Read more.Attempts to explain the 1080p gaming deficit observed by several reviewers.
Read more.Attempts to explain the 1080p gaming deficit observed by several reviewers.
To say it's caused by "optimization of the game uniquely to Intel platforms" seems more like sour grapes to me, it's not that games have been optimised uniquely to Intel platforms, that's unless AMD are positing that developers never optimised for any AMD platforms, it's that games have not been optimised for the new Zen architecture, yet.
CAT-THE-FIFTH (03-03-2017)
One theory I have read somewhere is that task schedulers may be shuffling threads from one four-core module to another. Since L3 cache is not shared across the modules (8+8 arrangement) this will cause cache misses and hit performance. I am also sure that memory latency will improve going forwards as motherboard manufacturers get to grips with the IMC.
All told I am not actually too worried. I don't believe it can do so well in 3DMark Firestrike/Timespy and yet be bad for games in general, even if that is a synthetic benchmark. Now that Ryzen is in the open, games programmers will pay more attention in time for the more game-oriented CPU releases.
Still, I have always thought there should be an option in operating systems along the lines of "disable SMT for this program". Plenty of software loses performance when using SMT - many EPS modelling HPCs/clusters have to disable SMT globally. It would be nice to keep the SMT cores for background tasks.
Last edited by CAPTAIN_ALLCAPS; 03-03-2017 at 11:53 AM. Reason: Clarification
CAPS LOCK IS NOT A BUTTON IT IS A WAY OF LIFE.
CAT-THE-FIFTH (03-03-2017)
It's actually sounds like a really plausible scenario. First Ryzen is a whole new architecture so its unlikely any game could be optimized for it. Secondly, if you see the reviews the benches show that the single core performance is just there where AMD said it will be, on par with Broadwell, and the multi core performance is even better given the SMT is better scaling than intels HT. So the grunt is there in the CPU, it won't disappear by starting a game instead of a benchmark. It needs optimization. I expect alot of patches in the next 1-2 months and a lot of performance from those in games.
CAT-THE-FIFTH (03-03-2017)
Can't help feeling compiler optimisations might also come into play. Intel spend a lot on compiler tech and integrating patches etc into the main line compilers, and that must surely give an advantage. With Zen so new, perhaps the toolchains simply haven't been updated as aggressively for AMD's architecture?
CAT-THE-FIFTH (03-03-2017)
I'm not saying it's not but they way AMD seem to be couching it is that developers have optimised for Intel platforms while ignoring AMD when that seems implausible at best, like i said it's not that developers have only ever optimised for Intel platforms while ignoring AMD, it's that they've not had time to optimise for the new Zen architecture.
It seems like the sort of answer you'd expect from a PR representative whose more concerned about protecting the brand than letting people know they recognise the problem and are working to fix it.
Last edited by Corky34; 03-03-2017 at 11:57 AM.
CAT-THE-FIFTH (03-03-2017)
It's a shame that the raw performance didn't show itself in consumer applications immediately, but the fact we found it in a variety of benchmarking tools - and the fact AMD seem confident they can organise fixes to the biggest concerns with relatively short shrift - means that by the time Ryzen 3 and 5 come along, we should be able to see them at their best. And damn is that what they need right now in the consumer space.
I'm sure they hoped that the benchmarking tool raw horsepower results would please the prosumers and small server admins, and that the general public either wouldn't be interested or would have their hearts set on 8 cores anyway and be happy regardless of short term optimization issues. It didn't quite turn out that way. AMD just need to focus on actively connecting with the customers now and showing them it's all in hand.
Corky34 (03-03-2017)
Its irrelevant why the reason is - if it was the odd game we would not care so much,but this is a major problem.
Its not isolated. Other websites tested - it performance regression in loads of titles with SMT. Thats at least 12 major titles affected by SMT problems,and some are not that small.
Just because a problem is systematic does not mean it is dead serious. AMD of course would always downplay this misfortunate circumstance, but it might be that they actually genuinely don't think it will be a problem for very long. Unrelated, but I'm hearing reports of people getting significantly better performance out of Ryzen chips on updated motherboards, too. Something to do with BIOS improvements.
On another note, are spoiler handles not a feature around here? My broken scrollwheel didn't appreciate this.
Look at my previous posts - 12 games at least show regressions.
Its a tepid response - if it needs to be fixed on a game by game and engine by engine case,how long is it going to take?
Even AMD is saying it will take as long as it needs to take for the devs to do the work.
If its not a general scheduling issue under Windows - how many games is AMD going to make sure will be fixed??
Every major game in the last few years?? Every one going forward??
Its almost like they didn't even bother trying to engage with some devs before launch.
The 4C/8T and 6C/12T models are the ones worst affected by this - so how can you recommend one for a gaming build and tell people,AMD will get devs to "fix" SMT issues at "some point".
Most people will see it as a bug and just buy Intel.
Its like they have had NOT BOTHERED to seed devs with ANY info about Ryzen in the last few months.
Edit!!
What if AMD cheats and makes sure there are fixes only for some popular benchmarking titles used in games??
Last edited by CAT-THE-FIFTH; 03-03-2017 at 12:50 PM.
It does when 12+ MAJOR games are affected. What about all the titles which are not common test titles? Plenty of popular games don't get tested.
Look at the screenshots of the AMA - AMD is saying its fixed when its fixed.
That means game by game and engine by engine.
How long is that is going to take??
There is no time-frame.
So whilst we are setting around for AMD to get the issue fixed,by that time Intel probably will start to get new CPUs like Skylake X and Coffee Lake out which will have far more predictable performance.
This makes the CPUs really unreliable for gaming once you get down to the 4C/8T and 6C/12T models and Intel only needs to drop Core i7 7700K pricing to pressure the 6C/12T R5 1600X.
This was the same with TLB except TLB didn't affect gaming.
Edit!!
No spoiler function here AFAIK.
Last edited by CAT-THE-FIFTH; 03-03-2017 at 12:54 PM.
Yes and I'm not denying there are teething problems, however from what i can tell the SMT problem is not a game developer thing as a Linux kernel commit was submitted in early February and simultaneous multithreading is the responsibility of the OS (afaik), it's worth noting that Microsoft skipped February's patch Tuesday so for all we know the SMT patch for Windows should have already been released, we won't really know until the 14th.
CAT-THE-FIFTH (03-03-2017)
Yeah,but somebody asked AMD that directly - look at the images above. They said it wasn't a windows problem but need a game by game optimisation of the "issue".
Now,it would be quite funny if a windows scheduler update actually solved it in many cases,though,wouldn't it?!
So, not looking good for Windows 7 users who are looking at Ryzen and want SMT.....how predictable :S
Main PC: Asus Rampage IV Extreme / 3960X@4.5GHz / Antec H1200 Pro / 32GB DDR3-1866 Quad Channel / Sapphire Fury X / Areca 1680 / 850W EVGA SuperNOVA Gold 2 / Corsair 600T / 2x Dell 3007 / 4 x 250GB SSD + 2 x 80GB SSD / 4 x 1TB HDD (RAID 10) / Windows 10 Pro, Yosemite & Ubuntu
HTPC: AsRock Z77 Pro 4 / 3770K@4.2GHz / 24GB / GTX 1080 / SST-LC20 / Antec TP-550 / Hisense 65k5510 4K TV / HTC Vive / 2 x 240GB SSD + 12TB HDD Space / Race Seat / Logitech G29 / Win 10 Pro
HTPC2: Asus AM1I-A / 5150 / 4GB / Corsair Force 3 240GB / Silverstone SST-ML05B + ST30SF / Samsung UE60H6200 TV / Windows 10 Pro
Spare/Loaner: Gigabyte EX58-UD5 / i950 / 12GB / HD7870 / Corsair 300R / Silverpower 700W modular
NAS 1: HP N40L / 12GB ECC RAM / 2 x 3TB Arrays || NAS 2: Dell PowerEdge T110 II / 24GB ECC RAM / 2 x 3TB Hybrid arrays || Network:Buffalo WZR-1166DHP w/DD-WRT + HP ProCurve 1800-24G
Laptop: Dell Precision 5510 Printer: HP CP1515n || Phone: Huawei P30 || Other: Samsung Galaxy Tab 4 Pro 10.1 CM14 / Playstation 4 + G29 + 2TB Hybrid drive
I really want to see some review site disable cores and compare AMD 4C vs Intel 4C and AMD 6C vs Intel 4C/8T for gaming.
SMT seems fine if you are not gaming though. For gaming its a work in progress best ignored until its reliable enough to be used just like with the first P4 models which had it.
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, or 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, whereas pre-DX12 scheduling and thread assignment is the responsibility of the OS, however that's just my best guess as I've not look into it to much.
There are currently 1 users browsing this thread. (0 members and 1 guests)