cancel
Showing results for 
Search instead for 
Did you mean: 

AMD Ryzen™ Community Update

Staff
Staff
10 40 180K

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 We have determined that an outdated version of the application was responsible for originating the incorrect topology data that has been widely reported in the media.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.

Product NameTrue Junction Temp (Example)tCTL Offset for Fan Policy
Temp Reported by tCTL
AMD Ryzen™ 7 1800X38°C20°C58°C
AMD Ryzen™ 7 1700X38°C20°C58°C
AMD Ryzen™ 7 170038°C0°C38°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:

  1. 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.
  2. 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.

40 Comments
Journeyman III
Journeyman III

Okay... So you're not going to address the actual problem regarding the latency problems caused by pulling information across 2 CCX's? If windows see's ryzen as a 4+4 chip it could fix the issue... If you don't want to fix this issue then there's no reason to buy your chips for gaming purposes.

Forerunner
Forerunner

Excellent news. Thx 4 sharing this info.

You could be tweeted a lot more if high end X370 motherboards were available to buy somewhere, out of stock or 15+ days shipping wait.

I will buy a Ryzen CPU the day I can buy the mythical X370 highend motherboard.Raiders .of.th.lost.x370.mobos.jpg

Raiders of the lost highend X370 mobos.

Journeyman III
Journeyman III

Does this mean that a performance improving AMD driver or Windows update is not planned or in the works for Ryzen?

Adept II
Adept II

i am on same boat

Journeyman III
Journeyman III

The thread scheduling issue wasn't SMT, but thread allocation per CCX block. It's already known that scheduling can be improved. People have turned off 4 cores and gotten compile times reduced by 35%. Clearly, when 4 cores are better than 8, there are scheduling issue, would you not agree?

Adept I
Adept I

I'm so sorry there are abusive comments - glad you can blank them out.

Thanks for Ryzen!!  8 cores / great clocks / soldered IHS & much, much else...  Zen is an enormous achievement that will benefit all of us in one way or another.

And good job on explaining what you've found re the community reports on threading anomalies etc.  It's nice to see AMD formally address these once you've had a chance to get the facts together.  The responsible sites will report it and some amount of controversy will die off.

My next CPU will be an 8-core Ryzen.  It's not only affordable, it's the only one with a modern geometry that's even close to being so.

Adept I
Adept I

Win10 scheduler may be operating properly but that does not mean it is fully utilizing the Ryzen architecture in an efficient manner. Are AMD really admitting that the scheduler throwing threads randomly over the 2 CCX's causing a performance hit is acceptable? Sure you can program around this in games but what about all those older games that the developers have not updated in years? If Windows scheduled threads more intelligently to allow threads to be on a single CCX surely this would improve performance (and power usage)? Anyhow. The fact that Windows 7 performs with a 10-20% improvement in DX11 seems telling and I doubt it can be attributed to difference in the OS architecture. But what do I know?

But I am sure this will get hidden/deleted, since it is a negative comment going against what has been stated by AMD.

Journeyman III
Journeyman III

Oh come on! Are you really saying that there isn't any topology malfunction?

Admitting there are issues with some game's understanding of Zen's topology and it has been shown all over the internet, like the 2 separate CCXs... but you guys say the scheduling is working fine?

How can you blame on games coding? The 2x4 design is the most important optimization due to PS4, Xbox One and PS4 Pro sharing this same 2x4 design.

Same with SMT, the performance losses in some games with SMT enabled is NOT because the scheduler does not see the difference in them, but because it is more likely to switch threads from one CCX to another when SMT is enabled sometimes.

What is going on here?

Or MSFT is telling they aren't going to made any scheduler changes for Ryzen's architecture, and now AMD is going to have to get every single game developer to go out of his way to optimize against Ryzen's architecture.. Or AMD is trying to not throw MSFT under the bus because they are big AMD customer and you are willing to take a little flack to keep the big customer happy?

This is a weak statement, really.

Journeyman III
Journeyman III

This is something that can be addressed as pointed out by deadman3000.  As things are now, you can have a simple 2-thread application in which, randomly, one thread is assigned to a different CCX than the other.  This visibly reduces performance.  You can run the same application again and, if you're lucky, have that same application's threads assigned to the same CCX--resulting in improved performance.  This has been demonstrated.  All an OS would have to do is prefer to put application threads on the same CCX when possible.

I'm a fan of the Ryzen processors, but I do hope someone addresses this issue.  I don't think changing the way threads are assigned will drastically improve performance in most cases, but it will improve performance in some cases and it will improve consistency--and, to me, that's worth the effort to address.  Please don't shut down criticism or alternate opinions--particularly when they're backed by evidence.  This is a new processor, and you guys did a great job.  Keep it up, but stay open.  That's what people love about AMD.

Journeyman III
Journeyman III

Are you guys aware that Hardware.fr found that the 2x CCX "Topology" of Ryzen is the cause of performance drops in some games? Their official review found 20% performance hit for Battlefield 1, 10% perf hit for several other titles. A lot of evidence from software engineers on tech forums all point to the same thing, cross CCX thread migration causes latency or cache thrashing so Ryzen's L2/L3 is not as effective under mis-scheduling.

Sure, Window's Scheduler is "fine", it is just not optimal for Ryzen's unique topology (which is similar to PS4, PS4 Pro, Xbox One) of 2x 4 cores. The scheduler is operating on the basis that Ryzen is a 1x8 cluster of cores. So yes, there are improvements to be made.

But it's just not only from optimizing games better, on the basic level, Window's Scheduler needs to be aware of Ryzen's cache/core topology and moves threads around less often. This is actually linked to your suggest of using the High Performance power management, as there's a lower penalty for CCX jumping under that power plan.

The better way to word these issues is simply: "Ryzen is a new architecture, and game engines and OS optimizations has room for improvement, both of which we will work very hard on moving forward". It's both diplomatic and address the real issue.

For those who want some technical explanation of what we're talking about when we raise CCX and such, view this video here, it explains it all in simple terms: The Zen in Ryzen: A Tale of Cache, Compute Complexes & Scheduling - YouTube

Exemplar
Exemplar

I did a bit of testing on my own, and the first game I tested showed a negative hit from SMT.  Gears of War 4.  I ran the benchmark in the game with SMT disabled in my ASRock Fatal1ty Pro.  Below are the results.

Gears of War 4 3_13_2017 7_44_01 PM.png

When I enable SMT, the CPU stats drop about 9% on average. But, as evidenced by the graph there are far more instances of random low FPS.  I guess it remains to be seen which behavior is in fact, the outlier, but in Gears of War 4, a modern DX12 title, there is definitely a negative impact.

Gears of War 4 3_13_2017 9_20_23 PM.png

But I will qualify this a say that even the SMT performance is way ahead of my old FX-8350.

Gears of War 4 2_25_2017 8_42_14 AM.png

Journeyman III
Journeyman III

Here are a few simple examples that clearly demonstrate the issue:

Assigning threads to specific CCX: https://www.youtube.com/watch?v=BORHnYLLgyY

Windows randomly selecting CCXs: https://www.youtube.com/watch?v=JbryPYcnscA

Adept I
Adept I

Conguroglation to rysen designer and amd

workers,i ask near 20 ryzen buyer ,are you happy withryzen , 100% say YES

People must be clever and dont belive intel seller gossip to harm ryzen

I hope all understand ryzen technology new 99% of games(software) cant relize and see ryzen cores truley

Like CPU-Z erorr in 1st week , after add ryzen works exclent , so games need time to update to see ryzen truely (if you check youtube  games test ryzen games can use  avrage 20% of cpu , 2core 4 threat only!!!!!!!???)

I promise ryzen 7 real and total power is 200% more from i7-7700

However compair 7700 with 4 core(produce for game) not fair with 8core cpu(optimize for workstation)

Thanks to all in amd, i belive after software update and VEGA graphic,we will see 100% moreryzen  power(speed)

Adept I
Adept I

what about all those older games that the developers have not updated in years?

Usually these run fast cause they're ...ermmm...old.

Nobody really cares about that as much I think.

On the other hand, I'm not so sure devs can decide what the scheduler should do, I've talk with some devs and they said windows makes all those decisions, not the application.

Can anyone (with the knowledge) confirm this?

I guess if any, windows should be aware a Ryzen CPU has two CCXs and should try to avoid sharing cache between them to avoid the latency penalty.

Adept I
Adept I

I think the scheduler/threads management optimization etc...whatever you wanna call it has to be tackled by all parts involved.

In general I would suggest to avoid throwing the hot potato to the other parties (we've seen it in other opportunities with all sort of companies) as much as possible cause the only winner if this happens is gonna be Intel.

Journeyman III
Journeyman III

Great, when can we get some fricken motherboards? This complete lack of x370 motherboards is pathetic. I've had my 1800 sitting on my empty case for nearly 2 weeks and not a single retailer can tell me when they can ship a motherboard...

Journeyman III
Journeyman III

Please allow us or comment on being able to disable AMD's PSP, open-sourcing requests are strawmen, we just want it off

Adept II
Adept II

keep these "Community Update's" coming! they are like a sitrep for amd users!, include content like user reported issues, solutions and game updates, when they are ready

Journeyman III
Journeyman III

keep'em coming , and use some raw language , don't be soo elegant

Challenger
Challenger

"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."

Hopefully, they will resolve induced latency via CCX intercommunicating, as it is easily reproducible.

Journeyman III
Journeyman III

should have replied this to main article.

Miniboss
Miniboss

Calm down people.  These issues that are being worked on DO NOT MAKE GAMES UNPLAYABLE...in fact, gaming on Ryzen in it current state is still way smoother and faster than the FX series.  I've been playing games on the AMD Twitch channel and they are performing just fine....and yes, even in the dreaded 1080p that everyone says is horrible.  You know what I find?  Buttery smooth gaming and plenty of power to run the most cpu heavy games out there right now such as H1Z1, The Culling, GTAV, BF1, ARK, RUST etc....   remember guys,  this is a new architecture,  on a new platform and there will be bugs.  These bugs WILL be taken care of, just be a little patient.  But in the meantime, gaming on Ryzen is just fine.

Miniboss
Miniboss

That's not AMD's fault.  They don't make motherboards.  It's the motherboard manufacturers you should contact.

Forerunner
Forerunner

derrickgott007​ is correct. Ryzen is an excellent CPU and wayyyyyyy better than FX-9590. The only big problem is availability of high end X370 motherboards right now. AMD bought all high end X370 motherboards for redteam+, Utube reviewers and master overclockers and left none for the rest of us. If there is a real problem with the Windows Os it will be fixed, just like when they fixed the parking and scheduler for FX CPU on Windows 7 (you had to apply two patches).

Adept II
Adept II

Trying to appeal to the MS Overlords because they have big big biiiig Server deals with them.

Exemplar
Exemplar

I also agree.  While I did see the FPS drop on the CPU tests for Gears of War when I enabled SMT, it's not like to processor performance is crippled.  I am now GPU bound on an R9 Fury X at 1080p whether I have SMT enabled or not.  Both results are miles ahead of my FX-8350 where I was effectively CPU bound half the time on the same benchmark.  The transcode and rendering performance are also top notch!  You will not be disappointed if you purchase one of these processors. 

A bigger issue for me was the still shaky XMP support for the motherboards.  You can check out my post in the red team rig area "monolith".  I gained a substantial amount of performance in 3Dmark by overclocking my DDR4 memory from 2133 MHz to 2667 MHz.  However, I had to set voltages and timings manually.  I couldn't get any of the XMP profiles to boot at all. 

Exemplar
Exemplar

I guess I still don't understand why there is a performance dip with SMT enabled.  If the issue is the added latency of the thread crossing to the other CCX, the probability of that happening is exactly the same if SMT is enabled (2X8) or disabled (2X4).  In either scenario, each CCX has the same number of logical processor's so the probability of jumping across a CCX is identical.  Why then a difference when turning SMT on and off?

Journeyman III
Journeyman III

To be fair it is some of AMD's fault they've passed loads of the high end motherboards off to redteam+ and youtube reviewers/review sites and all of those overclocker groups.

Adept I
Adept I

maybe that. Definitely XBox

Adept I
Adept I
Journeyman III
Journeyman III

From what I read so far it's not a scheduling issue. What needs to be understood is that there are 2 blocks of 4 cores with a local 8MB of L3 cache. The other 8MB of L3 cache is located "across slow fabric" on the other block of 4cores.

So what you should be comparing is:

Turn off SMT, so you exclude primary and secondary execution paths.

and then

run tests with 2 cores, where:

- core1 is in CCX1 and core2 is in CCX2.

and

- core1 and core2 are on the same CCX(1 or 2).

This should potentially show you if and how much, the access to L3 data 'on the other side' is creating how much latency.

It should be noted that running test2 as described here will be done vs 8MB of L3 cache local vs 2x8MB split with 'theoretical' latency issue.

Miniboss
Miniboss

To me the situation looks like the Ryzen will be revised on-die/hardware level - likely some designers clearly did not think far enuf about the implications of the current design as is. Better luck next time?

Adept I
Adept I

Hi AMD

About the temperatures.

That means if I see 80 in CoreTemp in reality is 60?

Journeyman III
Journeyman III

The more threads you have available the more Windows likes to move it around and thus your chance of jumping across to the other CCX is higher at any point in time. That's what's happening with SMT on vs off, it gets more erratic.

Basically Window's Scheduler needs to be told there is a performance hit for moving threads from 1-8 to 9-16 for Ryzen 7 so it limits the movement where possible.

If Microsoft does not cooperate with this, then game developers must do it manually as they can override the scheduler somewhat. But both the OS & games must be aware of Ryzen's core/cache topology.

Journeyman III
Journeyman III

There are some simple situations where scheduler could decide what to do with threads (for example application with 2 threads should prefer one CCX). But there are lots of other situations, where it can`t be universally decided. Think about game which spawns 11 threads, of which 4 are heavy duty and 7 make less important background job (like asynchronous AI simulation, prefetch map and level resources, updating other players positions on map etc.). If all these threads are of the same priority, OS would not know how to dispatch them correctly, since it doesn`t know which thread will need which data (from which CCX cache region) in the future. OS may only guess here - but how can it know that putting all Deus Ex, Doom threads on one CCX is better than spreading them across 2CCXs? In Battlefield 1, Thomb Raider and CS:GO situation is opposite - spreading all threads across 2CCXs results in higher performance.

5 Games On Ryzen 7 1700 8C/16T vs. 4C/4T (4+0) - YouTube

It is up to developers to optimize their software. But it is not a trivial task, unfortunately and many games will probably never get such updates. We can only wish that future software will be better lined with Zen architecture and next gen Zen will be simply better.

In my opinion Ryzen should not have NUMA architecture inside its own cpu/cache hierarchy. But remember that this design didn`t happen without a reason (time, cost, interconnections routing, tdp, chip layout, yield, etc).

Journeyman III
Journeyman III

Is is not a malfunction. It is consequence of Ryzen design. If you run only 1 CCX, some games work better (Deus Ex and Doom) but some get performance penalty (Battlefield1, Toms Raider, CS:GO). Scheduler doesn`t have knowledge required to decide which threads should work on one CCX. You can see this here:

5 Games On Ryzen 7 1700 8C/16T vs. 4C/4T (4+0) - YouTube

It is up to developers to optimize their software, really. Surely, there are some things MS can tweak in their scheduler, but in games with high thread count (Battlefield 1 and Deus Ex) it is up to developers to make good use of CPU resources.

Exemplar
Exemplar

While I agree that it can only get better.  My experience so far with Ryzen is pretty positive.  In productivity it performs as well as Intel CPUs twice as expensive.  While gaming isn't the chips strongest area it is still enough to keep users from being CPU bound in modern games, and light years ahead of AMD's previous offerings.  To make up nearly 6 years  of ground with this one release is truly remarkable.

pastedImage_0.png

Journeyman III
Journeyman III

May its a me issue but 1700x 16gb ddr3000 rx 480 and my benchmarks are lookign good but games feel very groggy. Also noticing CPU is barely getting above 20% utilization so maybe a GPU thing but still.

Adept I
Adept I

Does anybody know whether the temperature shown by AMD's Ryzen Master software is the temperature with or without said 20 degree offset?

Adept I
Adept I

Your temperature reporting is crap, really. My 1700X with a large cooler (be quiet dark rock) reports 55..60°C in Idle, goes up to 75°C with some light load and up to 95°C with maximum load from Prime 95. But at 75°C the BIOS (Version 0503) on my ASUS Prime B350 Plus sets all fans to maximum speed and there is no way to keep the system silent even when only browsing through the internet. This is not acceptable for me, especially because the previous BIOS (0406) was reporting correct temperatures (idle 40°C, max load 75°C) and made the fans work as expected.