AMD says seeding of dev kits will boost Ryzen gaming performance

[CAT-THE-FIFTH]

Yea sorry i skipped over those images.

TBH there does seem to be some misinformation and confusion WRT Ryzen information and what some people at AMD are saying (see only X chips come with XFR) and i suspect this SMT thing is the same, it's almost like Robert the technical marketing guy doesn't understand or has misinterpreted the question he was being asked, maybe he was referring to DX12 titles as (afaik) the scheduling and thread assignment under DX12 would/could be the responsibility of the game dev.

I really hope that is the case.

I still want to see if some site can disable two of the cores on a R7 1800X and the SMT and benchmark it against a Core i5 7600K and a Core i7 7700K. If it can give the former a good fight,then there is some hope here,since most people I know spend under £300 on a CPU anyway.

But this is what I expressed months ago,that I really hoped AMD didn't have some major issue during a launch like literally every CPU and GPU launch they had in the last few years - sadly it was very optimistic and somewhat deluded hope on my part. After the Phenom and FX8150 I should have realised AMD second releases always tend to be where things come together as a whole. I expect Ryzen 2 probably will be pretty great as a whole.

Intel OTH who probably are very worried about Ryzen have just been thrown a bone at least in one area,so I suspect the issue is music to their ears.

I was really hoping to actually have an AMD CPU in my system after being on Intel,but I know very well it might be another Intel one by the time all these issues are fixed,which makes me a very sad moose.

[Otherhand]

Good news. Yet again, I've just gone with Intel because I'm history's greatest monster but it's good to see AMD challenging them again. Hopefully these anomalies will be quickly levelled.

[Ozaron]

I was really hoping to actually have an AMD CPU in my system after being on Intel,but I know very well it might be another Intel one by the time all these issues are fixed,which makes me a very sad moose.

Is a 5-15% performance drop at 1080p (not even in all games, and especially after some games & Windows get optimized) worth being a sad moose for? (And quite possibly one which is slightly more out-of-pocket?) Just a thought.

What IPC or other advantage do you expect a new generation Intel lineup to bring, anyway? Their stagnation is exactly the reason people have backed AMD throughout Ryzen delivery.

Perhaps being idealistic is worth more than being realistic, for one CPU generation.

[CAT-THE-FIFTH]

Is a 5-15% performance drop at 1080p (not even in all games, and especially after some games & Windows get optimized) worth being a sad moose for? (And quite possibly one which is slightly more out-of-pocket?) Just a thought.

Yes,because it's less performance against the Intel equivalents and I have to be objective in having a CPU which I know works reliably.

ATM,even if I get a Ryzen CPU its not going to be anytime soon - it needs months of work on the SMT issue,motherboard BIOSes or even the fact I use a mini-ITX system. But that means Coffee Lake is edging closer and closer. Hopefully AMD by then will have the Ryzen APUs with Zen+ cores with greater IPC(might be just the launch ones).

The worst thing is if I go Intel again I feel like I am stealing from the poor to feed the rich.

[jimbouk]

Isn't this also showing that the majority of games also run faster with Hyperthreading disabled on the i7-6900K?

Less drastic and impact (except on Watch Dogs 2 where it's within 0.5%) but it's there none the less.

[Ozaron]

ATM,even if I get a Ryzen CPU its not going to be anytime soon - it needs months of work on the SMT issue,motherboard BIOSes or even the fact I use a mini-ITX system. But that means Coffee Lake is edging closer and closer.

Really wanted to build a Ryzen system for someone myself, but agreed that there's no point doing it just yet. I'd want 1600X for a start, rather than the R7s, so BIOS updates and SMT fix are likely to arrive before time anyway. It's... probably not a problem.

But I'm thinking within 3 months this will be a very solid system from a performance point of view. Will Intel do anything special with Coffee Lake? For once they will have a reason to, but what indication is there that they have any tricks to pull out of the expensive bag? Rather than wait for them to trot 10-15% better theoretical performance out and perhaps more cores for marginally less, once Ryzen is sorted could be a good time to put money where our mouths are.

[Bagpuss]

So once again a promise of jam tomorrow from AMD.

Sorry but that sounds too close to what they said about Bulldozer.

I'm not saying these chips won't be a good buy for some people but right now in gaming terms they're no match for the 7700K and are unlikely to ever be.

If you want to build a gaming Ryzen rig you'd be crazy to get these R7's, far better to wait and see what the R5's bring to the table.

[scaryjim]

So, not looking good for Windows 7 users who are looking at Ryzen and want SMT.....how predictable :S

Actually, we don't know if the problem also exhibits under Windows 7. That'd be an interesting test of how much it's down to the game engine and how much it's down to the OS scheduler.

I've got a spare copy of Win 7 somewhere if anyone would like to send me a Ryzen 7/mobo/RAM combo to test....

[CAT-THE-FIFTH]

Isn't this also showing that the majority of games also run faster with Hyperthreading disabled on the i7-6900K?

Less drastic and impact (except on Watch Dogs 2 where it's within 0.5%) but it's there none the less.

Outside one or two games,the Intel results are within a margin of error,but with AMD many are not.

The issue is that the KL CPUs,seem to not have issues with SMT scaling as a whole,as you only need to see reviews from websites like the Digital Foundry where the Core i7 is generally faster in minimums and frame-times than a Core i5.

A 10% performance drop on average is enough to make a Ryzen CPU look more like an IB CPU than something more like Haswell and is better to switch it off. If it were the odd game,I might dismiss it for AMD too,but checking only three reviews hinted that at least 12(maybe more) games have SMT issues.

Really wanted to build a Ryzen system for someone myself, but agreed that there's no point doing it just yet. I'd want 1600X for a start, rather than the R7s, so BIOS updates and SMT fix are likely to arrive before time anyway. It's... probably not a problem.

But I'm thinking within 3 months this will be a very solid system from a performance point of view. Will Intel do anything special with Coffee Lake? For once they will have a reason to, but what indication is there that they have any tricks to pull out of the expensive bag? Rather than wait for them to trot 10-15% better theoretical performance out and perhaps more cores for marginally less, once Ryzen is sorted could be a good time to put money where our mouths are.

On the flip-side at least the R5 1600X even in pure 6C mode will have 50% more cores than a Core i5 7600K. I just hope Intel does not drop Core i7 7700K pricing.

Well regarding Coffee Lake,its 14NM,so Intel can probably try and get it out quicker if they feel 8C Ryzen is being a threat.

The issue is the muddled AMD response in the AMA - which was basically SMT will be fixed in games when it happens. The probably is how long will it take to happen??

A month? 3 months?? 6 months?? 9 months?? The issue is that Coffee Lake will be much closer by then and it only takes Intel to price it at current Core i7 7700K levels,and the Coffee Lake Core i7 7700K equivalent at Core i5 7600K level to be enough to counter Ryzen in games.

If SMT works far more reliably in games by then,Intel will have far more problems,but if it doesn't Intel will probably win that one.

[Sumanji]

Its almost like they didn't even bother trying to engage with some devs before launch...

Its like they have had NOT BOTHERED to seed devs with ANY info about Ryzen in the last few months.

I think this is the key problem... surely they've had months (even years?!) to share details with game devs?

[Iota]

The issue is the muddled AMD response in the AMA - which was basically SMT will be fixed in games when it happens. The probably is how long will it take to happen??

Looking at the benchmarks for games, I've seen no reason to upgrade yet from my aged and slightly OC 2600k. I can still wait and see what happens, primarily because it isn't going to make a jot of difference to my gaming at 1080p. When I upgrade I tend to upgrade for the next 5 years (with exception to GPU upgrades in that time).

It entirely depends currently on pricing for more than 4 cores, I say that because I will go with whoever has the cheaper 8 core processer on my next upgrade, so if Intel remain in the over-priced gouging area they are and don't lower pricing to compensate for Ryzen, I couldn't care less if the performance of Ryzen is 5-20% off where Intel is considering cost.

[kompukare]

I think this is the key problem... surely they've had months (even years?!) to share details with game devs?

Well, some details were already known for ages as they told the Linux dev community.

But the biggest thing seems to be Windows' scheduler. That should sort out the contention issues and that turning off SMT often makes games faster (turning off HT on Intel processors still affects games too after all these years but not as much).

However, this does nothing for game developers who are not good at using threading. For instance, giving Bethesda some Ryzen systems or guiding them on how to code for them isn't going to magically turn them into threading experts. There is a reason I nickname them Bugsoft as their games tend to come out half finished and unlike DICE or even Id (which Bethesda publish) their programming has never been that great. It's just that for those like me who enjoy a large customisable open-world RPG, they are pretty much the only game in town.

Wonder if this seeding Ryzen dev kits has anything to do with a new console? Guess Project Scorpio is too soon to have Zen. But of course a new console using Zen is what will get devs to really optimise for their CPUs.

[Xlucine]

Wonder if this seeding Ryzen dev kits has anything to do with a new console? Guess Project Scorpio is too soon to have Zen. But of course a new console using Zen is what will get devs to really optimise for their CPUs.

Consoles have all been using AMD cores for years, and the old AMD cores are easily beaten by intel in games - why would ryzen in consoles be any different?

[kompukare]

Consoles have all been using AMD cores for years, and the old AMD cores are easily beaten by intel in games - why would ryzen in consoles be any different?

Because the cluster of 2 x 4C/4T Jaguar cores is not the same as 8C/16T Ryzen and the most important take away from Zen so far is something people seem to have forgotten: just because something is x86 (or x64/AMD64) does not mean that there are not optimisations which can be made for the specific architecture. The same would apply if a console used Intel Atom cores: the idiosyncrasies of Atom would mean that optimising for Atom is not the same as optimising for i5/i7 Core mainstream or even i7 HEDT. Cache, memory latencies, logical topologies, SMT/CMT or lack thereof, new instructions, x87/SSE/AVX all make a big difference.
Anyway, unsure if Microsoft would be willing to pay much for the CPU on Scorpio anyway and I don't think AMD are willing to give Zen away like they were Jaguar. For the true next-gen consoles, 4C/8T or 8C/8T is far more like. Microsoft or Sony are far more likely to spend silicon area on Vega or Navi than the CPUs. Still at roughly 100mm², a 4C/8T @ 3GHz+ Zen might be possible and it would be strategically advantageous to AMD to have console games optimised for their architecture.

[CAT-THE-FIFTH]

If the XBox Scorpio uses Ryzen instead of Jaguar that would be a biggie IMHO!! It means literally every major title outside the few PC exclusives would have to be optimised for Ryzen.

[Xlucine]

Because the cluster of 2 x 4C/4T Jaguar cores is not the same as 8C/16T Ryzen and the most important take away from Zen so far is something people seem to have forgotten: just because something is x86 (or x64/AMD64) does not mean that there are not optimisations which can be made for the specific architecture. The same would apply if a console used Intel Atom cores: the idiosyncrasies of Atom would mean that optimising for Atom is not the same as optimising for i5/i7 Core mainstream or even i7 HEDT. Cache, memory latencies, logical topologies, SMT/CMT or lack thereof, new instructions, x87/SSE/AVX all make a big difference.
Anyway, unsure if Microsoft would be willing to pay much for the CPU on Scorpio anyway and I don't think AMD are willing to give Zen away like they were Jaguar. For the true next-gen consoles, 4C/8T or 8C/8T is far more like. Microsoft or Sony are far more likely to spend silicon area on Vega or Navi than the CPUs. Still at roughly 100mm², a 4C/8T @ 3GHz+ Zen might be possible and it would be strategically advantageous to AMD to have console games optimised for their architecture.

Jaguar is a low power construction core, right? So with everyone optimising games for a jaguar based SOC, construction was easily beaten by intel's core