AMD officially launches Ryzen processors

[Boo Berry]

1 hour ago, Andre S. said:

AMD isn't releasing chipset drivers for Windows 7, so why should Microsoft support it? And then they don't prevent the OS from running, it just won't receive updates - they would need to test these updates on these new processors and just don't want to put the additional investment into a legacy OS.

 

It's impressive Microsoft is still releasing updates for a 7 year old OS, and that they've kept adding support for new processors until this year.

Totally agree. If users can't accept that (or upgrading to Windows 10) they can always give Linux a shot.

[spaceelf]

Figures the guy gets to the end of our driveway and leaves without dropping it off.  bloody hope he comes back today o.o

[Boo Berry]

20 minutes ago, Andre S. said:

Btw,

https://www.reddit.com/r/Amd/comments/5ybrxn/ryzen_7_is_actually_behaving_like_a_dual_4c8t/

But then AMD says this:

https://community.amd.com/community/gaming/blog/2017/03/13/amd-ryzen-community-update

 

Is it just me or something doesn't compute (no pun intended) here? How is Windows 10 scheduler behaving fine if it randomly assigns threads between "CPU Complexes"? How are developers supposed to optimize for that? Some technical details are missing here.

Gotta love the comments here!

[Andre S. Veteran]

Also btw, Ryzen 5: http://www.anandtech.com/show/11202/amd-announces-ryzen-5-april-11th

 

I think the Ryzen 5 1500X will be the value sweet spot for gaming. Aggressive clock speeds, full L3 cache, and simple 4-core configuration (likely 1 CCX), all below 200$. It will face tough competition with the i5-7500 and i3-7350K.

[Draconian Guppy]

1 hour ago, Andre S. said:

Btw,

https://www.reddit.com/r/Amd/comments/5ybrxn/ryzen_7_is_actually_behaving_like_a_dual_4c8t/

But then AMD says this:

https://community.amd.com/community/gaming/blog/2017/03/13/amd-ryzen-community-update

 

Is it just me or something doesn't compute (no pun intended) here? How is Windows 10 scheduler behaving fine if it randomly assigns threads between "CPU Complexes"? How are developers supposed to optimize for that? Some technical details are missing here.

So i'm confused, comments say Ryzen is not a true 8 core?

[Andre S. Veteran]

10 minutes ago, Draconian Guppy said:

So i'm confused, comments say Ryzen is not a true 8 core?

It's 8 cores but the L3 cache is divided in two, each group of 4 cores (a CCX) having fast access to its own half. When they need access to the other half, it is about as slow as in a dual-CPU system. So, inferior to Intel's solution in that respect, where all cores on e.g. Broadwell-E having the fast low-latency access to all L3 cache. Depending on what your program is doing this may more or less impact performance. The killer is if your program is only using 4 threads and Windows randomly decides to put half on CCX1 and half on CCX2. And then AMD says Windows doesn't have an issue with scheduling there, so it is already aware of that configuration? Not clear what's going on.

[Andre S. Veteran]

Oh well AMD's apparently right, Windows does correctly accounts for CCX and SMT according to tests done by pcper: https://www.pcper.com/reviews/Processors/AMD-Ryzen-and-Windows-10-Scheduler-No-Silver-Bullet

 

Kinda disappointing, silver bullets are nice

[Alera]

34 minutes ago, Andre S. said:

Also btw, Ryzen 5: http://www.anandtech.com/show/11202/amd-announces-ryzen-5-april-11th

 

I think the Ryzen 5 1500X will be the value sweet spot for gaming. Aggressive clock speeds, full L3 cache, and simple 4-core configuration (likely 1 CCX), all below 200$. It will face tough competition with the i5-7500 and i3-7350K.

 

The Ryzen 5s will be Hexacore 3+3 and Quadcore 2+2, there will not be any 4+0 unfortunately, will have to wait for the Raven Ridge APUs to see that in action which somewhat funnily could make APUs the better choice for gaming other than simply choosing a 7700K.

 

But $220 for a 6 core is a very aggressive price and will surely be a good bargain for some out there. Probably lacking in games a little until any fixes can be applied.

[Luc2k]

16 hours ago, Circaflex said:

I think it is just a misunderstanding. I absolutely agree, AMD is not the only one to have bugs; however my point is over the last 10 years they have always had supposed "stellar performance," which is spun from their PR team. It usually never pans out; look at their high-end GPUs over the last few years. They continue to get beat by nVidias last generation and the follow up by AMD is always "let our drivers mature," or "coming next update." We saw some of the same non-sense with the FX resurrection, which was an embarrassment to the original FX series that kicked some serious butt, circa 2003.

 

Intel had had, still do to be honest, serious bugs with Skylake. They continue to say "fixes incoming." Believe it or not, I am not bias between the two companies. I'd love to run a great gaming rig with and AMD chip. I have very fond memories of running their chips and having a blast overclocking them many moons ago; I am just hesitant.

At least recently, I think its a case of fans and trolls driving everyone into a frenzy with unrealistic expectations. For example, AMD never hyped game performance before the release of Ryzen. They only showed some head to head and said it was similar to the 6900K. PR did fail with Fury and Bulldozer however.

 

Fair outlook and not something I can fault you for.

 

7 hours ago, xendrome said:

What exactly did I "outright lie" about?

About AMD GPU drivers - point 2 of your comment

 

7 hours ago, xendrome said:

The problem here is people like you that will only post positive propaganda and try to suppress anything negative about their beloved product. So by myself and others posting various articles about the negatives or truths of the product it hurts your feeling and you get personal by calling me a liar...

You must have me confused with somebody else as I've tried to be fairly objective in my posts and acknowledged the negatives of Ryzens launch (that would affect me) - motherboard issues, availability and gaming performance to a lesser extent. The part about performance is because manual OC mitigates it enough.

 

Ryzen is not my beloved anything, it is just a product that ticks most of the boxes on my upgrade needs list, but yes, I did call you out. The reason is that, when trying to discuss any of the articles or comments you've posted, you just ignored and carried on. Articles like the supported memory frequency which, at least partially, was proven untrue by both AMD and buyers after release. Or that hit piece from Toms Hardware about cooler screws which had no specifics and half of it was just "if" "should" and "could". Do you an update on that one with specific boards, now that they're are out, as it could be a "negative" piece that would actually help potential buyers.

Edited March 16, 2017 by Luc2k

[Andre S. Veteran]

5 minutes ago, Alera said:

The Ryzen 5s will be Hexacore 3+3 and Quadcore 2+2, there will not be any 4+0 unfortunately, will have to wait for the Raven Ridge APUs to see that in action which somewhat funnily could make APUs the better choice for gaming other than simply choosing a 7700K.

My bad for not reading. Wow, so unfortunate to keep that split for a 4 core part. But probably done to reduce costs. One can always dream some of the disabled cores will be unlockable (as in Athlon X3).

[Alera]

Just now, Andre S. said:

My bad for not reading. Wow, so unfortunate to keep that split for a 4 core part. But probably done to reduce costs. One can always dream some of the disabled cores will be unlockable (as in Athlon X3).

 

Would be a nice bonus for sure but they will likely do an Intel and laser the "faulty" cores off if they're really just using cut down Ryzen 7s for this. It will likely be harder too, the Ryzen's seem to like symmetry so unlocking 2+2 to 3+3 or 4+4 are possibly the only options so it'll be interesting to see if it's possible!

[xendrome]

9 minutes ago, Luc2k said:

About AMD GPU drivers - point 2 of your comment

 

You really had to dig through my comments about Nvidia graphics drivers related to their release cycle vs AMD? It's totally true, there were times when AMD was going MONTHS without releasing anything new, plus that has nothing to do with Ryzen which is what we are talking about.

[Draconian Guppy]

1 hour ago, Andre S. said:

It's 8 cores but the L3 cache is divided in two, each group of 4 cores (a CCX) having fast access to its own half. When they need access to the other half, it is about as slow as in a dual-CPU system. So, inferior to Intel's solution in that respect, where all cores on e.g. Broadwell-E having the fast low-latency access to all L3 cache. Depending on what your program is doing this may more or less impact performance. The killer is if your program is only using 4 threads and Windows randomly decides to put half on CCX1 and half on CCX2. And then AMD says Windows doesn't have an issue with scheduling there, so it is already aware of that configuration? Not clear what's going on.

Thanks for the thorough explanation.

 

Another question, how come this seems to be an issue only on 1080p ?

[Luc2k]

32 minutes ago, xendrome said:

You really had to dig through my comments about Nvidia graphics drivers related to their release cycle vs AMD? It's totally true, there were times when AMD was going MONTHS without releasing anything new, plus that has nothing to do with Ryzen which is what we are talking about.

I didn't dig through anything, it was just a topic I participated in. What you've posted hasn't been true for quite some time now and I would think that up to date information is valued on a tech site. From what I've noticed, you dislike AMD, and AMD makes Ryzen, so I think my point still stands.

3 hours ago, Mockingbird said:

Of cause it's going to support Windows 7 (64-bit) since AMD A-Series and AMD E-Series support Windows 7 (64-bit)

Going by board manufacturer drivers, Windows 7 to 10, both 32-bit and 64-bit, are supported. For example, AsRock Taichi doesn't support APUs. I mean, they might work, but the board has no way to output onboad video.

[Andre S. Veteran]

2 minutes ago, Draconian Guppy said:

Another question, how come this seems to be an issue only on 1080p ?

The lower the settings, the less the GPU becomes a bottleneck and the more you see differences between CPU.

[Draconian Guppy]

1 minute ago, Andre S. said:

The lower the settings, the less the GPU becomes a bottleneck and the more you see differences between CPU.

But wouldn't a high demand game tax the cpu more and this issue be more evident?

Sorry i need explanations like i'm five

[Andre S. Veteran]

20 minutes ago, Draconian Guppy said:

But wouldn't a high demand game tax the cpu more and this issue be more evident?

Sorry i need explanations like i'm five

For the same game, increasing resolution only increases GPU load, not CPU load. The CPU just tells the GPU "1920x1080" rather than "3840x2160", it's not any more difficult for it. Same goes for most graphical settings, the GPU will take all the additional burden.

[Noir Angel]

My guess is that the issue with CPUs being divided into 2 core complexes has something to do with yields. Ryzen CPUs are pretty complex (4.8 billion transistors and ~2KM of signal wiring on the 8 core variants), and the aggressive pricing probably made that particular design tradeoff necessary. As programming bottlenecks are removed and applications are updated to properly reflect the way Ryzen CPUs are designed I'd personally expect it to be a bit less of an issue going forward.

[Draconian Guppy]

33 minutes ago, Javik said:

My guess is that the issue with CPUs being divided into 2 core complexes has something to do with yields. Ryzen CPUs are pretty complex (4.8 billion transistors and ~2KM of signal wiring on the 8 core variants), and the aggressive pricing probably made that particular design tradeoff necessary. As programming bottlenecks are removed and applications are updated to properly reflect the way Ryzen CPUs are designed I'd personally expect it to be a bit less of an issue going forward.

From the pcper Andre S. Posted:

 

Quote

Closing Thoughts

What began as a simple internal discussion about the validity of claims that Windows 10 scheduling might be to blame for some of Ryzen's performance oddities, and that an update from Microsoft and AMD might magically save us all, has turned into a full day with many people chipping in to help put together a great story. The team at PC Perspective believes strongly that the Windows 10 scheduler is not improperly assigning workloads to Ryzen processors because of a lack of architecture knowledge on the structure of the CPU.

In fact, though we are waiting for official comments we can attribute from AMD on the matter, I have been told from high knowledge individuals inside the company that even AMD does not believe the Windows 10 scheduler has anything at all to do with the problems they are investigating on gaming performance. 

In the process, we did find a new source of information in our latency testing tool that clearly shows differentiation between Intel's architecture and AMD's Zen architecture for core to core communications. In this way at least, the CCX design of 8-core Ryzen CPUs appears to more closely emulate a 2-socket system. With that, it is possible for Windows to logically split the CCX modules via the Non-Uniform Memory Access (NUMA), but that would force everything not specifically coded to span NUMA nodes (all games, some media encoders, etc) to use only half of Ryzen. How does this new information affect our expectation of something like Naples that will depend on Infinity Fabric even more directly for AMD's enterprise play?

Doesn't look like there are any bottlenecks to be removed, except for game specific improvements  

[spaceelf]

Finally set up.  Some weirdness in the install, but it's done now.  Not a super friendly place to be an early adopter, apparently.

 

Holy ######## crap this is how computing should be, compared to my last box it's amazing.

 

I haven't really done benchmarks yet (or finished the build, since I was troubleshooting,) but considering I'm not doing any normal progression it probably won't mean that much to anyone but me.

[Alera]

16 hours ago, Draconian Guppy said:

From the pcper Andre S. Posted:

 

Doesn't look like there are any bottlenecks to be removed, except for game specific improvements  

1

Higher speed RAM and future improvements in Zen 2/3 that adds HBM2 or similar to the die should help with the inter-CCX communication but right now most motherboards are struggling to get higher than 2933MHz (3000) RAM working and some refuse to go about 2667MHz. I think this is how AMD are going to hit their projected 15% IPC improvement, I expect a little higher clock speeds and some small changes to the dies to boost performance just that bit closer to Intels arch.

[spaceelf]

9 minutes ago, Alera said:

Higher speed RAM and future improvements in Zen 2/3 that adds HBM2 or similar to the die should help with the inter-CCX communication but right now most motherboards are struggling to get higher than 2933MHz (3000) RAM working and some refuse to go about 2667MHz. I think this is how AMD are going to hit their projected 15% IPC improvement, I expect a little higher clock speeds and some small changes to the dies to boost performance just that bit closer to Intels arch.

They're apparently not supposed to work with 3000 as the memory controller doesn't have a divider for it.  3200 is supposedly fine.

[Draconian Guppy]

10 minutes ago, Alera said:

Higher speed RAM and future improvements in Zen 2/3 that adds HBM2 or similar to the die should help with the inter-CCX communication but right now most motherboards are struggling to get higher than 2933MHz (3000) RAM working and some refuse to go about 2667MHz. I think this is how AMD are going to hit their projected 15% IPC improvement, I expect a little higher clock speeds and some small changes to the dies to boost performance just that bit closer to Intels arch.

yeah, but what I meant is, for the current Zen, not unreleased products.

[spaceelf]

16 hours ago, Draconian Guppy said:

From the pcper Andre S. Posted:

 

Doesn't look like there are any bottlenecks to be removed, except for game specific improvements  

This build is the most beautiful thing I've ever used I can tell you that.  I haven't played too many games yet since I'm used to not doing too much else when downloading, but I was pretty far below average for PC builds I think.

[Alera]

28 minutes ago, Draconian Guppy said:

yeah, but what I meant is, for the current Zen, not unreleased products.

 

 

Yeah, I know! That's why I mentioned the higher speed RAM, it affects the Infinity Fabric speed which runs at half the speed of the memory in your system so 2133MHz DDR4 is 1066MHz inter-CCX speed, 2933MHz/1466MHz, 3200MHz/1600MHz is possibly the highest speed they might reach but also possible might be 3600MHz/1800MHz which is almost double. The effect that the CCX penalty has is 40-60ns latency from core to core within a CCX but 140ns latency from CCX to CCX. So even with conservative estimates, AMD and AIBs could likely fix this with better BIOS support for faster RAM making it much closer to the Intel side of things and improving performance across the board.

 

Of course, game engine optimisations will help too, Vulkan and OpenGL have already shown that Ryzen can keep up with the 7700K at 1080p when cores are being utilised properly, 165 FPS for both 1800X and 7700K with the 6950X reaching 162 FPS in that test compared to Intels 190 FPS in DX12 and 1800X only hitting 129 FPS.