AMD News Galore: Threadripper, EPYC, Ryzen Pro Processors, Integrated Vega Graphics For APUs, 7nm Ryzen Roadmap

[kyotokid]

Exactly my point, in a round about way. Tailoring a cpu specifically for just amd gpus seems pointless, ppl pushing the kind of power you describe would also likely be using that DGX-1 tesla or similar if gpu rendering as firepros don't have anything quite as close. So it'd have to be a play on wording that all that power with all those pcie lanes etc will actually be able to make better use of amd gpus, vrs designing a cpu to take on the Xeon Reign that cuts out a very large portion of the work base.

..the other downside of any these these CPUS (Ryzen, Threadripper, Epyc) is you cannot run an older version of Windows like 7 or 8.1 as the first two no doubt will not support anything but W10/Linux and Epyc, only Linux. Now nothing against Linux, but if you are a 3D CG artist there isn't much in the way of software that supports it (even the "industry standard" 3DS Max doesn't, let alone enthusiasts' programmes like Vue, Carrara and such).

Bedsides the force fed updating, telemetry reporting, a ridiculous personified digital assistant, and the possibility, (even though MS won't say it yet) of becoming a cloud based subscription OS, W10 presents a fairly major issue for those who work with GPU rendering in that it also "reserves" a noticeable portion of GPU VRAM. So instead of nearly 12 GB, that 1,200$ Titan XP effectively has around 9 GB available for rendering instead. W7 and Linux on the other hand have an almost unnoticeable impact

Speaking of Xeons, I still am pretty settled on going ahead with my dual 8 core Sandy Bridge Xeon, 128 GB quad channel, dual 1080 Ti workstation (I still also use biased engines that do not support GPU rendering hence the boatload of memory). Yeah only 32 CPU threads total but it will run on W7 Pro. While a single Threadripper workstation with 32 CPU threads and enough PCIe lanes to support 4 GPUs (at a lower price than Intel's Broadwell-E series) would be really nice, dealing with W10s shortcomings isn't worth it.

Oh and just saw brief mention of Intel rolling out the first i9 CPU on Ars Technica (why they keep avoiding even numbered CPUs is anyone's guess) though the top end in that family is to have 12 not 16 cores.

 

[bit_user]

I find it a little bit of too much to design any of those cpus around performance bonuses for amd gpus. That's gotta be either a coincidence or a play on words or something.

Initially, I took Lisa Su's statement to mean just that they had equipped the CPUs with enough PCIe 3.0 lanes to provide full x16 links to n GPUs.

However, AMD is a key (founding?) member of the HSA Foundation. Perhaps she means that it's designed to support their GPUs in a fully HSA-compliant system (which I think is likely). Although HSA seems primarily designed around APUs, I know they've stated it can/will also support discrete GPUs.

http://developer.amd.com/resources/heterogeneous-computing/what-is-heterogeneous-system-architecture-hsa/
http://www.amd.com/en-us/innovations/software-technologies/hsa
http://www.hsafoundation.com/

 

[bit_user]

Oh and just saw brief mention of Intel rolling out the first i9 CPU on Ars Technica (why they keep avoiding even numbered CPUs is anyone's guess) though the top end in that family is to have 12 not 16 cores.

IMO, i9 always should've been used for the extreme desktop line. It causes too much confusion to call them i7's.

As for skipping even, maybe they just wanted a numbering scheme that better reflected their relationship to each other. I don't really know, but at least their desktop CPU numbering scheme makes more sense than mobile, where you can get i3's, i5's, and i7's that have 2 cores / 4 threads.

 

[genz]

Interesting thing is that Zeppelin's die so far has contained large s space dedicated to GPU. If this is indeed 4 Zep dies that leaves either a massive space for a Greenland IF GPU onboard, or a massive empty space for them to pile the cores on later!

 

[jimmysmitty]

Correct they gave no name but they verbally did state a 2018 refresh before zen 2. Slide three they list 14nm+ below 14nm aka the refresh. They also verbally mentioned it.. Link to slide deck below.

http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-analystday

That's somewhat encouraging.

Rumours have Coffee Lake launching with a 6 core i7 on the mainstream platform.That could well (speculating here) mean hyperthreaded i5s. *If* that turns out to be true and *if* Intel stick with the traditional i7/i5 prices (both are massive "ifs" at this point), then the entire first gen Ryzen lineup starts to look a whole lot less attractive. A (hypothetical) 4C8T i5 would get very close to the top end Ryzen 5 lineup in multithreaded tasks while maintaining Intel's single threaded and gaming advantages. A (hypothetical) 6C12T i7 would do the same to the entire Ryzen 7 lineup.

AMD have a really competitive CPU lineup right now, but Intel is hardly going to sit still. Obviously it's all "ifs" and speculation at this point, but right now AMD's prime selling point is cores/threads per dollar. If Intel attack them on that front then AMD will need to answer with bumps in IPC or clock speed (probably both) to stay competitive in the $150 - $500 market. If Ryzen 2 is tied to 7nm and likely 2019, I think there will be quite a lot riding on a Zen+/Zen1.5 lineup on 14nm.

I am actually wondering if Intel is pushing LGA2066 to the mainstream and enthusiast as a single socket. It would be ideal honestly, make it easier and it is obviously possible since they will have a Kaby Lake X with only dual channel memory support.

If AMD has to push clock speed to compete with Intel I feel they will fail till their next process at least. The 14nm they have is not nearly as good for clock speed as Intel.

 

[InvalidError]

I am actually wondering if Intel is pushing LGA2066 to the mainstream and enthusiast as a single socket. It would be ideal honestly, make it easier

I doubt much of the mainstream sees the appeal of buying a $200+ motherboard where 2/3 of the features don't work with mainstream CPUs due to lack of PCIe lanes and extra memory channels on the CPU.

Last I heard, mainstream Coffee Lake may end up using LGA1151v2.

 

[JamesSneed]

I am actually wondering if Intel is pushing LGA2066 to the mainstream and enthusiast as a single socket. It would be ideal honestly, make it easier

I doubt much of the mainstream sees the appeal of buying a $200+ motherboard where 2/3 of the features don't work with mainstream CPUs due to lack of PCIe lanes and extra memory channels on the CPU.

Last I heard, mainstream Coffee Lake may end up using LGA1151v2.

Yeah once we see reviews I don't think many will bite on the quad cores in the 2066 socket. Its a very hard sell as the ones that would bite are going to delid, OC the hell out of it, likely on water. Most of those same people would gladly snag the skylake-x models since they have the budgets. Those on a budget would just get the 7700K which is a cheaper CPU and motherboard with what I expect to be nearly identical performance. The whole but you can upgrade later argument is a hard sell to someone on a tight budget. Those new quads on 2066 are just awkward.

 

[InvalidError]

The whole but you can upgrade later argument is a hard sell to someone on a tight budget. Those new quads on 2066 are just awkward.

The whole upgrade argument is an even tougher sell when there is almost no mainstream software that requires any such thing, octo cores are already relatively affordable on AMD's mainstream platform and hex cores are going to be standard on Intel's mainstream CPUs with Coffee Lake.

The worst part for LGA2066 is the amount of extra added cost on those motherboards needed to re-route some of the CPU's PCIe lanes via switch ICs to at least enable people to pick between features when the installed CPU doesn't have enough PCIe lanes to feed everything. Finding a motherboard that allows you to enable all the features you want with a 16 lanes CPU is going to be a nightmare.

 

[jimmysmitty]

I am actually wondering if Intel is pushing LGA2066 to the mainstream and enthusiast as a single socket. It would be ideal honestly, make it easier

I doubt much of the mainstream sees the appeal of buying a $200+ motherboard where 2/3 of the features don't work with mainstream CPUs due to lack of PCIe lanes and extra memory channels on the CPU.

Last I heard, mainstream Coffee Lake may end up using LGA1151v2.

The majority of motherboard features pull from the PCH PCIe pool. The majority of CPU PCIe lanes go to PCIe slots.

As for pricing, the boards most people stick to are around the $200 range for mainstream. There are $250 dollar x99 boards that are just as good.

Also, Kaby Lake X has only 16 PCIe lanes yet is going to be on LGA2066. So what would stop Intel from putting Coffee/Cannon Lake on it instead to simplify the platforms and offer more value in easier upgrades?

While you are probably correct there is no real information out there about it. Best we have is a possible 6 core i7 and that socket but I can only find a few random posts from not very reputable sites about Coffee Lake.

 

[bit_user]

Also, Kaby Lake X has only 16 PCIe lanes yet is going to be on LGA2066. So what would stop Intel from putting Coffee/Cannon Lake on it instead to simplify the platforms and offer more value in easier upgrades?

You mean instead of 1151,v2 or whatever? One word: cost. Those extra memory channels & PCIe lanes need to be routed, and that costs money. Not to mention support for 140+ W TDPs. Unless you're talking about TDP-restricted boards with only 16 lanes & 2 memory channels actually connected, which seriously hurts the upgrade argument.

The whole proposition sounds like a real headache, to me. And does 2066 even support an iGPU? Given that the Kabylake-X CPUs have theirs disabled, I'd guess not.

 

[jimmysmitty]

Also, Kaby Lake X has only 16 PCIe lanes yet is going to be on LGA2066. So what would stop Intel from putting Coffee/Cannon Lake on it instead to simplify the platforms and offer more value in easier upgrades?

You mean instead of 1151,v2 or whatever? One word: cost. Those extra memory channels & PCIe lanes need to be routed, and that costs money. Not to mention support for 140+ W TDPs. Unless you're talking about TDP-restricted boards with only 16 lanes & 2 memory channels actually connected, which seriously hurts the upgrade argument.

The whole proposition sounds like a real headache, to me. And does 2066 even support an iGPU? Given that the Kabylake-X CPUs have theirs disabled, I'd guess not.

How would Coffee Lake differ from Kaby Lake X? Kaby Lake X has only dual channel DDR4 and 16 PCIe lanes all on the LGA2066 socket. The channels wont be an issue. You can run a LGA 2011 board in single, dual or triple channel besides quad channel so the system can turn the extra channels off without having to make two distinct motherboards. The PCIe is the same. Right now if you buy a board that has dual x16 and dual x8 slots (4 slots total) then you need a 6850K or better to be able to use all 4 slots. However the board still supports the 6800K and just has 28 PCIe lanes so 3 x8 and 1 x4 slots.

Plus most board manufactures already have specs coming out and the Asus Prime X299-Deluxe looks to show support for both Kaby Lake X and Skylake X:

https://www.asus.com/Motherboards/PRIME-X299-DELUXE/

All it will do is turn off what is not supported by the CPU in PCIe lanes and the memory channels. I doubt there will be any X299 boards geared only to Kaby Lake X, however if they do make a dual channel only board it might allow the manufactures to make cheaper boards that still can support the higher end Skylake X CPUs but with fewer memory channels and less traces/layers. I don't think that would be a horrible idea.

And of course it wouldn't support iGPUs. However since we have until 2018 maybe they will launch a different chipset. It is just an idea but a very plausible one since it would just be Kaby Lake X.

 

[InvalidError]

Plus most board manufactures already have specs coming out and the Asus Prime X299-Deluxe looks to show support for both Kaby Lake X and Skylake X:

Having support for both CPUs isn't a problem. What is would be what gets neutered when you use a KLX CPU (which is just regular Kaby Lake on an LGA2066 substrate) on x299. With the board's functionality practically reduced to z270 level or worse depending on which features the motherboard manufacturer attached to the throw-away lanes, it makes very little practical sense. Even more so when you consider that KLX will be superseded by Coffee Lake bringing two extra cores to the mainstream platform whenever that launches.

 

[bit_user]

How would Coffee Lake differ from Kaby Lake X?

If you're proposing a Coffee Lake-X, then I apologize. I'd agree it would make no less sense than Kaby-X. I thought you were saying mainstream Coffee Lake would use LGA2066.

If I did understand you correctly, then I think the x299 is just too expensive for mainstream. Plus, IE has some good points about the complexities of supporting all 3 PCIe configs.

You can run a LGA 2011 board in single, dual or triple channel besides
quad channel

Um, proof? I have a LGA2011 v1 board and you definitely can't run triple-channel on it. I'd bet LGA2011-3 is the same.

 

[jimmysmitty]

As massively complex as these new cpus are as compared to older cpus, as much as they are packing into the die just to enable that amount of pcie lanes, clock speeds, mc, everything, I find it a little bit of too much to design any of those cpus around performance bonuses for amd gpus. That's gotta be either a coincidence or a play on words or something. It'd be far easier to have designed the 500 series to be better compatible with the new cpus, than the other way around.

...makes much better sense as then they would be compatible with either brand of GPU and thus maybe net more sales.

A workstation with Dual Epycs and 1 TB of 16 channel physical memory would be pretty shredding with biased raytrace render engines that don't support GPU rendering like Vue, Carrara, and Renderman/3DL. It could probably even knock out a big scene in Iray in a more reasonable amount of time.

Yes but if you look at past marketing, AMD always markets their whole platform. They make more money for that than just one or the other so their marketing will always sell their platform as more beneficial together rather than "mix and match". However if you actually do the testing, Radeon will not benefit from AMD just due to the platform.

As for memory configurations, it may depend on the manufacture but this says you can:

http://www.legitreviews.com/ddr3-memory-performance-analysis-on-intel-x79_1779

And that is for X79, LGA 2011-V1, so I assume X99 would easily be able to run the same configurations.