Further proof that Windows 10 is causing a perf hit with the Ryzen CPU

[zMeul]

1 minute ago, cj09beira said:

Don't you see that you are the one with bullshit logic,

a game from 2013 isn't multi-threaded == games don't use multiple cores

complete bullshit

if you go and look for old cpus you will see constantly that the ones with most cores will last the longest.

oh right?

multicore CPUs just arrived on the market, eh?

multicore CPUs have been around since 2005 and I have yet to see a purposely build game that multithreads efficiently

[79wjd]

1 hour ago, cj09beira said:

if we have two cpus at around the same frame rate in 1440p, and one is sitting at 98% utilization and the other at 48%, which is the more powerful one, which will be the best in the long one, yes the one which isn't at 98%, thats ryzen in a nutshell right now

No one (intelligent) is denying that a 1700 is more powerful than a 7700k. It obviously is, but that doesn't make it better for gaming. If games continue to prioritize single core performance, then a 7700k is better. If games begin to spread the load better across more cores, than a 1700 is better. But both of those futures are equally likely, therefore, the better option is the one that is better right now (which is a 7700k). If you choose to gamble on more cores becoming more important, then that is a personal choice and you are choosing to risk current performance in the hopes that future performance will be better. I would personally get a 1700 over a 7700k and take that gamble myself, but that doesn't make it any less of a gamble.

[Citadelen]

On 2017-3-5 at 3:00 AM, ARikozuM said:

I doubt this is an OS issue. 

It very much is, Windows 10 needs to have its thread scheduler patched for Zen, it swaps the programs cores are using between each Zen cluster, which means core 0 evicts some information it needs to the 8mb of victim cache in the first cluster, then Windows 10 moves the program to run on core 5, the program suddenly doesn't have the information it needs causing a huge latency spike and quite adversely affecting performance. This is why disabling SMT gives a boost in performance. Not only this, But it assumes each thread has 8mb of cache to work with, clearly the R7 does not have 128mb of cache, you can imagine the havok that would cause...

[cj09beira]

28 minutes ago, zMeul said:

oh right?

multicore CPUs just arrived on the market, eh?

multicore CPUs have been around since 2005 and I have yet to see a purposely build game that multithreads efficiently

there are a lot of games that use well multiple cores, if they didn't exist you would not be seeing a 7700k at 98% usage :-|

there isn't much more to be gained from ipc and clock speeds so the way to go is more cores, software will follow, slowly but it will,

don't forget even the consoles today have 8 cores.

23 minutes ago, djdwosk97 said:

No one (intelligent) is denying that a 1700 is more powerful than a 7700k. It obviously is, but that doesn't make it better for gaming. If games continue to prioritize single core performance, then a 7700k is better. If games begin to spread the load better across more cores, than a 1700 is better. But both of those futures are equally likely, therefore, the better option is the one that is better right now (which is a 7700k). If you choose to gamble on more cores becoming more important, then that is a personal choice and you are choosing to risk current performance in the hopes that future performance will be better. I would personally get a 1700 over a 7700k and take that gamble myself, but that doesn't make it any less of a gamble.

i don't believe that the likelihood is the same for both cases, its much more likely that it will start using more cores, because of the rise of dx12, vulkan, 8 cores on consoles, programing for more cores is starting to make sense as most people have at least 4 threads.

as i said the way forward in performance is core count so programing for it is the way to go, although its a gamble as to how much time it will take, but then again as people forget to mention by the time the bottleneck would start shifting to the cpu in normal cases the games will be better threaded.

[Jahramika]

On 3/4/2017 at 9:51 PM, Praesi said:

Ya. Give it time till Intel whipe the Floor with it again

How 5.3 Ghz next chip? Lol

[Jahramika]

On 3/4/2017 at 10:33 PM, ARikozuM said:

Just saying that AMD set a precedent with Witcher 3 (the latest that I can think of) and a driver update fixed everything. It seems too often that reactions such as "hur dur Windoes x sucks for gaming" when it could be as easy as let a programmer make a new driver to address the hardware for better performance.

Amd set X99 platform on its ass like it was suppose to do before optimization.

[ARikozuM]

2 hours ago, cj09beira said:

Snip

Just because a program uses multiple cores doesn't mean that it's multi-threaded. Physics calls aren't passed off to other cores. Each core is working on independent tasks until the data has to be consolidated. BF1, for example, uses multiple cores but the majority of calculations are single-threaded since the work occurs in sequence. 

2 hours ago, Citadelen said:

Snip

AMD could have very well made a driver for the OS given the amount of time they've had to prepare for the launch.

[cj09beira]

8 hours ago, ARikozuM said:

Just because a program uses multiple cores doesn't mean that it's multi-threaded. Physics calls aren't passed off to other cores. Each core is working on independent tasks until the data has to be consolidated. BF1, for example, uses multiple cores but the majority of calculations are single-threaded since the work occurs in sequence. 

AMD could have very well made a driver for the OS given the amount of time they've had to prepare for the launch.

by definition a game that uses multiple cores is multi-threaded, you can say they don't balance the load between threads, but doesn't mean its not multi-threaded

[Lynx]

There is one interesting (fantasy) theory from Wendell about why AMD give so little time for the software/gaming dev, microsoft, and motherboard manufacture to optimize their product for Ryzen.

Because of the stuff that happen in the past where AMD offer a faster CPU than Intel, and Intel decided to do some stuff that latter judged as unfair, AMD want to keep everything in secret, not give Intel any early hint about Zen architecture and its performance, give Intel little time to respond, so they could disrupt the market.

it's in the beginning of this video.

And also at around 20 minutes, Wendell said that in GTA V, his intel machine (kaby lake i7?) got stutter sometimes, while there's no such thing with Ryzen 7. But well, just watch Wendell in this show, I would just skip when the others are talking, because well, Wendell.

[laminutederire]

On Sunday, March 05, 2017 at 6:37 PM, Wolther said:

sure. But how many people actually stream, video edit or even record their plays. 

Most gamers play on single threaded games (competitive e-sports), and in that case AMD's current cpus are worse at compared to Intel's offerings

In those games any ryzen cpu is just fine.

I mean come on 300+fps on ultra 1080p in Counter strike, what more would you want seriously? Okay a 7700k will get you 400fps, but at that point you're just nitpicking. Out of the box cs go won't run at more than 300 fps anyway, that shows you how inappropriate your example is.

[Citadelen]

11 hours ago, ARikozuM said:

-snip-

That is Mircosofts responsibility, AMD can't make updates for Windows. I can say without hesitation that if AMD could get a patch like that out before launch they would've.

[Master Disaster]

So Jay made a great point in his latest ryzen video...

 

The 1700X is as bad at gaming as the 7700K is at productivity.

 

That's a pretty fair analysis IMO.

[Kobz360]

2 minutes ago, Master Disaster said:

So Jay made a great point in his latest ryzen video...

 

The 1700X is as bad at gaming as the 7700K is at productivity.

 

That's a pretty fair analysis IMO.

He said the 7700K is worse in professional workloads than Ryzen is at gaming. Or in other words, Ryzen does gaming much better than the 7700K can do workstation tasks 

[cj09beira]

8 minutes ago, Master Disaster said:

So Jay made a great point in his latest ryzen video...

 

The 1700X is as bad at gaming as the 7700K is at productivity.

 

That's a pretty fair analysis IMO.

Didnt he say that it wasn't as bad at gaming as the 7700k was at pro tasks?

[nizmoz44]

Not windows 10 fault. It's AMD that needs to have the correct drivers to be compatible with the operating system to work correctly.  

[Guest]

I'm actually worried that AMD royally screwed something up with the scheduler that could be irreparable until the next gen chips. 

 

Hoping I am wrong. 

[TechGod]

12 hours ago, Alexokan said:

I'm actually worried that AMD royally screwed something up with the scheduler that could be irreparable until the next gen chips. 

 

Hoping I am wrong. 

Nope this is NOT the case. the scheduler is OS side, not hardware side. 

[79wjd]

33 minutes ago, TechGod said:

Nope this is NOT the case. the scheduler is OS side, not hardware side. 

However, if it's a pain to fix and not enough people have Ryzen, then...well, no fix. I'm not saying this is the case, just that something being an OS fix doesn't necessarily make it an easy, or even guaranteed, fix. 

[TechGod]

15 minutes ago, djdwosk97 said:

However, if it's a pain to fix and not enough people have Ryzen, then...well, no fix. I'm not saying this is the case, just that something being an OS fix doesn't necessarily make it an easy, or even guaranteed, fix. 

Sure absolutely, thankfully the Ryzen CPUs are a big release. 

[Taf the Ghost]

3 hours ago, djdwosk97 said:

However, if it's a pain to fix and not enough people have Ryzen, then...well, no fix. I'm not saying this is the case, just that something being an OS fix doesn't necessarily make it an easy, or even guaranteed, fix. 

The Linux builds had a scheduler fix in 24 hours.  It's why Ryzen is an utter beast over there.

[Taf the Ghost]

19 hours ago, Lynx said:

There is one interesting (fantasy) theory from Wendell about why AMD give so little time for the software/gaming dev, microsoft, and motherboard manufacture to optimize their product for Ryzen.

Because of the stuff that happen in the past where AMD offer a faster CPU than Intel, and Intel decided to do some stuff that latter judged as unfair, AMD want to keep everything in secret, not give Intel any early hint about Zen architecture and its performance, give Intel little time to respond, so they could disrupt the market.

it's in the beginning of this video.

And also at around 20 minutes, Wendell said that in GTA V, his intel machine (kaby lake i7?) got stutter sometimes, while there's no such thing with Ryzen 7. But well, just watch Wendell in this show, I would just skip when the others are talking, because well, Wendell.

Intel has literally paid billions in fines because of their anti-competitive practices against AMD. So there's a lot of logic to that.

 

Also, Naples, the AMD server-chips based on Zen, are about to murder Intel's server business.  That's probably more what this is about, as we're just a few months from that launch, so the details can be worked out with the R7 launch at the software level.  Welcome to the bleeding edge!

[TechGod]

19 hours ago, Lynx said:

There is one interesting (fantasy) theory from Wendell about why AMD give so little time for the software/gaming dev, microsoft, and motherboard manufacture to optimize their product for Ryzen.

Because of the stuff that happen in the past where AMD offer a faster CPU than Intel, and Intel decided to do some stuff that latter judged as unfair, AMD want to keep everything in secret, not give Intel any early hint about Zen architecture and its performance, give Intel little time to respond, so they could disrupt the market.

it's in the beginning of this video.

And also at around 20 minutes, Wendell said that in GTA V, his intel machine (kaby lake i7?) got stutter sometimes, while there's no such thing with Ryzen 7. But well, just watch Wendell in this show, I would just skip when the others are talking, because well, Wendell.

That's a theory I could get behind. 

[LAwLz]

3 hours ago, Taf the Ghost said:

The Linux builds had a scheduler fix in 24 hours.  It's why Ryzen is an utter beast over there.

How big was the improvement in terms of percentage points?

[Morgan MLGman]

8 minutes ago, LAwLz said:

How big was the improvement in terms of percentage points?

I'd like to know that as well! Though the improvement overall should be lower than on Windows 10 as W10 has CPU power management issues also to be fixed on Ryzen so the overall boost should be theoretically higher on W10 when compared to previous benchmarks.

[MageTank]

4 hours ago, Taf the Ghost said:

Intel has literally paid billions in fines because of their anti-competitive practices against AMD. So there's a lot of logic to that.

 

Also, Naples, the AMD server-chips based on Zen, are about to murder Intel's server business.  That's probably more what this is about, as we're just a few months from that launch, so the details can be worked out with the R7 launch at the software level.  Welcome to the bleeding edge!

I wouldn't be so sure. Knights Landing and Skylake-E (AVX-512) are coming. Intel has plenty of tricks up their sleeves when it comes to the HPC market, and are willing to flex those muscles if AMD gets too close with Naples. I won't say AMD doesn't stand a chance, simply because I do not know yet, but I certainly wouldn't say they are going to "murder Intel's server business" because that's simply not going to happen in a single generation.