News AMD previews Ryzen 3rd generation at CES

[Hitman928]

More info to come. . .

 

[Hitman928]

https://www.thestreet.com/investing...coming-products-14833554?puc=CNBC&cm_ven=CNBC

Mark Papermaster interview. Pulled out some quotes:

I asked if AMD's third-gen Ryzen Mobile chips might rely on TSMC's 7nm+ manufacturing process, which is set to enter volume production later this year and delivers moderate performance and power improvements relative to standard 7nm.

Papermaster declined to say whether this will be the case. But he did suggest that AMD is looking to launch third-gen Ryzen Mobile in 2020, a point in time when 7nm+ volume production should be well underway. "If you look at the [product launch] cycle, next year would be a natural time for that migration," he said. "We'll have more details on that as we get closer to that time."

Papermaster said AMD is mindful of the single-threaded performance gap that has remained for Ryzen, and promised his company will deliver "very exciting gains" in this area while maintaining its multi-threaded performance lead. "What you will see with our third-generation Ryzen really is simply outstanding gaming performance," he declared.

When asked whether AMD is interested in adding similar [RTX] cores to its GPUs -- they're not present on the Radeon VII -- Papermaster indicated AMD would do so if there's strong customer interest. "We look at fixed-function acceleration, and I have to say it's very much customer-driven," he said, while noting AMD has long integrated accelerators within its GPUs that improve the performance of audio and video codecs.

When asked about his thoughts on ARM in the data center, Papermaster stated AMD "looked hard at ARM" and decided -- given that ARM's software ecosystem was still immature -- it made sense for AMD to direct its attention towards launching high-performance x86 products. At the same time, he admitted that ARM-based server CPUs will continue providing competition.

 

[jpiniero]

Actually AMD is rumored to be using Samsung's 7 EUV node next.

 

[mattiasnyc]

About the reserved PCI-E lanes, Zen 1 die had 4 unused pcie lanes on AM4, if the Ryzen I/O die still has the 32 PCI-E lanes they may be able to run a IGP on it, if they are PCI-E 4.0.

The Zen 1 die had four unused lanes?

 

[HurleyBird]

The Zen 1 die had four unused lanes?

That's why 1P Epyc supports 128 lanes. 32 x 4 dies.

 

[Thunder 57]

The Zen 1 die had four unused lanes?

Rumor is AM4 can't handle them though and we would need a new socket with more pins. So I've read.

 

[JimKiler]

I think you would be surprised. Not to say gaming isn't on many peoples minds here.

EDIT

As for why I agree I am glad to see more AVX2 performance, simple. x265 transcoding.

Transcoding would be a valid reason then. I did not realize it uses AVX2 instructions.

 

[Abwx]

Rumor is AM4 can't handle them though and we would need a new socket with more pins. So I've read.

"Our sources tell us that after unlocking the feature via a BIOS update, the older motherboards supply a PCIe 4.0 x16 connection to the first slot on the motherboard, but the remainder of the slots revert to PCIe 3.0 signaling rates. That's because any trace routing on the motherboard that exceeds six inches requires newer redrivers and retimers that support PCIe 4.0's faster signaling rates. That means the PCIe slot nearest to the CPU will easily support PCIe 4.0, while the other slots, including M.2 ports, will run at a PCIe 3.0 signaling rate."


Update: We spoke with AMD representatives, who confirmed that 300- and 400-series AM4 motherboards can support PCIe 4.0. AMD will not lock the out feature, instead it will be up to motherboard vendors to validate and qualify the faster standard on its motherboards on a case-by-case basis. Motherboard vendors that do support the feature will enable it through BIOS updates, but those updates will come at the discretion of the vendor. As mentioned below, support could be limited to slots based upon board, switch, and mux layouts.



https://www.tomshardware.com/news/amd-ryzen-pcie-4.0-motherboard,38401.html

 

[Shivansps]

The Zen 1 die had four unused lanes?

yes, TR has 64 lanes and EPYC 128, this is only possible if the Zen 1 die has 32 PCI-E, chances are there is no enoght pins on AM4.

 

[Thunder 57]

Transcoding would be a valid reason then. I did not realize it uses AVX2 instructions.

Yup, that will have me looking at Zen 2. When doing x264 -> x264, it's damn fast. x265 -> x264 though and I pay the AVX penalty.

"Our sources tell us that after unlocking the feature via a BIOS update, the older motherboards supply a PCIe 4.0 x16 connection to the first slot on the motherboard, but the remainder of the slots revert to PCIe 3.0 signaling rates. That's because any trace routing on the motherboard that exceeds six inches requires newer redrivers and retimers that support PCIe 4.0's faster signaling rates. That means the PCIe slot nearest to the CPU will easily support PCIe 4.0, while the other slots, including M.2 ports, will run at a PCIe 3.0 signaling rate."

Update: We spoke with AMD representatives, who confirmed that 300- and 400-series AM4 motherboards can support PCIe 4.0. AMD will not lock the out feature, instead it will be up to motherboard vendors to validate and qualify the faster standard on its motherboards on a case-by-case basis. Motherboard vendors that do support the feature will enable it through BIOS updates, but those updates will come at the discretion of the vendor. As mentioned below, support could be limited to slots based upon board, switch, and mux layouts.

https://www.tomshardware.com/news/amd-ryzen-pcie-4.0-motherboard,38401.html

I had read that, and, if true, it would be unprecedented. Using an old board and chipset for a new CPU and also upgrading the old board's first slot to PCIe 4? That would be impressive and reminds me of the days when CPU upgrades made sense rather than platform upgrades.

I have my doubts mobo manufactures will go out of their way to do that though, as they will want to sell new boards. On the other hand, they may want to build good will and a fan base. It should be interesting.

 

[Shivansps]

Yup, that will have me looking at Zen 2. When doing x264 -> x264, it's damn fast. x265 -> x264 though and I pay the AVX penalty.

I had read that, and, if true, it would be unprecedented. Using an old board and chipset for a new CPU and also upgrading the old board's first slot to PCIe 4? That would be impressive and reminds me of the days when CPU upgrades made sense rather than platform upgrades.

I have my doubts mobo manufactures will go out of their way to do that though, as they will want to sell new boards. On the other hand, they may want to build good will and a fan base. It should be interesting.

The ironic part here is that an A320 and a old cheapo B350 probably supports PCI-E 4.0 because there is no extra stuff in there, expensive boards with switches and mux for extra stuff may not.

 

[Accord99]

The ironic part here is that an A320 and a old cheapo B350 probably supports PCI-E 4.0 because there is no extra stuff in there, expensive boards with switches and mux for extra stuff may not.

That actually happened with cheap P67 motherboards getting PCI-e 3.0 with Ivy Bridge processors.

https://www.tomshardware.co.uk/gigabyte-msi-pcie-3.0-gen3-third-gen,news-36482.html

 

[PotatoWithEarsOnSide]

If one mobo manufacturer provides a BIOS update for it, then I suspect that they all will.
However, my bet is that they'll limit BIOS updates to cheaper versions, knowing that consumers will likely want all the bells and whistles. Some will benefit, whilst most will need to stump up for a new mobo.
For the most part I don't think it'll matter, since there's nothing that can max out the bandwidth in that 1st PCIe slot anyway...at least not yet.

 

[DrMrLordX]

Interesting that AMD will be able to support PCIe4 on old x370 boards. Is Radeon VII even PCIe4.0 capable? I thought they disabled that.

 

[Topweasel]

About the reserved PCI-E lanes, Zen 1 die had 4 unused pcie lanes on AM4, if the Ryzen I/O die still has the 32 PCI-E lanes they may be able to run a IGP on it, if they are PCI-E 4.0.

Its a wiring issue at this point. It actually had 8 unused lanes a further 8 were turned off when using RR. 90% sure it wasn't about internal communication between CCX and iGPU it was for wiring to the video outputs. I don't know about downstepping PCIe 4.0 to PCI-e 3.0 to keep the 16x first slot. Or just go with 8x PCIe 4.0 in the slot. But then that means all Ryzen's lose one of it's new feature to support something that doesn't even apply that much, business PC's aren't even a dGPUless as people suggest and RR desktop chips fit that hole just fine. When I am ordering Dells if I need i5 or i7 selections it almost always involves at least some small dGPU. This would apply just the same to AMD's hierarchy.

 

[ozzy702]

Its a wiring issue at this point. It actually had 8 unused lanes a further 8 were turned off when using RR. 90% sure it wasn't about internal communication between CCX and iGPU it was for wiring to the video outputs. I don't know about downstepping PCIe 4.0 to PCI-e 3.0 to keep the 16x first slot. Or just go with 8x PCIe 4.0 in the slot. But then that means all Ryzen's lose one of it's new feature to support something that doesn't even apply that much, business PC's aren't even a dGPUless as people suggest and RR desktop chips fit that hole just fine. When I am ordering Dells if I need i5 or i7 selections it almost always involves at least some small dGPU. This would apply just the same to AMD's hierarchy.

We have around 15k desktops, maybe 500 of them have dgpus. Maybe 5% of the desktops I see in small and medium sized businesses while consulting have dgpus these days, they just aren't needed for the vast majority of use cases. Sorry, but if you want a piece of the volume business, you'll need to compete with Intel's offerings and offer an IGP/APU.

 

[Topweasel]

We have around 15k desktops, maybe 500 of them have dgpus. Maybe 5% of the desktops I see in small and medium sized businesses while consulting have dgpus these days, they just aren't needed for the vast majority of use cases. Sorry, but if you want a piece of the volume business, you'll need to compete with Intel's offerings and offer an IGP/APU.

All but 20 Desktops in my business use dGPU's. But that's because Almost all of them are I7's. That's the major point and why I stated RR takes care of it. The I have 2000 sales people doing sales stuff and other call center or day to day app stuff are going to be mostly on i3's and low level i5's. A selection RR is perfect for. From there no body is missing out on 8c or in the future 12c and 16c desktops because the company doesn't want to spend $20-$30 on a simple dGPU. If anything it's like my Ryzen Pro 1700 at work. The $30 extra no one batted an eye because it was much better than spending the extra $1200 and getting less memory to get an 8 core CPU.

 

[ozzy702]

All but 20 Desktops in my business use dGPU's. But that's because Almost all of them are I7's. That's the major point and why I stated RR takes care of it. The I have 2000 sales people doing sales stuff and other call center or day to day app stuff are going to be mostly on i3's and low level i5's. A selection RR is perfect for. From there no body is missing out on 8c or in the future 12c and 16c desktops because the company doesn't want to spend $20-$30 on a simple dGPU. If anything it's like my Ryzen Pro 1700 at work. The $30 extra no one batted an eye because it was much better than spending the extra $1200 and getting less memory to get an 8 core CPU.

Is RR being offered by larger vendors like HP and Dell to business clients? I haven't ordered anything in awhile so I haven't looked. I agree that RR is enough for the majority of users.

 

[beginner99]

Interesting that AMD will be able to support PCIe4 on old x370 boards. Is Radeon VII even PCIe4.0 capable? I thought they disabled that.

Good point. Why does it matter if there isn't a single consumer GPU with pcie 4.0? And navi will probably be slower than vega20 so no need for it either unless they can do something fun with ccix.

 

[Shivansps]

Its a wiring issue at this point. It actually had 8 unused lanes a further 8 were turned off when using RR. 90% sure it wasn't about internal communication between CCX and iGPU it was for wiring to the video outputs. I don't know about downstepping PCIe 4.0 to PCI-e 3.0 to keep the 16x first slot. Or just go with 8x PCIe 4.0 in the slot. But then that means all Ryzen's lose one of it's new feature to support something that doesn't even apply that much, business PC's aren't even a dGPUless as people suggest and RR desktop chips fit that hole just fine. When I am ordering Dells if I need i5 or i7 selections it almost always involves at least some small dGPU. This would apply just the same to AMD's hierarchy.

if they are going for a basic non-gaming oriented IGP they could just wire it to a PCI-E 1x and be done with it. Raven IGP is wired to a full x16 3.0, this is why they had to cut down the main PCI-E to x8, x16 is not needed on a non-gamning oriented IGP of the size of Vega 3-6, it can be done while still offering the full 24 lanes needed for AM4.

 

[Tuna-Fish]

Its a wiring issue at this point. It actually had 8 unused lanes a further 8 were turned off when using RR. 90% sure it wasn't about internal communication between CCX and iGPU it was for wiring to the video outputs

if they are going for a basic non-gaming oriented IGP they could just wire it to a PCI-E 1x and be done with it. Raven IGP is wired to a full x16 3.0, this is why they had to cut down the main PCI-E to x8, x16 is not needed on a non-gamning oriented IGP of the size of Vega 3-6, it can be done while still offering the full 24 lanes needed for AM4.

Not having enough PCI-E lanes on the CPU is not the issue. They are relatively tiny, they can fit more. The issue is pins on the socket. AM4 APU loses 8 PCI-E lanes not because those are used to connect to the GPU (that connection doesn't leave the chip), it's because the APU needs to get the display signal out some way, and they use the 32 data pins of PCI-E 8x for that. In principle, they could make an APU with 16x PCI-E for the GPU that would allow using the full width when the IGP was not in use. They have chosen not to do this.

Rumor is AM4 can't handle them though and we would need a new socket with more pins.

Pins on a socket are crazy expensive to AMD and Intel -- they add per-unit manufacturing cost they have to pay for every cpu that fits into that socket, whether they use it or not. AMD chose to have a larger consumer socket to be able to fit M.2 and some USB lines direct from the CPU, and a wider power delivery. I think it's unlikely that future mainstream sockets will be larger than AM4.

 

[scannall]

Is RR being offered by larger vendors like HP and Dell to business clients? I haven't ordered anything in awhile so I haven't looked. I agree that RR is enough for the majority of users.

Yes, both HP and Dell have a lot of models using various Ryzen APU's. Just go to their site and search Ryzen. Lots to choose from.

 

[dacostafilipe]

Even if nothing being changed I would still consider it a different arch just like I would consider Conroe and Nahelam to be different Arch. If you a fundamentally changing the the whole communication system even for packaging reasons. I think it becomes a different arch.

We don't have official data on how the chiplets are connected, but I doubt it's that different from Zeppelin : SDF on IOC, connected to the CCX using CCM.

If that's the case, the changes are more physical than architectural, so, IMO it's not a new architecture.

You do have a point, but let's just agree to disagree and call it a day

 

[Atari2600]

I'd have to check why the 8 lane drop but I thought I remembered it being wiring that is shared with the video outputs. The other are lost because they have no wiring at all because of the sockets support for AM4 and their decision to not go LGA. It isn't as simple as put a 1-3 unit iGPU for remeadial tasks, AM4 is packed a little tight for Zen's possible feature set and no desktop user wants AMD to lose more PCIe just so that people can pretend it's costing them business sales.

Fair 'nuff. Didn't realise it required losing a full block.

 

[stockolicious]

Fair enough. And a much easier solution they could offer to OEMs is a rebate on CPU + GPU combo from AMD. If you take both from us you get xx% off the normal price.

there is no doubt that AMD will try and push that advantage - It is a big advantage from a supply chain and cost perspective as both parts cost less than their competitors + additional rebate

 

[Jan Olšan]

if they are going for a basic non-gaming oriented IGP they could just wire it to a PCI-E 1x and be done with it. Raven IGP is wired to a full x16 3.0, this is why they had to cut down the main PCI-E to x8, x16 is not needed on a non-gamning oriented IGP of the size of Vega 3-6, it can be done while still offering the full 24 lanes needed for AM4.

I'm fairly sure that is not true, putting the iGPU on a PCIe bus logically woud have no advatages and would be inefficient. It is much more closely coupled than that.
The only reason the APU doesn't have a full ×16 GPU slot connectivity is that AMD didn't put full 32 lines worth of PHYs on the chip. It is only x8 to conserve die space.
Look at socket FM2+, Athlon X4 845 (Carrizo) already started that, despite Kaveri offering x16 before it and being APU.

Naturally, the display output lines conflicting with the x16 slot connectivity is also not true. I didn'T check but it is very unlikely, because you would need some switching hardware on motherboard for that - makes no sense really a mobo vendors would hate that because extra cost.