Next Generation Hardware Speculation with a Technical Spin [pre E3 2019]

[anexanhume]

Those commits were done by AMD.

C) Sony developer Code Reviews optimizations for Zen 2.

Simon Pilgrim commits plenty on his own, including AVX512.

 

[BRiT]

Simon Pilgrim commits plenty on his own, including AVX512.

We're talking about this Specific Rumor that is tied to this Specific Commit, that was done by an AMD employee.

As far as thIs rumor goes, it is Non Sequitur.

 

[anexanhume]

We're talking about this Specific Rumor that is tied to this Specific Commit, that was done by an AMD employee.

As far as thIs rumor goes, it is Non Sequitur.

The commit itself states that it's simply a clone of the Zen1 capabilities for now, so you could reasonably assume that any Zen1 commits prior were placeholders for Zen2 optimizations as well.

 

[BRiT]

The commit itself states that it's simply a clone of the Zen1 capabilities for now, so you could reasonably assume that any Zen1 commits prior were placeholders for Zen2 optimizations as well.

None of which is evidence of PS5 being Zen2. Hence, this rumor is entirely Non Sequitur.

 

[anexanhume]

None of which is evidence of PS5 being Zen2. Hence, this rumor is entirely Non Sequitur.

The rumor is not predicated on the link. It was completely independent of that. It existed well before that was made public.

 

[BRiT]

The rumor is not predicated on the link. It was completely independent of that. It existed well before that was made public.

Once again, we were talking about the latest rumor that is linked to that latest commit.

 

[Deleted member 11852]

B) Sony’s lead developers work with LLVM and use it internally for their tools and have been working with it for a long time. I suppose we should see The PS3 processor in LLVM then ? I’m trying to explore precedent.

Try not to get hung up on precedent. If you only accept ideas and concepts for which there is precedent, you automatically reject everything which is new.

But the precedent is surely, PS4?

 

[milk]

Once again, we were talking about the latest rumor that is linked to that latest commit.

Ok, but how many different letters are in the word "commit"? Five. As in, PS5, maybe?

 

[anexanhume]

Once again, we were talking about the latest rumor that is linked to that latest commit.

There is no rumor. It’s not a questionable piece of information sourced from a journalist. It’s a hard fact in a public-facing repository. We’re debating over what conclusions to draw, not whether to believe the source.

 

[Shifty Geezer]

There is no rumor. It’s not a questionable piece of information sourced from a journalist. It’s a hard fact in a public-facing repository. We’re debating over what conclusions to draw, not whether to believe the source.

You stated that Zen 2 was confirmed and cited the commit as evidence:

Zen 2 confirmed for PS5.
https://reviews.llvm.org/D58343

The rumour is from you, that PS5's has Zen 2, as confirmed by an LLVM commit. Without concrete evidence, it's a rumour, and the evidence has not been explained as proof of your claim.

 

[London Geezer]

You stated that Zen 2 was confirmed and cited the commit as evidence:

The rumour is from you, that PS5's has Zen 2, as confirmed by an LLVM commit. Without concrete evidence, it's a rumour, and the evidence has not been explained as proof of your claim.

But is it a concrete rumour?

 

[iroboto]

Those commits were done by AMD.

C) Sony developer Code Reviews optimizations for Zen 2.

okay welp, I screwed that up. That really makes this commit to Znver2 pretty irrelevant on whether PS5 is going Zen2.
As per this article:
https://www.phoronix.com/scan.php?page=news_item&px=AMD-Znver2-LLVM-Clang-9
Seems to coincide with AMD just getting the major compilers ready to support Zen2 which they started back in Oct 2018.

This is the basic patch introducing the next-generation AMD Zen CPU to the GCC compiler collection. At this stage it's just the basic implementation and carries over the same cost tables and scheduler data from Znver1. So it doesn't reveal any major breakthrough changes, but in digging through the code, it does confirm some new CPU instructions that will be supported by these next-gen Zen CPUs... On top of the Znver1 instructions, Zen 2 is adding:

- Cache Line Write Back (CLWB)

- Read Processor ID (RDPID)

- Write Back and Do Not Invalidate Cache (WBNOINVD)

That's it in terms of new instructions, at least what's enabled by these patches. It's possible there might be some other new instructions supported by Zen 2 that AMD doesn't want to reveal at this time, just like the scheduler cost tables haven't yet been tuned, etc. This patch is basically a starting point so the GCC 9.1 stable update due out in 2019 can at least handle -march=znver2 and that march=native targeting will also work for these next-gen AMD processors.

So to me it's just looks like due diligence by AMD to make sure their processors are properly supported. I see no link here.

 

[iroboto]

Try not to get hung up on precedent. If you only accept ideas and concepts for which there is precedent, you automatically reject everything which is new.

But the precedent is surely, PS4?

Agreed, I'm just exploring. Yea I would assume PS4 would have been precedent. Back in 2013 I'm not sure if we have access to those commits, but would we find something like Simon dropping commits that optimize for what became PS4 architecture, before PS4 was released?

AVX512 discussion is interesting because I don't think AVX512 is even confirmed for Zen 2? Perhaps my GoogleFu is falling behind, but aside from some rumours, I don't see anything about confirmation of that instruction set.

AFAICS at this moment, Simon's optimizations are being used for Intel processors. I will loop back to my original theory that these commits support tool optimization (studios need more bang for buck for their hardware, video encoding etc) and not the console hardware.

Are we sure there isn't some sort of mixup on AVX with AVX512 in the code? As Jaguar supports AVX

 

[anexanhume]

Agreed, I'm just exploring. Yea I would assume PS4 would have been precedent. Back in 2013 I'm not sure if we have access to those commits, but would we find something like Simon dropping commits that optimize for what became PS4 architecture, before PS4 was released?

AVX512 discussion is interesting because I don't think AVX512 is even confirmed for Zen 2? Perhaps my GoogleFu is falling behind, but aside from some rumours, I don't see anything about confirmation of that instruction set.

AFAICS at this moment, Simon's optimizations are being used for Intel processors. I will loop back to my original theory that these commits support tool optimization (studios need more bang for buck for their hardware, video encoding etc) and not the console hardware.

Are we sure there isn't some sort of mixup on AVX with AVX512 in the code? As Jaguar supports AVX

AVX512 commits are expressly called out as such. It’s not a stretch to think that Zen 2 could support AVX512 by double cycling its 256 bit ALUs, just as their current architectures do on 128-bit units for AVX2.

Someone pointed out (I believe Phoronix) that AVX512 was not in the list of new instructions when Znver2 first popped up, but I believe it was only required to submit instructions new to LLVM, not the architecture.

If you look through Simon’s commits, a lot of them reference the Skylake architecture, one with a more basic AVX512 implementation, rather than the newest which have added more instructions. It seems to be somewhat of a reference point.

You stated that Zen 2 was confirmed and cited the commit as evidence:

The rumour is from you, that PS5's has Zen 2, as confirmed by an LLVM commit. Without concrete evidence, it's a rumour, and the evidence has not been explained as proof of your claim.

Since we’re venturing into pedantry here, what constitutes concrete evidence? Are we going to keep debating this even after it releases, a la the Xbox One X doesn’t use Jaguar discussion? Clearly the only thing that would truly satisfy this is a full disclosure of the architecture, which we won’t get until reveal time, if at all. I think we’re being a bit dense with the rigor we’re expecting at this point, especially since 0 people in this thread seem to contend that Sony isn’t using a Zen derivative.

 

[AlNom]

especially since 0 people in this thread seem to contend that Sony isn’t using a Zen derivative.

I think the only way that would happen is if they were willing to pull a Nintendo, and that might come down to cost/benefit to the company as a whole (not feeding some console war), and be pretty far out there with a number of additional factors aligning to make that decision.

 

[iroboto]

AVX512 commits are expressly called out as such. It’s not a stretch to think that Zen 2 could support AVX512 by double cycling its 256 bit ALUs, just as their current architectures do on 128-bit units for AVX2.

Zen 1 doesn't have it. I'll be honest, not sure if that means they can just jump it with Zen 2. I guess this is a wait and see type of thing. If it was a new CPU architecture entirely I could see AVX512 as a definite.

Are we going to keep debating this even after it releases, a la the Xbox One X doesn’t use Jaguar discussion?

I'm the only one still left on this, and only because we found reference in a textbook about it. The only reason this point is worth discussing isn't because it's performance matters, but whether that has implications into next gen.

since 0 people in this thread seem to contend that Sony isn’t using a Zen derivative.

I have some reservations still. I think there are a lot of fantastic reasons to go Zen 2. At the heart of my reservation is mainly price. Followed closely secondly by backwards compatibility. You have to look at this from AMD's perspective. You have a processor that can finally compete with Intel. Are you really going to sell it at rock bottom prices and have all your fabs producing 120 Million APUs for nothing? There either is going to be good margins on this coming APU or it's not Zen. It's unlikely to be a chiplet either, so binning isn't a possibility.

At the bottom barrel of PC parts I can find the following:
4 Core Ryzen 3 (120 CAD)
Mobo (109 CAD)
Memory (145)
GPU Radeon 580 (319)
Sitting at 693 dollars and haven't gotten into anything else yet like a disc drive, power, case, fans and storage.

And that's pretty bottom of the barrel. The idea we're going to get this 8 core Zen 2 + 11 TF of power at 499 CAD? hmmmm. I was way wrong before and learned my lesson with Scorpio thinking it was going to be Zen 1.

Going to quote a blog post from Agner: https://www.agner.org/optimize/blog/read.php?i=838

Let us compare the execution units of AMD's Ryzen with current Intel processors. AMD has four 128-bit units for floating point and vector operations. Two of these can do addition and two can do multiplication. Intel has two 256-bit units, both of which can do addition as well as multiplication. This means that floating point code with scalars or vectors of up to 128 bits will execute on the AMD processor at a maximum rate of four instructions per clock (two additions and two multiplications), while the Intel processor can do only two. With 256-bit vectors, AMD and Intel can both do two instructions per clock. Intel beats AMD on 256-bit fused multiply-and-add instructions, where AMD can do one while Intel can do two per clock. Intel is also better than AMD on 256-bit memory writes, where Intel has one 256-bit write port while the AMD processor has one 128-bit write port. We will soon see Intel processors with 512-bit vector support, while it might take a few more years before AMD supports 512-bit vectors. However, most of the software on the market lags several years behind the hardware. As long as the software uses only 128-bit vectors, we will see the performance of the Ryzen processor as quite competitive. The AMD can execute six micro-ops per clock while Intel can do only four. But there is a problem with doing so many operations per clock cycle. It is not possible to do two instructions simultaneously if the second instruction depends on the result of the first instruction, of course. The high throughput of the processor puts an increased burden on the programmer and the compiler to avoid long dependency chains. The maximum throughput can only be obtained if there are many independent instructions that can be executed simultaneously.

This is where simultaneous multithreading comes in. You can run two threads in the same CPU core (this is what Intel calls hyperthreading). Each thread will then get half of the resources. If the CPU core has a higher capacity than a single thread can utilize then it makes sense to run two threads in the same core. The gain in total performance that you get from running two threads per core is much higher in the Ryzen than in Intel processors because of the higher throughput of the AMD core (except for 256-bit vector code).

So here, Agner doesn't believe in 512-bit vector support coming soon for AMD, thus AVX512 is in question again.
b) Ryzen sounds amazing for server loads. But game code is a bit different from that. Not sure if all this extra juice is required.

AMD has a different way of dealing with instruction set extensions than Intel. AMD keeps adding new instructions and remove them again if they fail to gain popularity, while Intel keeps supporting even the most obscure and useless undocumented instructions dating back to the first 8086. AMD introduced the FMA4 and XOP instruction set extensions with Bulldozer, and some not very useful extensions called TBM with Piledriver. Now they are dropping all these again. XOP and TBM are no longer supported in Ryzen. FMA4 is not officially supported on Ryzen, but I found that the FMA4 instructions actually work correctly on Ryzen, even though the CPUID instruction says that FMA4 is not supported.

I mean, what would happen if your customized CPU was using those instruction sets and AMD didn't carry it over. A bit annoying I imagine, and depending on how deep you customized it, you're going to have to do the same thing all over again with the next architecture. Time is money, and if the goal is to keep the console at $399 USD, we need to make some very efficient trades.

 

[Gubbi]

AVX512 commits are expressly called out as such. It’s not a stretch to think that Zen 2 could support AVX512 by double cycling its 256 bit ALUs, just as their current architectures do on 128-bit units for AVX2.

The problem with supporting AVX-512 is that there are 18 (EIGHTEEN!!!) distinct subsets of instructions in AVX-512, they are:
F, CD, ER, PF, 4FMAPS, 4VNNIW, VPOPCNTDQ, VL, DQ, BW, IFMA, VBMI, VNNI, VBMI2, BITALG, VPCLMULQDQ, GFNI and VAES

I can see why AMD won't try to be compliant until the dust settles.

They are going to have the raw computational throughput of two AVX-512 units with their four full width AVX2 units and they'll have similar bandwidth to/from their FP units as Intel does. There are some useful instructions in AVX-512, like gather and scatter that they'll miss out on, but it's pretty easy to saturate a memory subsystem with 6+ cores without those instructions anyway.

Cheers

 

[Shifty Geezer]

Since we’re venturing into pedantry here, what constitutes concrete evidence?

It's not pedantry. You stated, absolute, that Zen2 was confirmed for PS5. The evidence presented for such an statement must be similarly absolute and undebatable. If you want to say, "does this point to Zen2 confirmation in PS5?" then you'll get a very different discussion. If you want to stick by your point that Zen2 is confirmed for PS5, link to an unquestionably clear leaked document or executive announcement. Everything else, even a DF confirmed insider rumour, is speculation, not confirmation.

 

[Xbat]

Surely though I doubt Sony or Microsoft would of spent the R&D to get Zen 1 or any other Arch to 7nm when AMD were already designing Zen 2 for 7nm?

 

[anexanhume]

I think the only way that would happen is if they were willing to pull a Nintendo, and that might come down to cost/benefit to the company as a whole (not feeding some console war), and be pretty far out there with a number of additional factors aligning to make that decision.

Lisa Su has stated AMD is working with both Sony and MS on their “secret sauces.” They’re also predicting a semi custom revenue bump coming.

Zen 1 doesn't have it. I'll be honest, not sure if that means they can just jump it with Zen 2. I guess this is a wait and see type of thing. If it was a new CPU architecture entirely I could see AVX512 as a definite.

What constitutes a new architecture in your mind? What has been disclosed on. Zen 2 shows a revamped front-end, and a vastly different FP arrangement.

I have some reservations still. I think there are a lot of fantastic reasons to go Zen 2. At the heart of my reservation is mainly price. Followed closely secondly by backwards compatibility. You have to look at this from AMD's perspective. You have a processor that can finally compete with Intel. Are you really going to sell it at rock bottom prices and have all your fabs producing 120 Million APUs for nothing? There either is going to be good margins on this coming APU or it's not Zen. It's unlikely to be a chiplet either, so binning isn't a possibility.

Lot of supposition there. What matters to AMD is which fabs the console makers will be using. If it’s 7nm, what’s the difference to them which design occupies that die space. Wouldn’t it make more sense to have them use a standard Zen 2 chiplet so that their binning options for console, desktop or mobile expand vastly? The idea that using the latest would somehow hamper their ability to compete with Intel or magically lose revenue does not seem to have a basis to me.

At the bottom barrel of PC parts I can find the following:
4 Core Ryzen 3 (120 CAD)
Mobo (109 CAD)
Memory (145)
GPU Radeon 580 (319)
Sitting at 693 dollars and haven't gotten into anything else yet like a disc drive, power, case, fans and storage.

And that's pretty bottom of the barrel. The idea we're going to get this 8 core Zen 2 + 11 TF of power at 499 CAD? hmmmm. I was way wrong before and learned my lesson with Scorpio thinking it was going to be Zen 1.

Why are we comparing retail prices? When has that ever been a sound basis for costing of a console BOM?

Going to quote a blog post from Agner: https://www.agner.org/optimize/blog/read.php?i=838


So here, Agner doesn't believe in 512-bit vector support coming soon for AMD, thus AVX512 is in question again.

That seems to be reading into what the blog post says. Moreover, Zen 1 is the first of a new approach for AMD. To suggest we have all this precedent for what Zen 2 will be seems to be liberally misappropriating history.

It's not pedantry. You stated, absolute, that Zen2 was confirmed for PS5. The evidence presented for such an statement must be similarly absolute and undebatable. If you want to say, "does this point to Zen2 confirmation in PS5?" then you'll get a very different discussion. If you want to stick by your point that Zen2 is confirmed for PS5, link to an unquestionably clear leaked document or executive announcement. Everything else, even a DF confirmed insider rumour, is speculation, not confirmation.

*Checks thread title* yup, we’re good here. Was I given a scepter of you’re-not-allowed-to-disagree with-my-conclusions? Not that I saw, but I could be wrong I suppose.