxide is in a close partnership with AMD...

 

AMD did state to PcPer that they send out over a thousand development kits to developers to help optimize for Ryzen.

Now whether that's for current games or only for upcoming we have no idea. We also don't know what performance difference that'll make.

For all we know it could be that Ryzen is going to be in the Xbox Scorpio and AMD wants developers to work on that and Ryzen in a similar fashion for games.

Still a lot of ifs, and buts, but only time will tell. We do know AMD is now a major partner with Bethesda, which previously used Gameworks and was a The Way It's Meant To Be Played partner.

[ivan134]

@HKZeroFive

 

AMD released this on their channel. Should be relevant enough to add to OP.

 

[ApolloFury]

4 hours ago, Bouzoo said:

Please, someone tell me how RAM speed doesn't affect performance.

Nope, it doesn't affect performance at all.

/jk

[Coaxialgamer]

5 hours ago, Morgan MLGman said:

FYI :

People keep bringing up that GF video. It's not even a very real world comparison. They used an overclocked titan x with an i5 at 1080p. No one was pairing a 200$ cpu with a 1000$ cpu back then, and neither are they now. 

 

Sure you could make the argument that the 1070 is just as fast for 400$, but gpu workloads have gotten significantly heavier since the  witcher 3 came out. 

 

I'd argue that the 1070 gives you less frames in modern games right now than the titan x did with 2015 games. That means that the amount of work the cpu has to do has stagnated. 

 

The point is, faster ram will only really help when you cpu is already near its limit. The game is running well over 60fps already and anyone expecting to play at 120/144hz should not be getting an i5 anyway. 

If you did that benchmark again with a 7700k or a 6700k you'd see a considerably smaller performance boost. 

 

 

 

[Humbug]

5 hours ago, HKZeroFive said:

 

To be honest, this is a double edged sword for me. I recognised that there was something off with Ryzen's performance seeing how its IPC in multithreaded non-gaming workloads wasn't that far behind than Intel's offerings. It's good in the sense that with the right amount of development and effort, we can see large performance gains such as this in current and future titles. It's bad in the sense that in today's gaming developer world, the culture has always seemed to be put in as little effort as possible for a maximum amount of money; most developers will not put in the time and effort to patch their already existing games for Ryzen

so far the only other to patch an existing game is valve who patched dota2 for Ryzen. So both games which got patched are multiplayer games which means the games are still played by thr community and people benefit from more performance. On the other hand do not expect single player games to get patched because the devs normally leave those alone once the initial rush dies down, unless it's a game breaking bug. Also some devs like Blizzard who even though they maintain a big game (starcraft2) are not technically inclined so they will not spend time doing things like CPU optimizations or better multicore utilization, they only focus on gameplay stuff.

 

I think what matters here is just the proof of concept... so we now know that future games can and will be optimized for Ryzen. Now that devs have these CPUs in hand... All these years devs were doing all their work and optimization targeted at intel CPUs. Things were hand tuned to work well on intel, now they will be hand tuned to work well on both intel and AMD ryzen. At least AMDs dev program should make sure of that....

 

 

[Doobeedoo]

It's definitely nice bump in gains. Maybe we won't see as such improvements in current games but potentially in newer ones though. None the less, improvements are welcome and it can only get better not worse so.

Also, let's remember how in general games for quite some time aren't really praised for being optimized really, and that being for HW that's already out and known.
Them working with Bethesda for bringing optimization with Ryzen and Vega is nice though. Can't wait to see in general how games run in future. Not just getting optimized for these new architectures, but in general being optimized and released with good performance. Not bad coding and bad ports. Also utilizing new APIs along better, learning it and understanding it more, put more effort into good coding to truly bring it's potential out.

[Citadelen]

2 hours ago, Misanthrope said:

-snip-

Initial devoper work is promising, that's what we do know, as for everything else, well, your view is no more valid than mine as there both based on assumptions.

[alamox]

7 hours ago, WereCat said:

So the baseline is i7 6900k at stock speeds with 2400MHz DDR4?

Then they take stock clocked 1800X which has similar per core performance on 2400MHz RAM and compare it to 6900k and say "This is up to 22% faster!", then they take 3200MHz RAM and say "This is up to 31% faster!".... why don't they also take 3200MHz RAM with 6900k and compare it with that?

Because from the graph I can only guess that all improvements come just from faster RAM.

 

EDIT:

Nvm.. I totally ignored the blue lines, though it still seems to me like the RAM performance was the main factor still.

 

i dont't think you understood the slides, the blue graph is before the patch, the orange graph is after the patch.

the 6900k is there just because it's there it's irrelevant, the slides show you that before and after patch, and also with low and high Ram, because Ryzen infinity fabric is tied to the speed of the Ram.

[WereCat]

10 minutes ago, alamox said:

 

i dont't think you understood the slides, the blue graph is before the patch, the orange graph is after the patch.

the 6900k is there just because it's there it's irrelevant, the slides show you that before and after patch, and also with low and high Ram, because Ryzen infinity fabric is tied to the speed of the Ram.

I do, but as I said...the 2nd graph confused me as it appears that the Ryzen is 22% and 31% faster than the 6900k.

Which is not true because the graph just shows the % improvements based on the 1st graph when comparing the improvements between game versions.

[Citadelen]

2 hours ago, Coaxialgamer said:

Spoiler

People keep bringing up that GF video. It's not even a very real world comparison. They used an overclocked titan x with an i5 at 1080p. No one was pairing a 200$ cpu with a 1000$ cpu back then, and neither are they now. 

 

Sure you could make the argument that the 1070 is just as fast for 400$, but gpu workloads have gotten significantly heavier since the  witcher 3 came out. 

 

I'd argue that the 1070 gives you less frames in modern games right now than the titan x did with 2015 games. That means that the amount of work the cpu has to do has stagnated. 

 

The point is, faster ram will only really help when you cpu is already near its limit. The game is running well over 60fps already and anyone expecting to play at 120/144hz should not be getting an i5 anyway. 

If you did that benchmark again with a 7700k or a 6700k you'd see a considerably smaller performance boost. 

 

They used an i5 to illustrate how in CPU bottlenecked situations RAM frequency can provide a rather substantial CPU performance uplift.

 

@WereCat The top section of the graph is tested in the high preset and it does not include a 6900K, the bottom section is in extreme present, and does include a 6900K.

Edited March 30, 2017 by Citadelen

[LukaH]

this is exciting! hope devs can do the same in other games. i5 in gaming my foot!

[Taf the Ghost]

5 hours ago, EminentSun said:

The pre-update benchmarks were nearly identical becoming a nearly 7 fps gap afterwards.

 

It seems likely many games have other issues with ryzen that have been overshadowing the performance delta from ram speeds. Leading people to dismiss the effect of ram speed. Hopefully we'll see more games gain this level of improvement as ryzen5/3 rolls out.

All AAA games are tuned to Intel's quirks.  Ryzen has different quirks.  Tuning around those quirks and suddenly performance looks in line with IPC, Clock Speed & Thread count.  Though this is really much more of a Game Engine issue, which will be adjusted over time.

 

There's still a few odd ones, though everyone using Tomb Raider really needs to retire that benchmark. Like a few other things that have come out with all of the benchmarking, some games are just not worth using for Benchmarks because their run-to-run consistency is terrible.  The average of Bad Information is just Averaged Bad Information.  (On Intel + Nvidia, Tomb Raider has these really high-peak effects, making the averages look a lot better than the frametimes show as a baseline. 7700k is still better than Ryzen in the game, but it's a test that doesn't show much.)

 

2400 to 3200 RAM speed showing about the 7-8% improvement in another game. That's been pretty consistent in a lot of applications.

[Speaker1264]

20 hours ago, Misanthrope said:

If it was just "not all" or "not most" I wouldn't have such a problem but the more fair statement would be "Almost no other games" will receive patches or launch with Ryzen optimizations.

 

AMD just doesn't has the market share it makes no sense to optimize for them beyond "good enough" which is "make it run 60 FPS 1080p at decent but not max settings"

 

For AMD to turn the tide they need to move more units and then maybe you'll see Ryzen optimizations reach a small niche of games much like Radeon optimized games that are around which are still a minority but one potentially worth considering when purchasing hardware.

 

Dude, you realize that most games were optimized for Intel systems and not Ryzen systems because Ryzen didn't exist when they were optimizing their game? This Ashes of the Singularity optimization just proves that all previous game benchmarks don't mean diddly squat, unless you never plan on playing a new release. All new games coming out should now be getting optimizations for both Intel and Ryzen, and the benchmark margins for up-coming game releases will shrink or disappear completely when compared to earlier benchmarks that were done with games that were optimized for Intel but not Ryzen.

 

For this reason when the Ryzen R5's are getting benchmarked I'm only going to be looking at the Ashes of the Singularity benchmarks to compare performance, since it's the only game currently optimized for both Intel and Ryzen, imo.  It's the only game where we will be able to see Ryzen's true performance, and that is the true performance that we are likely to see in future game releases.

[MandelFrac]

17 hours ago, Citadelen said:

The point is that gains of these magnitude are possible. They got this out in a few weeks, I don't see it being to hard for future games to be patched for AMD CPUs. Of course it remains to be seen if they do, but I do not think it's as bad as you make it out to be.

Since it comes down to nothing more than core affinity to keep the threads and jobs grouped together such that reaching across to the other CCX's L3 cache is minimized. Other than that the optimization is almost exactly what you would do for Intel. B/C of that double-sized L2 cache, you could change the block/shard sizes for the CPU-side physics and AI algorithms and potentially get more gains from Ryzen, but otherwise there's no real optimization to do other than group the threads together so minimize CCX crosstalk. It's not hard.

[MandelFrac]

1 hour ago, Speaker1264 said:

 

Dude, you realize that most games were optimized for Intel systems and not Ryzen systems because Ryzen didn't exist when they were optimizing their game? This Ashes of the Singularity optimization just proves that all previous game benchmarks don't mean diddly squat, unless you never plan on playing a new release. All new games coming out should now be getting optimizations for both Intel and Ryzen, and the benchmark margins for up-coming game releases will shrink or disappear completely when compared to earlier benchmarks that were done with games that were optimized for Intel but not Ryzen.

 

For this reason when the Ryzen R5's are getting benchmarked I'm only going to be looking at the Ashes of the Singularity benchmarks to compare performance, since it's the only game currently optimized for both Intel and Ryzen, imo.  It's the only game where we will be able to see Ryzen's true performance, and that is the true performance that we are likely to see in future game releases.

Actually it doesn't mean that. If other games had their threads grouped in such a way that CCX crosstalk was already minimal, then there's no big gain to be had like there was here. There are some gains to get from having a double-sized L2, but they won't be anywhere near as big.

 

And let's not kid ourselves. The games industry is a good 8 years behind on optimization techniques CPU side and is 5 years behind in instruction sets (still little to no AVX usage even though it doubles the throughput over SSE which IS in use today).

[Citadelen]

6 minutes ago, MandelFrac said:

Spoiler

Since it comes down to nothing more than core affinity to keep the threads and jobs grouped together such that reaching across to the other CCX's L3 cache is minimized. Other than that the optimization is almost exactly what you would do for Intel. B/C of that double-sized L2 cache, you could change the block/shard sizes for the CPU-side physics and AI algorithms and potentially get more gains from Ryzen, but otherwise there's no real optimization to do other than group the threads together so minimize CCX crosstalk. It's not hard.

 

In the quote in the OP they talk about how the instruction scheduler is also optimised for Intel, this would have to be patched as well, the improvements we've seen took them around 400 hours to implement.