AMD's Robert Hallock (previously the Head of Global Technical Marketing for AMD and now working full time on the CPU side of things) has posted a comprehensive Ryzen update, covering AMD's official stance on Windows 10 thread scheduling, the performance implications of SMT, Windows power management settings, and more. The post in its entirety is reproduced below, and also available from AMD by following this link.
(Begin statement:)
It’s been about two weeks since we launched the new AMD Ryzen™ processor, and I’m just thrilled to see all the excitement and chatter surrounding our new chip. Seems like not a day goes by when I’m not being tweeted by someone doing a new build, often for the first time in many years. Reports from media and users have also been good:
- “This CPU gives you something that we needed for a long time, which is a CPU that gives you a well-rounded experience.” –JayzTwoCents
- Competitive performance at 1080p, with Tech Spot saying the “affordable Ryzen 7 1700” is an “awesome option” and a “safer bet long term.”
- ExtremeTech showed strong performance for high-end GPUs like the GeForce GTX 1080 Ti, especially for gamers that understand how much value AMD Ryzen™ brings to the table
- Many users are noting that the 8-core design of AMD Ryzen™ 7 processors enables “noticeably SMOOTHER” performance compared to their old platforms.
While these findings have been great to read, we are just getting started! The AMD Ryzen™ processor and AM4 Platform both have room to grow, and we wanted to take a few minutes to address some of the questions and comments being discussed across the web.
Thread Scheduling
We have investigated reports alleging incorrect thread scheduling on the AMD Ryzen™ processor. Based on our findings, AMD believes that the Windows® 10 thread scheduler is operating properly for “Zen,” and we do not presently believe there is an issue with the scheduler adversely utilizing the logical and physical configurations of the architecture.
As an extension of this investigation, we have also reviewed topology logs generated by the Sysinternals Coreinfo utility. We have determined that an outdated version of the application was responsible for originating the incorrect topology data that has been widely reported in the media. Coreinfo v3.31 (or later) will produce the correct results.
Finally, we have reviewed the limited available evidence concerning performance deltas between Windows® 7 and Windows® 10 on the AMD Ryzen™ CPU. We do not believe there is an issue with scheduling differences between the two versions of Windows. Any differences in performance can be more likely attributed to software architecture differences between these OSes.
Going forward, our analysis highlights that there are many applications that already make good use of the cores and threads in Ryzen, and there are other applications that can better utilize the topology and capabilities of our new CPU with some targeted optimizations. These opportunities are already being actively worked via the AMD Ryzen™ dev kit program that has sampled 300+ systems worldwide.
Above all, we would like to thank the community for their efforts to understand the Ryzen processor and reporting their findings. The software/hardware relationship is a complex one, with additional layers of nuance when preexisting software is exposed to an all-new architecture. We are already finding many small changes that can improve the Ryzen performance in certain applications, and we are optimistic that these will result in beneficial optimizations for current and future applications.
Temperature Reporting
The primary temperature reporting sensor of the AMD Ryzen™ processor is a sensor called “T Control,” or tCTL for short. The tCTL sensor is derived from the junction (Tj) temperature—the interface point between the die and heatspreader—but it may be offset on certain CPU models so that all models on the AM4 Platform have the same maximum tCTL value. This approach ensures that all AMD Ryzen™ processors have a consistent fan policy.
Specifically, the AMD Ryzen™ 7 1700X and 1800X carry a +20°C offset between the tCTL° (reported) temperature and the actual Tj° temperature. In the short term, users of the AMD Ryzen™ 1700X and 1800X can simply subtract 20°C to determine the true junction temperature of their processor. No arithmetic is required for the Ryzen 7 1700. Long term, we expect temperature monitoring software to better understand our tCTL offsets to report the junction temperature automatically.
The table below serves as an example of how the tCTL sensor can be interpreted in a hypothetical scenario where a Ryzen processor is operating at 38°C.
Power Plans
Users may have heard that AMD recommends the High Performance power plan within Windows® 10 for the best performance on Ryzen, and indeed we do. We recommend this plan for two key reasons:
- Core Parking OFF: Idle CPU cores are instantaneously available for thread scheduling. In contrast, the Balanced plan aggressively places idle CPU cores into low power states. This can cause additional latency when un-parking cores to accommodate varying loads.
- Fast frequency change: The AMD Ryzen™ processor can alter its voltage and frequency states in the 1ms intervals natively supported by the “Zen” architecture. In contrast, the Balanced plan may take longer for voltage and frequency (V/f) changes due to software participation in power state changes.
In the near term, we recommend that games and other high-performance applications are complemented by the High Performance plan. By the first week of April, AMD intends to provide an update for AMD Ryzen™ processors that optimizes the power policy parameters of the Balanced plan to favor performance more consistent with the typical usage models of a desktop PC.
Simultaneous Multi-threading (SMT)
Finally, we have investigated reports of instances where SMT is producing reduced performance in a handful of games. Based on our characterization of game workloads, it is our expectation that gaming applications should generally see a neutral/positive benefit from SMT. We see this neutral/positive behavior in a wide range of titles, including: Arma® 3, Battlefield™ 1, Mafia™ III, Watch Dogs™ 2, Sid Meier’s Civilization® VI, For Honor™, Hitman™, Mirror’s Edge™ Catalyst and The Division™. Independent 3rd-party analyses have corroborated these findings.
For the remaining outliers, AMD again sees multiple opportunities within the codebases of specific applications to improve how this software addresses the “Zen” architecture. We have already identified some simple changes that can improve a game’s understanding of the "Zen" core/cache topology, and we intend to provide a status update to the community when they are ready.
Wrap-up
Overall, we are thrilled with the outpouring of support we’ve seen from AMD fans new and old. We love seeing your new builds, your benchmarks, your excitement, and your deep dives into the nuts and bolts of Ryzen. You are helping us make Ryzen™ even better by the day. You should expect to hear from us regularly through this blog to answer new questions and give you updates on new improvements in the Ryzen ecosystem.
(End statement.)
Such topics as Windows 7 vs. Windows 10 performance, SMT impact, and thread scheduling will no doubt still be debated, and AMD has correctly pointed out that optimization for this brand new architecture will only improve Ryzen performance going forward. Our own findings as to Ryzen and the Windows 10 thread scheduler appear to be validated as AMD officially dismisses performance impact in that area, though there is still room for improvement in other areas from our initial gaming performance findings. As mentioned in the post, AMD will have an update for Windows power plan optimization by the first week of April, and the company has "already identified some simple changes that can improve a game’s understanding of the 'Zen' core/cache topology, and we intend to provide a status update to the community when they are ready", as well.
It is refreshing to see a company publicly acknowledging the topics that have resulted in so much discussion in the past couple of weeks, and their transparency is commendable, with every issue (that this author is aware of) being touched on in the post.
I don’t know how they can say
I don’t know how they can say that the scheduler is working properly when we can see that it’s not currently, sufficiently accounting for the massive diff in l3 latencies which requires core grading.
I think this is simply
I think this is simply because AMD fully complied with Windows 10 and expected from Microsoft a good sportsmanship. AMD never expected such an outright BACKSTABBING on Microsoft’s side. Basically, AMD’s in the “la-la-la-la, I can’t hear you!” mode right now in regards to problems with Windows 10, because AMD fully expected great performance oN Windows 10 and never foresaw such problems.
lol.
lol.
Yeah, Microsoft is basically
Yeah, Microsoft is basically Turkey.
Why do i have a slight
Why do i have a slight feeling that you’re just a butthurt Amd Fanboy and nothing else?
Relax dude, there is no need to spend all your life defending a company which always manage to screw things up.
Why post anonymously? We all
Why post anonymously? We all know who these childish comments come from.
Check your butthurt feelings
Check your butthurt feelings twice over, then.
I won’t be repeating myself anymore on this: my current main system is on i7 2600K.
There are problems they wont
There are problems they wont acknowledge its starting to look very shady imo…
The problem is not only in
The problem is not only in that AMD itself has gone full-blown delusional denial due to the utter shock of being backstabbed by M$, but also because such “””””tech xxxperts””””” as PcPer are trying to swipe everything under the rug and are acting like problem doesn’t exist at all. PcPer is basically CIA during the period of Snowden’s leaks. They’re in full damage control mode.
…I’m beginning to wonder
…I’m beginning to wonder whether we’re being trolled by 4chan or something, these comments are next level stupid.
…or Intel shills :p
…or Intel shills :p
Grow up.
Grow up.
Could you just stop posting
Could you just stop posting to this site!
ESAD (c) You-know-who
ESAD (c) You-know-who
Go back to getting your tech
Go back to getting your tech news from 4chan
Damage control for who
Damage control for who exactly? Nobody is making anyone out to be a clear winner here, but you sure seem to be locked into believing that everything everyone writes is a conspiracy theory. Still wearing the tin foil hat I see…
Before Ryzen PCPer was
Before Ryzen PCPer was supposedly Intel Nvidia shills. Now youre AMD shills? Can these people stfu already? Its like listening to Bernie Sanders and Clinton voters bitching constantly.
Your evident political
Your evident political affiliations are, unsurprisingly, as basic as your comments.
Well i also post a lot about
Well i also post a lot about very specific details of semiconductor design and engineering on here. Funny how you assume that my political voting record affects my trchnical knowledge, but if youd like to discuss the intricacies of semiconductor engineering, supercomputer architecture, just about any aerospace or mechanical engineering topic, im all for it.
But my point remains. For years PCPer has been called Intel and Nvidia shills for stating facts, from FCAT exposing the fact that Crossfire didnt work, to covering AMDs lackluster CPUs.
Now theyre being called AMD shills? Ironic how the technically illiterate pop out of the woodwork when their brand isnt doing better than another brand.
I have an i7-6700k and a 1070, but AMD has released what appears to be a good CPU architecture.
Im not crazy about two dies being used with a chip to chip interconnect rather than a monolithic die with a lower latency core to core bus, but they managed to make Ryzen compete with chips that cost 2-3x as much from Intel, on a small socket.
And unlike intel, theyre using solder on their small chip. AMD has gone all out with Ryzen and i am glad that intel will have to stop their “bump the clock and call it a new chip” kaby lake horse shit or 10 core $1700 CPUs that their own quad core architecture CPUs beat in a lot of cases.
AMD isnt the only one eating into Intel’s dominance either. Nvidias GP100 is destroying Knights Landing in some cased, which in many ways is a let down of an architecture after years of hype.
Fujitsu is working on the most advanced supercomputer architecture in the world, while their now 6 year old K Computers SPARC architecture is still #1 on HPCG and Graph500, beating machines based on Intel CPUs with nearly 10x the peak FLOPS. They also deployed a 34 core CPU with HMC a year before Intel, with 10x the performance of K. No one, including Intel, really has anything to compete with the likes of that. Especially since its successor will likely beat anything Intel has for exascale.
Skylake Purley, which is on the same socket as Knights Landing, and an absolutely massive chip, cant be a letdown, otherwise Intel is going to lose serious market share. If you can buy Zen based 32 core x86 CPUs that perform close to Skylake 28 cores, for significantly less money, AMD can start to really come back.
“Im not crazy about two dies
It’s a single chip/die solution.
https://cdn.arstechnica.net/wp-content/uploads/2017/02/ryzen-die.jpg
Thought one of the big issues
Thought one of the big issues was getting the core groups, which talk over a fabric that is different than the core to core topology within each core complex was a major source of latency.
Isnt each group of cores a NUMA node that has an inter and intra core communication fabric on chips with more than 4 physical cores?
It’s the bumpkin problem
It’s the bumpkin problem Gaming has a SERIOUS bumpkin problem, so don’t expect things to ever be civil!
Gaming with its FPS, and other CS:GO types of games will always attract the angry Bunpkins and their urban equivalents, the JOE Six-Packs!
Gaming has a serious Bumpkin problem!
Replace the Bumpkins/Joe Six-Packs with ROBOTS and Boycott any factories’ products in any countries that employ Bumpkins/Joe Six-Packs, ditto for any agricultural production that employs the Bumpkins/Joe Six-Packs!
ROBOTs are that answer to the Bumpkin/Joe Six-pack related problems of the world!
Let the Bumpkins/Joe Six-Packs go back to the 12th century in their isolated communities while the rest of the world goes about living!
What the fuck are you
What the fuck are you attempting to incoherently say?
Hey! Hey! Hay! Joe-Sixpack
Hey! Hey! Hay! Joe-Sixpack now don’t you have no heart attack!
They’ve moved your jobs on overseas and given them to the Red Chinese!
They’ll get Mad Tax Breaks to move the factories back here!
But they’ll be using ROBOTS so that’s just one more bitter pill!
They don’t need you they’ve used your vote!
To get those Mad Tax Breaks!
Now don’t it just get you goat!
The cost of ROBOT labor is now cheaper than the shipping!
So they’ve got their mad tax breaks and you are up the river!
You still don’t have a job, and you still will be a slob!
So maybe they’ll feed you some more cornless corn cobs!
‘caues they’ve punked you all over again for the umpteenth time!
And you are still left without one thin dime!
It seems to me that things
It seems to me that things are actually going quite well for a completely new architecture. There is going to be differences that will need to be worked out, either by software developers or compiler writers. There could still be some OS tweaks also. Scheduling processes isn’t a simple problem to solve. Multi-threaded programming isn’t simple either, especially if threads need to share varying amounts of data. I am probably going to build a Ryzen system, but I want to wait for slightly more mature motherboards. I have had issues with “board revision 1.0” before. I had an early AGP board a long time ago that just had issues. It died in a bios flashing accident.
> “already identified some
> “already identified some simple changes that can improve a game’s understanding of the ‘Zen’ core/cache topology, and we intend to provide a status update to the community when they are ready”
This highlights they know that Zen’s 2x CCX design is causing a performance problem in some games more than others. Topology of Ryzen is basically 2 groups of 4 cores, it is not a single group of 8 cores, and Windows does not recognize this unique topology as it freely moves threads around whenever it feels like it.
Technically, the scheduler is fine, it’s doing what it’s told. However, it’s doing a bad job at handling Ryzen’s unique topology. So it can be improved.
Pretty simple changes actually.
This is not a simple thing to
This is not a simple thing to fix from the scheduler. The scheduler does not know which thread need to comunicate with each other with high granularity. If it has 8 threads, how would it know whether to schedule them across the 8 physical cores or schedule all 8 on a single CCX? One way may be better than the other, or it may not. It is really up to the developer, since they know what each thread is doing and how much shared data there is.
I love Ryzen, but it’s clear
I love Ryzen, but it’s clear that they took “the budget approach” to get octo-core to the masses by joining two quad-core modules, rather than building “native” octo-core CPUs. If the CCX issues can be fixed with software, that’s fantastic, but then it passes the buck to the individual software makers.
Can AMD work with Microsoft to get this done more at a lower level?
Well Intel does the same
Well Intel does the same thing on the server cpus with many cores. It’s done for a reason and it’s here to stay so operating systems/programs just has to learn to live with it just as we had to learn to live with cache when that was added.
Correct. Adding complexity to
Correct. Adding complexity to the scheduler will add latency to thread allocation, which will slow *everything* down. Scheduler code needs to be as tight as possible and adding enough features to detect which threads were sharing which memory and group them by CCX would slow it down considerably. The solution is that the optimization needs to happen at the application (game) level.
If it needs to be done at the
If it needs to be done at the game level, how many devs are realistically going to do that, especially for console ports? I doubt the likes of EA, Ubi, Acti etc are going to be making that a top priority.
Therein lies the rub.
Therein lies the rub.
Well, their workload in DX12
Well, their workload in DX12 is quite likely to be higher than previous versions because they do have control that has been unprecedented. But that price is high as well. In theory they could exploit DX12 to improve perf on Ryzen pretty specifically, but how many devs would consider that a good use of time and manpower?
This is why something at the
This is why something at the OS level should identify applications that use four or fewer threads and keep them on core complex. This doesn’t sound that difficult, but I don’t have a programming background.
The problem is that even if
The problem is that even if such a change was easy to implement, media encoding apps would suffer because they need raw compute power more than they need to share across threads. That's why it is the apps themselves that need to be CCX aware.
Well, don’t be surprised if
Well, don’t be surprised if you see a 2CCX CPU in the next generation consoles. Also PS4 Pro uses 2×4 Jaguar cores already. So it’s not a matter of if they want to do it. Its more of a matter, If they can get away with it by being GPU limited before the Port.
At the same time I believe that AMD will have a say in it, pushing for optimization for multiple ccx complexes
Also you guys, WTF is
Also you guys, WTF is this?
https://archive.is/AzDnA
This is a baby throwing a
This is a baby throwing a tantrum and reveling this is a Nvidia/Intel shill site. I cant believe he posted something as immature as that.
Turns out getting sprayed
Turns out getting sprayed with AMD fanboy vitriol takes a toll on one’s psyche.
Jeremy’s version was
Jeremy’s version was hilarious..
^
^
It was a response to a
It was a response to a comment by an idiot. Probably should have stayed in the comments rather than being made a post on the site, and I have no problem with it having been removed.
The only idiot here is the
The only idiot here is the one you’re seeing each and every morning in the reflection of your mirror.
Just because it’s so for you
Just because it’s so for you and your mirror does not mean it apply to everybody else.
I don’t have any mirrors in
I don’t have any mirrors in my house, but nice try anyway.
You’re that horrified by your
You’re that horrified by your own reflection. How sad.
Nice try, but no cigar.
Nice try, but no cigar. Unlike your fat dirty ass, I workout and wash myself every day. And I have a well-paid job. And I live in my own house.
Not only that but even Master
Not only that but even Master Chen’s reflection is so horrified that it looks horrible exactly like the horrible Master Chen that even the reflection itself refuses to even show up on any reflective surface; thereby reducing the resultant Horror by 1/2 or more(Depending on the amount of horrible Master Chen Reflections that themselves are reflected on any reflective surface) for any reflection related horror related horrors of reflections of horrible Master Chen!!
There is of course the matter of those recursive reflections that go on until infinity! So in Master Chen’s case we use the Horror reducing reflection reducing formula of (Infinity_Of-Horrible-Master-Chen-Recursive-Reflections – Infinity_Of-Horrible-Master-Chen-Recursive-Reflections + (1 * (original horror inducing Master-Chen))).
Which still leaves the very non reflective in any reflective surface(because of a very horrified Master Chen reflection’s refusing to reflect the Horror that is Master Chen’s refection), original Master Chen, that is still visible to the necked eye! And one fine and very durable very opaque paper bag can be employed to solve that horrible issue!
Edit: refection
to:
Edit: refection
to: reflection
You are definitely a virgin.
You are definitely a virgin.
Youre deinitely a FUCKWIT
Youre deinitely a FUCKWIT
You need to reflect on your
You need to reflect on your horrible reflection! Now if only your reflection could be coaxed out of hiding! The poor horrible horrified thing that it is!
Shaved, huh? Good. Let the
Shaved, huh? Good. Let the whole world see their failings.
No, the men in the white
No, the men in the white uniforms are coming to get you! As you are a tad bit obsessive in the pathological sense in your train of irrational reasonings. you are a classic case of the obsessive with your compulsive driven and repeated attempts to apportion blame in an unfair and unreasonable manner on one website among many in an industry that has its industry wide and very systemic problems.
You are unable to properly petition the proper elected officials and/or other agencies tasked with the job of properly addressing the grievances of any individual or groups of individuals. You appear to be more interested in the perverse joy of smiting your irrationally perceived wrongdoers than you are interested in truthfully addressing any problems that are endemic in the entire technology and technology reporting industry.
So Really! You need to take this matter up with you elected officials or any of the agencies tasked with addressing consumer/trade issues.
But Knowing you and your MO, like everyone knows across these fourms on this and other websites, you can not do it on your own and thus the men in the white uniforms are coming to get you to that nice place where they have the trained professionals who can help you to get over yourself and get on with your life.
Learn to not to wall.
Learn to not to wall.
Learn to how to live
Learn to how to live surrounded by padded walls!
Terry, log in already.
Terry, log in already.
You need to reflect upon your
You need to reflect upon your horrible reflection!
The post they just deleted 🙂
The post they just deleted 🙂 The latency Issues affect any gaming platform that uses more than 4 cores so saying this processor will scale well with the future isn’t realistic
Games can scale well past 4
Games can scale well past 4 physical cores on Ryzen just fine, but they need to be aware of the architecture to do so. The scheduler is not meant to be smart enough to figure out exactly which game threads need to share what data, but if the game engine itself segments the threads properly, it will be fine.
What a joy to read your post.
What a joy to read your post. It is obvious that you do not have an ax to grind. It is nice that you give AMD credit for doing the right thing and trusting them about the improvements that will come as developers get how to best code for ryzen. Wish others on the staff were able to be objective and appreciative of all the fucking amazing stuff we are so lucky to be able to buy and enjoy.
The thing I dont get, is why
The thing I dont get, is why is windows7 better at handling Dx11 titles compared to Windows10 ?
Cant Microsoft integrate some the old windows7 optimization into windows10 ? if indeed this is just code difference VS scheduler .
it make no sense to see windows7 perform better at gaming,
or is it expected ? is this also true with the i7-7700k ?
I don’t think there’s been
I don’t think there’s been any credible benchmarks that demonstrate Ryzen has larger performance degradation on Win 10 vs Win 7 when compared to intel’s CPUs. It’s been mostly random forum posts and reddit posts that get circulated.
Yes they can just strip out
Yes they can just strip out all that windows 10 spyare, adware, bloatware, UWP(TIFKAM renamed) abstraction layers and orher uncessary to gaming crud that is bogging the system down with uncessary CPU processor cycles stealing thread/cache bandwidth inducing activity that clogges the Infinity Fabric nonsense!
Windows 10 is so full of crud and it’s clogging everything a real OS should not have to worry about. Windows 10 is a big land grab by M$ of your system for M$’s monetary milking! Enough with the crud M$ get rid of all that crap!
All you CPU cycles and cache coherency bandwidth are belong to windows 10 and M$! So game on that info fools!
Thats why i used windows 8.1
Thats why i used windows 8.1 with windows update disabled on my new machine. Fuck microsoft and windows 10.
Maybe the upcoming windows 10
Maybe the upcoming windows 10 “game mode” will help. Hey, one can always dream, right?
Simple explanation is that
Simple explanation is that Windows 7 is not trying to spread out the load evenly among cores.
Also, I’m wondering if those Windows 7 tests have been done with or without the Bulldozer hotfixes applied.
No windows 7 spreads the load
No windows 7 spreads the load evenly and windows 7 has less of the adware/bloat and spyware/bloat and UWP(TIFKAM renamed) translation layers/Bloat!
So it’s windows 10 that is spreading out the loads of bloat/crap to fill up all available cores/threads on Ryzen with loads of Bloat-work performing software threads running on CPU core processor threads that are Bloat-work running threads that are shifted all over the place across Ryzens’ hardware cores/threads causing the cache subsystems to thrash about with all that windows 10 bloat clogging the L3/l2/l1(D$ and I$) and causing all the Infinity Fabric bandwidth bloat/crap brownouts.
So windows 7 is not affected as much as windows 10 with all of 10’s bloat/crap cycles stealing CPU core cache/subsystems taxing musical chairs processor core/thread dispatch style bloatware processor cores/threads taxing by that craploads of pure insanity that is the bloat/adware/crapware/UWP laden OS known as windows 10!
Windows 10 main selling point
Windows 10 main selling point is dx12. Of course DX 11 isn’t going to be as good in it. It’s so it appears that the performance uplift from using dx12 vs dx11 is greater. Maybe leading to more people ditching older win versions for 10. Obviously win 10 is the only OS that is able to make that comparison.
They have been adding the same spyware to older versions of Windows as well via win update.
The disparity with Windows 7
The disparity with Windows 7 is much more mundane: Windows 7 does not implement the various low-power nodes Ryzan (and other modern CPUs) support, so is always in FULL SPEED FULL POWER FINAL DESTINATION mode. Switch Windows 10 to the same power mode, and the disparity disappears.
Different driver models
Different driver models (WDDM), different drivers and also different versions of DirectX 11.
https://msdn.microsoft.com/en-us/library/jj863687.aspx
Since the topology of the 8
Since the topology of the 8 core Rizen is basically 2 groups of 4 cores. And the problem is basically caused by Windows not recognizing this property. We can expeculate that this problem is not going to pass when AMD releases the 4 core processors (if they are all in 1 group and not in 2 groups of 2)…
And I’m glad AMD is paying
And I’m glad AMD is paying close attention of what People are saying
There does still seem to be
There does still seem to be an issue, not with SMT, but with some types of workloads straddling both CCXes. This link shows significant performance gains in three games when one CCX is disabled: http://www.hardware.fr/articles/956-24/retour-sous-systeme-memoire-suite.html
That said, three other games showed no difference, and four non-gaming tests showed no difference either.
So it might be easier to fix this in the games rather than in the Windows scheduler.
You can literally just set
You can literally just set core affinities to a running game in Task Manager in the mean time.
why you removed the original
why you removed the original title? I love seeing AMD’s fanbois squirming
you forgot to mop your chin
you forgot to mop your chin off, you’re dribbling.
What about this result that
What about this result that shows there is a problem with the scheduler currently not understanding the CCXs? http://www.hardware.fr/articles/956-24/retour-sous-systeme-memoire-suite.html
Yeah. People say there is
Yeah. People say there is nothing wrong with the scheduler, and while it may not be doing anything ‘wrong’ with the info it has, the lack of understanding about the CCXs does mean the scheduler is at fault and needs to be fixed to understand the CCXs.
I don’t think the scheduler
I don’t think the scheduler has a good way of knowing what threads need to be grouped without the developers giving it a hint. Scheduling threads on the same CCX that is already running 4 threads may be very good for some applications and very bad for others.
True.
Some applications with
True.
Some applications with little inter-communication might prefer to have more L3 (such as 7zip), while others that have high inter-communication (such as battle field 1) prefer to be on the same CCX.
How could the scheduler know where to put what thread? A) a new feature for the developers to give a hint to the scheduler or B) the scheduler could be improved to analyse the behavior of applications and make an intelligent decision.
If naples is going to be using 4 x (2 x CCX) for 32 cores, then it seems like the server industry might start pushing for an intelligent scheduler (some kind of AI or something).
How do the results you linked
How do the results you linked support the claim you’re making?
Hardware.fr shows that there is software that suffers from high inter-CCX latency, other software that basically doesn’t care, and then there’s 7-Zip, which apparently enjoys the extra per-thread cache afforded by running it across multiple CCXes.
The latest kernel change for
The latest kernel change for Ryzen scheduling on Linux didn’t actually improve numbers by much. I wouldn’t expect Windows to do any better if they did work with Ryzen topology to eliminate having threads accessing required resources in the other CCX.
Nothing strange here. What
Nothing strange here. What everyone should be expecting to read.
As with graphics cards in the past, where everything was getting optimized for Nvidia and AMD cards and drivers where looking as bad performers with compatibility issues, the same is true for the Ryzen CPUs.
There are games where the optimization for the Intel architecture don’t let Ryzen perform as expected. So what? Ryzen is more than enough for today’s applications and games. Only fanboys and crybabies who measure the meaning of their own existence with benchmark results should have a problem here.
In the future when Ryzen’s maximum performance will be more needed, games will be optimized for it’s architecture. Especially if Scorpio comes with Ryzen cores in it. In that case, all games ported from Scorpio, will be already optimized for the Ryzen architecture.
Yawn.
Yawn.
If you want to yawn, yawn
If you want to yawn, yawn properly with your body and don’t type it out here. We don’t want to read about your bodily functions.
What’re you gonna do next? Sneeze?
I’m gunna vigorously scratch
I’m gunna vigorously scratch mah family jewels and outgas some sulfurious compounds!
Just a different operating
Just a different operating system, accept it. NONSENSE!
Sir, I respectfully refuse.
i7 doesn’t show a similar performance difference between
Win7 and Win10. This is obviously fixable, but requires
responsible parties to take ownership and not cry “Fake
News.”
Well the OS(windows) has been
Well the OS(windows) has been optimized for that i7/other i series SKUs and their market share for damned near a decade and windows 10’s excessive services/bloat/spyware/telemetry sure can require some many thousands of software threads to be multitasked across as many hardware based processor cores/threads as are available.
The i7/other series have got a ring-bus(since Sandybridge) that is better for some things but worse at others compared to AMD’s more point to point inside the CCX unit and more point to point from CCX unit to CCX unit interconnect fabric(?). So yes there will be a Need in the OSs and Applications(For applications that manage their own Core/thread affinity) for some processor/platform connection fabric topology awareness.
AMD and M$ need to get that optimization work done and that same work needs to be done for an Linux Kernel based OS builds, ditto for any BSD based Kernel based OS builds. The server/HPC market will damn sure get its Linux Kernel based scheduling ducks in order for any OSs/applications and API ecosystems, as will for that server/HPC market that will do the same for any who use any BSD based Kernel OS builds.
OS optimization is an eternally ongoing process, If the OS is being maintained correctly! And if the systems software/OS software engineers are allowed to focus their resources on the OS and are not forced to focus instead on any Silly UI and Silly UWP Run-time nonsense and other coding that is more engineered for extracting UWP software, cloud services fees, and ad/personal metrics gathering related monetary/revenue streams!
In every benchmark Win10 won
In every benchmark Win10 won in performance over Win7. That could be CPU or GPU benchmarks, anything that I have found on the net. Yet with Ryzen win7 has better performance over win10 on some applications, and its not a windows problem? It baffles me
So basically they’re saying
So basically they’re saying problems, what problems?
While it’s to be commended that they’ve made a public statement on the issues being talked about in the community it doesn’t exactly instil confidence that they’ve all but denied any issues with the way Ryzen is configured (shared memclk and internal Data Fabric clocks)
https://thetechaltar.com/amd-ryzen-clock-domains-detailed/
And they’ve also denied there are any issue with the way Windows 10 deals with workloads that span CCX’s.
So basically they’re saying
So basically they’re saying problems, what problems?
Hey! Johnny Tightlips! You hear somethin’ ’bout some problems?
Johnny Tightlips: I ain’t sayin’ nothin’!
Glad to see AMD talking about
Glad to see AMD talking about this, bulldozer was left silent for so long before they said anything about community thoughts.
Really want to set the windows 7 vs windows 10 argument because from every windows 7 vs windows 10 benchmark Intel wise the windows 10 bench us usually faster.
Although I do agree with some of the other comments, windows 10 is bogged down by so much unnecessary bloat that it just feels sluggish and hanky at times. Really wish there was a stripped down version out there that had just the essentials.
thats because they knew it
thats because they knew it was a dog.
they’re response now shows the think Zen is a winner
It’s a winner for AMD and at
It’s a winner for AMD and at that 12% more IPC metric than the initial 40% IPC gain that they promised over their excavator micro-arch! But it was never expected to be an outright IPC winner over Intel’s latest In that IPC metric!
Ryzen kicks Intel’s A$$ in that all around Price/Performance metric! And I can get an 8 core/16 thread Ryzen 7 1700 for $10 dollars less than the reduced proce of the 4 core/8 thread 7700K, and I can overclock that 1700 to perform like the 1800X!
So that’s 8 AMD Ryzen 1700 x86 cores for less than the price of 4 i7 7700K x86 cores, and the 1700 is the best all around deal for overclocking at a great saivings! So that’s also a great win for AMD’s Ryzen 1700 customers!
I’ll take that 1700 “dog” any day for any all around workload usage and that 1700 “dog’ still games OK for me!
So, you deleted that POS of
So, you deleted that POS of Jeremiah’s, but you’re still keeping Allyn’s BS intact? Something tells me PcPer needs to reevaluate everything.
You have to wonder what is
You have to wonder what is happening to a site that attacks and ridicules its own readers in posts? Not this one, but the deleted one and Allyn’s for sure. Really out of control children it seems.
I think it is merely a
I think it is merely a symptom of reading too many article comments.
it the post is inflammatory
it the post is inflammatory and incendiary, it will result in angry comments aimed at pcper and staff
the comments on Hilbert’s site might be contentious, but never directed at Hilbert or Guruf3d
that is because he is critical but respectful and appreciative and does not act entitled or superior or smug
there is never a hint of bias in Hilbert’s reviews or posts
same can be said of Tim’s and Sebastian’s posts
Allyn is the main culprit of this troubling trend, but I have no doubt Ryan is right there with him
If you say so. I would like
If you say so. I would like to think I have been pretty evenhanded with my coverage of the different manufacturers. Even when I give kudos, but then talk about a weakness I see in a product, MC and you are quickly all over it implying bias, money, or shilling. Eventually you sorta stop caring about what the other person says to you and you want to start shoveling it back. I see that doesn't help the situation, so I guess it is just ignore from here on out. Enjoy your time at PCPer!
Just ignore them Josh,
Just ignore them Josh, they’re just a set of arseholes.
^
No nom noms for the bridge
^
No nom noms for the bridge dwellers
This guy gets it.
Walrus
This guy gets it.
Walrus below this comment of mine does not.
Master Chen is a paragon of
Master Chen is a paragon of virtue and technical knowledge. He is generous, forthright, and gracious. He treats others with dignity and respect.
“The sarcasm is strong with
“The sarcasm is strong with this one”
– Master Troll
You actually did it. Wow.
You actually did it. Wow.
sarcastic or not, that
sarcastic or not, that response is the way to go
thanks josh for being a stand up guy
If we stated AMD stands for
If we stated AMD stands for Advanced Micro Devices it would set off angry comments.
you really don’t get it
why
you really don’t get it
why do you think hilbert never gets any shit
he presents things in a balanced, fair, and unbiased way
he is appreciative and respectful rather than superior and entitled
he does not go out of his way to prove he is smarter than his readers and doesn’t believe he is
allyn can’t stop chiming in, though couched in a attempt to act unbiased. to prove he was right about the scheduler
big fuckin deal
I agree.
But saying they’re
I agree.
But saying they’re going outta their “way to prove he is smarter….”, is a bit inaccurate.
I think any “chiming in” is a direct result of their integrity being called into question. And that is a COMPLETELY different scenario altogether.
However, I still agree. Better to turn the other cheek. Easier said than done although. Lots of empathy for the Per’s staff… It’s very difficult to take out the garbage without getting dirty yourself.
thanks for your thoughtful
thanks for your thoughtful reply
i still have to stand by what i said based on this somewhat rhetorical question
why is it that his integrity is so often questioned while Hilbert’s never is?
PCPer articles tend to get
PCPer articles tend to get into the technical details that most consumer level sites dont, largely because their readers are average consumers who dont care enough to get that in depth.
They also dont pretend to be robots that dont have opinions. I think its fine that they have opinions, even if i dont agree with them all.
Its not like the professional sites like semiengineering, hpcwire, the next platform or even top500 arent FILLED with personal opinions of the people who write there.
I rather enjoy calling them out when theyre wrong about things on those sites as well.
But being technical somewhere in between a consumer and professional site is one of the main reasons that i read PCPer and not other tech sites. Its actually interesting to read the technical details, not just the typical consumer oriented bullshit that most tech sites spew.
Thanks for the feedback. It’s
Thanks for the feedback. It's a balancing act between a giving readers a list of specifications and what my thoughts/opinion of something is. I try to stick to the facts and wait towards the end share any opinions I have of them while pointing out that they are just that; opinions. I don't mind if you do not agree with them, heh, it is actually a good thing that you are able to take the information from here and elsewhere and make up your own mind on things. Everyone usually is able to have productive discussions in the comments either way which is good even if its telling me how wrong i am :-). constructive critism is fine/a good thing.
I wasnt implying that i
I wasnt implying that i disagree often. Your articles tend to be spot on. Its more directed at the trolls in the comments section saying “well this other tech site that likes EVERYTHING doesnt have its integrity questioned!”
The reason for that is, they pretend abstain from having an opinion.
You do find THOSE website comments sections with tons of ignorant “oh i wish i hadnt bought this because i didnt know better” type comments or “damn why did i just buy a shitty Maxwell GPU?” AS IF PASCALL WASNT ANNOUNCED YEARS IN ADVANCE!
Thats the difference between a news regurgitation “tech site” and a tech site that gets into the nitty gritty of microarchitecture, storage controller architectue or even the type of semiconductor being used, which PCPer does do.
This is literally the only consumer PC oriented site worth a damn, because it focuses on details like a professionally oriented one(like HPCWire, The Next Platform or Semiengineering etc.) I read those sites as well but they cover totally different subjects than PCPer or other “normal” tech sites.
Thank you and the rest of the PCPer staff for being different and standing apart from the drooling masses.
AMD means “American Marvelous
AMD means “American Marvelous Designs”.
AMD—> Allyn Malventano
AMD—> Allyn Malventano Destroys
lol
lol