Thursday, March 30th 2017

Ashes of the Singularity: Escalation Update Brings Improved Performance to Ryzen

Some outlets are reporting that Stardock's Ashes of the Singularity is about to receive the much-referred-to patch that allows for improved performance on AMD's Ryzen line of processors. If you remember, rivers of ink flowed regarding AMD's Ryzen performance in gaming, with its monstrous, high-performance 8-core, 16-threaded design sometimes delivering performance below expectations. At the time, AMD clarified how Ryzen is a distinctive CPU architecture, similar yet fundamentally different from Intel's x86 implementation, promising upcoming patches from game developers that would allow Ryzen's architecture to truly deliver.

After Creative Assembly and Oxide Games vouched to improve Ryzen support, Oxide seems to be the first developer with a patch available (from version 25624 to 26118) that improves performance by up to 30%. Reportedly, it took the developers around 400 work-hours to improve the game code in respect to its execution on AMD hardware.
The update leads to AMD's Ryzen 7 1800X effectively edging out Intel's 7700K in all scenarios, whose better overall gaming performance is the prevailing argument towards preferring the Intel solution. Even so, it would seem that Intel's 6900K still edges out the Ryzen 7 1800X on the Extreme, 1080p preset. But its 7 FPS lead over the 1800X when paired with 2400 MHz DDR4 memory does come with a cost of more than double the 1800X's.

Performance improvements are greater when the Ryzen processors are paired with higher-performing memory (a nod to the way the CCX's Infinity Fabric inter-CCX communication is dependent on memory throughput for increased performance). It would seem that this update improves Ashes' handling of work threads on the Ryzen CPUs, limiting the amount of workload that hops between CCXs - which incurs in a heavy latency penalty for Ryzen processors. Thus, the load on Infinity Fabric would be alleviated, allowing for its increased throughput to carry only game-critical data between both CCXs, whilst not having to also deal with performance-dropping, inter-CCX tread-hopping.

This is an interesting development, which some probably didn't think would actually happen - bold "future performance improvement" claims have been shouted on their way down from rooftops, after all. And while a single developer (out of two who committed to improving performance) doesn't represent an entire industry, it does give AMD credence in its promises of latent performance on their Ryzen CPUs. Best of all: this happens with no performance penalty for AMD's arch-rival Intel processors. We have to wait and see, but fingers crossed for Creative Assembly's changes (when they come, if they come) to join Oxide Games' own improvements opening up the industry to some relevant, platform-specific improvements (and at the same time, agnostic, as in, with no penalties for Intel).

AMD's partnership with Bethesda is looking more and more interesting by the day.
Sources: Tom's Hardware, PC Perspective
Add your own comment

39 Comments on Ashes of the Singularity: Escalation Update Brings Improved Performance to Ryzen

#26
pantherx12
RCoonIt's a little depressing that the only reason people have heard of/buy your game is because of benchmarking. I literally do not know anyone that owns this game aside from myself and benchmarkers, and most of us have never actually played it.
Hey it worked for crisis, completely average shooter but it was so over the top at the time that people got it just to see if they could run.
dozenfuryIt's a nice jump from a patch alone, but I own AOTS and it's always seemed more like an AMD gfx demo than a game. I'm also a bit skeptical of other devs spending 400+ hours to optimize their games for Ryzen. The AOTS devs are very connected to AMD so they are definitely a special case. I'm far more interested in the performance in more mainstream games like BF1, Rise of Tomb Raider, ME: A, etc. that aren't significantly subsidized by 1 vendor to get special treatment and give misleading results. Granted most games get that to a smaller degree with "Best played on...", but not usually to the degree of AOTS.
400 hours could be distributed over 400 employees potentially so the work can be done In less than 24 hours.

Still even 4 people working on it that's only two weeks of work :)
Posted on Reply
#27
efikkan
dozenfuryIt's a nice jump from a patch alone, but I own AOTS and it's always seemed more like an AMD gfx demo than a game. I'm also a bit skeptical of other devs spending 400+ hours to optimize their games for Ryzen. The AOTS devs are very connected to AMD so they are definitely a special case. I'm far more interested in the performance in more mainstream games like BF1, Rise of Tomb Raider, ME: A, etc. that aren't significantly subsidized by 1 vendor to get special treatment and give misleading results. Granted most games get that to a smaller degree with "Best played on...", but not usually to the degree of AOTS.
Yes, that game is nothing more than a "tech demo"/benchmark. "No one" uses it except for benchmarking.

In terms of rendering performance, the only way to "optimize for Ryzen" would be to remove bloat; removing wasted CPU-cycles. 400 hours to optimize probably means they spent 50-100 hours browsing through the code, a few hours tweaking a few functions and then the remaining in QA. They have clearly found some low-hanging fruit, no one will do a major overhaul of production code for something minor, and that would require >10k hours. Most games would not have enough low-hanging fruit to help out enough.
pantherx12400 hours could be distributed over 400 employees potentially so the work can be done In less than 24 hours.

Still even 4 people working on it that's only two weeks of work :)
You clearly have never worked on a large software project. Even if you had 400 developers in a code base, just the overhead of starting the work would consume more than 400 hours…
Posted on Reply
#28
pantherx12
efikkanYou clearly have never worked on a large software project. Even if you had 400 developers in a code base, just the overhead of starting the work would consume more than 400 hours…
Whilst to assumption about me is correct I did use the word potentially in my statement.

That's why I offered up the more realistic example of a four person team but still I accept I could be wrong about that as well.

Still though they've achieved this improvement with about a month of time with Ryzen, 400 hours over a month isn't all that much especially for a studio that only seems to have the one game to work on.
Posted on Reply
#29
efikkan
pantherx12Still though they've achieved this improvement with about a month of time with Ryzen, 400 hours over a month isn't all that much especially for a studio that only seems to have the one game to work on.
Yes, one of my points was that 400 hours is "nothing" for a big project. As the project grows larger, changes becomes more time consuming and risky, so in 400 hours you can't really do a lot with a game engine. Larger projects always have to do a lot of QA for all changes. So if you have found something that yields a measurable performance gain it has to be some low-hanging fruit. Most game engines wouldn't have that, so you can't expect most developers to spend 400 hours and get similar gains.
Posted on Reply
#30
pantherx12
efikkanYes, one of my points was that 400 hours is "nothing" for a big project. As the project grows larger, changes becomes more time consuming and risky, so in 400 hours you can't really do a lot with a game engine. Larger projects always have to do a lot of QA for all changes. So if you have found something that yields a measurable performance gain it has to be some low-hanging fruit. Most game engines wouldn't have that, so you can't expect most developers to spend 400 hours and get similar gains.
I'm not expecting performance improvements like this across the board, it would just be nice to see other games that are clearly underperforming get some patches.

The games where performance is adequate likely won't be touched.
Posted on Reply
#31
HTC
According to this, the AotS improvements do not originate from thread scheduling fixes.

This means that there's the possibility that there may be more performance gains should a windows patch covering thread scheduling be released later on.
Posted on Reply
#32
CounterZeus
'Tech demo' or not, I'm glad they show it's possible within reasonable amount of work/time. Ryzen has a lot of computing power spread across those 8 cores and it's nice they have it translated to real world gaming performance. This can only benefit us all.
Posted on Reply
#33
john_
This is an indication, but not a proof, about what we might see in the future, in most AAA titles. And that's consistent performance from Ryzen, not some odd cases where Ryzen performs like an overclocked FX. I think this example makes it easier to go with AMD over Intel, even if gaming is important for the person buying a new computer. Only people who need the absolute maximum of frames per second score, should still stay with Intel.
Posted on Reply
#34
HD64G
Manu_PTNo one ever said ryzen is bad for gaming. Is just that Intel similar and lower priced are better.
I can remember of many posts talking about Ryzen being weak for gaming. And they tended to forget that a new platform is getting better with newer BIOS and update in OS and games. Ryzen has another good point also: RAM speed scaling due to its arch.
Posted on Reply
#35
Nergal
Very good job by AMD, if I hadn´t bought me a 6700 last year, I would be buying a Ryzen right now.
Posted on Reply
#36
notb
So, once again, this awful game emerges. :(
I'm really not surprised Oxide decided to make this patch. They're struggling to sell this game at all... Yet, it always was a favorite of AMD GPU owners, so I'm not surprised Ryzen crowd also had to be taken care of.

I'm sorry to say this, but I don't think these AoS benchmarks are doing any good to Ryzen marketing strategy.

We know that most games today are hardly benefiting from more than 4 cores.
AMD released Ryzen saying that in games it's on par with 6900K (costing twice as much). And it is true, although most graphs ignored the fact that a cheaper 7700K is also in the same league gaming-wise (sometimes faster...).

Now we're back to AoS, which clearly has been optimized for more than 4 cores since release.
But what happened?
Even after this patch Ryzen looks like a competitor to 7700K which has half the cores. 6900K is way faster...

So maybe it's not about games not being optimized for 8 cores? Maybe Ryzen is simply fairly slow and difficult for coders - it only has the multi-thread advantage in some price segments.
At this point I'm quite interested how Ryzen will look when more games make use of 8C/16T potential. :P

BTW:
This test was made using the "Crazy preset" which is said not to change much in terms of gaming experience - it's basically a built-in benchmark. The original Ryzen review used the "Extreme preset" in which the gap between 7700K and 6900K was much wider. I'd prefer to see that one remade - to check where the Ryzen lands this time.
Posted on Reply
#37
Frick
Fishfaced Nincompoop
pantherx12Hey it worked for crisis, completely average shooter but it was so over the top at the time that people got it just to see if they could run.
Two things: One. I am not a shooter fan, but I still finished Crysis. TBH I loved it. It was mediocre on normal difficulty, but if you turned it to the highes difficulty, the game for some reason completely changed and became glorious. It felt designed for the highest difficulty. Everything fell in place then. Two. It ran on everything, not just on the highest settings. I played the MP Demo on an Athlon 64 3000+ at 2.4Ghz, 1GB RAM and a Radeon x1950pro, on 1280x1024 on High settings. I know exactly what you mean though, I just want things to be correct.
Posted on Reply
#38
thesmokingman
efikkanYes, that game is nothing more than a "tech demo"/benchmark. "No one" uses it except for benchmarking.
Tech demo or not the game is actually fun once you get over the learning curve. I used to play it a lot last year when I had more free time. Then the game got more and more advanced, the ai is too smart now, or its effing cheating? lol Then I got too busy, my kid joined a mtb team and I literally have not played a game in a year. But ashes as I remember it is a large game, its for diehard warfare gamers on a massive scale. That's why not many play it because its not a simple romp thru game.
Posted on Reply
#39
notb
thesmokingmanTech demo or not the game is actually fun once you get over the learning curve. I used to play it a lot last year when I had more free time. Then the game got more and more advanced, the ai is too smart now, or its effing cheating? lol Then I got too busy, my kid joined a mtb team and I literally have not played a game in a year. But ashes as I remember it is a large game, its for diehard warfare gamers on a massive scale. That's why not many play it because its not a simple romp thru game.
I remember reviews of AoS that ended up in mediocre scores like 6/10 and they didn't really mention the high AI level as a major drawback. What they actually said is that this game is BOOOORING. :)

AoS looks and feels like a (heavily multi-thread) benchmark which was later turned into a game. So you can control this vast amount of units and the game is quite challenging technically, but there isn't much else. It lacks a story, design finesse and some brain-challenge as well (other than having to click a lot).

Overall, you could right that this game is too difficult/time-consuming for many to play. As you've said: in time people tend to have other things to do than just playing games.
The issue with AoS is that it also hasn't been BOUGHT very often in the first place. :)
Posted on Reply
Add your own comment
Nov 18th, 2024 22:31 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts