Level1News: 2017-01-10 Tech on the HoRyzen & Other Tech News | Level One Techs

I think you're right when you say AMD is targeting the "middle of the road" for Ryzen. While they have been comparing their 8-core to the 6900K, I think it is just to demonstrate that they are competitive purely on CPU performance. I wouldn't say they are necessarily targeting the X99 platform as a whole. X99 motherboards are prohibitively expensive. Quad channel RAM and a bunch of PCIE lanes are nice, but really do drive up the cost of the motherboard. I think AMD is looking to combat that and keep motherboard price down to make themselves more competitive.

Perhaps from your perspective it is lacking but I think for a majority of AMD's target audience it is enough. IE Gamers/small scale content creators,"enthusiasts" ect... Very few run multi-GPU, 10Gig Ethernet is almost non-existent in the consumer space and while M.2 is becoming more popular, IDK if it is necessarily the end all be all. While there will be niche cases that do require more I think the platform is relatively balanced in terms of features to price which I expect to be lower than Intel.

Seems to me that AMD is trying to give more cores to the people and make more cores mainstream. Hopefully killing of dual cores for good.

As for workstations, servers, ect I'm fairly confident AMD will offer a higher-end more expensive platform with more connectivity for those that want it.

That being said I think it will have more than the 16+4 you described. The linked article refers to at least 24. Other's put it at more. Just a wait and see thing. As for the platform lasting 4 years I'm not so sure. From the few articles I was reading, AMD was stating that the Zen architecture will last for 4 years not necessarily the platform. We could see an AM4+ with more features later on

ibert · January 10, 2017, 9:56am

when you @wendell and @ryan talked about the loss off white color jobs i rememberd this news item i red just this morningt: http://www.rtvnoord.nl/nieuws/172480/Robots-maken-ook-de-witte-boorden-overbodig
is about a small company that has developed software to do business valuations without the need for a big (pricy) accountant firm. in the netherlands there are a limited amount of these firms and they have the marked cornered. Now small and medium company's can afford to do valuations to help them focus on their strenght instead off trying things out until they fail.(witch would have saved me)

On the subject of I.O.T. i can remember a episode of the tek where you @wendell mentioned a waffle iron with wifi or internet connection before : ) i think you are at least two years ahead of us.

alextulu · January 10, 2017, 9:59am

So, Alexa responded to what was being said on TV.

This is exactly why voice commands are stupid.

At the very least there should be a button, that you need to press before using voice commands.

alextulu · January 10, 2017, 10:07am

I agree.

I think that people who review hardware for a living (like Wendell) have forgotten the sacrifices, that most people have to make, when building a computer, because most people have a limited budget.

willrun4fun · January 10, 2017, 11:14am

@wendell that tie in the outro... I had to watch it again to decide it was real. Looked like post processing at first.

RGBLEDCPU · January 10, 2017, 11:23am

I remember seeing x4 NVMe in multiple articles about X370. Maybe it will depend on the motherboard. However, bottlenecking a Samsung 960 Pro (I plan to buy one) is a deal breaker for me.

Necron_99 · January 10, 2017, 1:11pm

I can't find the audio only option now.

wendell · January 10, 2017, 2:10pm

This is what I meant by right size. It's right size right now but in a few years it seems like not enough. Especially for the workstation class machines. Also there is 16 pcie (gfx) plus 4x pcie for all the USB and pcie.2 and other PCH features. And 4 more lanes just for SATA and nvme. However I wouldn't want to drop down to x2 nvme just for one SATA device. Which seems like it might be common for regular folks using the platform.. some more connectivity than kaby lake but less efficient if one SATA device is going to take up x2 worth of pcie bandwidth . But way less connectivity than low end x99

Etheoma · January 10, 2017, 2:58pm

Really Wendell have you not heard of AMD's Naples server class CPU's with upto 32 core, 64 PCIe lanes, quad channel memory, quad sockets etc.

And really most people don't need more than 16 PCIe lanes for there GPU 8x PCIe gen 2 is enough to not bottleneck a Titan XP 8x PCIe gen 3 should be more than enough.

And really 1 NVMe drive is enough for the vast majority, I think you have gotten confused with Intel putting a lot of the server stuff in there Enthusiast platform. What you should be using in a workstation is server grade stuff not half consumer half server.

What AMD is doing is what Intel did until the enthusiast platform you have your consumer chipset and your server chipset no enthusiast chipset.

Which really was only used to segregate the market so they could keep selling 4 core 8 thread CPU's to the consumer at a static price and have an excuse to charge even if you wanted more cores.

And with Vulkan and DX12 I think yes the time is right to be moving to 8 cores in the mainstream, just like with the i5 2500 was truly the time to move to quad core with Crysis showing fair quadcore scaling.

Now you have Doom and AotS showing fair octa core scaling and the time is right as said to be moving to octa core CPU's.

wendell · January 10, 2017, 3:09pm

This was, somewhat surprisingly, not part of CES given that AMD "demoed" or at least talked about them together last august.

I mostly agree, that's why I said "right size" -- but more in a sec.

Would disagree here -- If the target market isn't workstations, it's gamers/"enthusiasts" prettymuch everyone still has spinning rust for cheap bulk storage. It looks as though the NVMe would have to run at x2 if you want even a single sata device. Which is silly. In a year or two, maybe 2tb NVMe will be affordable.

So I don't know if you saw the demo system of their workstation cards. It looked like an Asus Z10PEDL motherboard rocking two Intel Xeon CPUs. That's what I'm concerned about. I would be really, really surprised if we saw naples at all in 2017. I just don't think it's that important to AMD.

Workstations are a special case anymore, apart from servers and desktops. Servers? AMD should do that hybrid arm thing. Arm servers etc. Workstations I think are starting to demand higher clocks first AND a lot of cores. So two 8 core cpus at higher clocks is probably more desirable than 32 cores lower clocked? I guess depends on your workload.

But it didn't seem as though Naples was even in a quasi demoable state. It also didn't seem as though Vega was ready for mass production, allthough maybe dangerously close to ready for mass production?

the lead time from these press releases (august, 2016) and "real product" concerns me a bit. And the 4 year lifetime with these limitations concerns me a bit. If you said X99 is a 4 year platform, it makes sense. When x99 first came out -- 40 PCIe lanes? PCIe3? Crazy! But then here we are a few short years later and SLI + 10 gigabit isn't that weird. Two NVMe in a workstation isn't that weird. And AMD set themselves up for this by comparing to X99.

So the flip side of this is for games and non-workstation workloads that 5ghz clock will be tough to beat. And intel can drop a 6 core cannon lake later this year. If it is still overclockable to 5ghz that may be a real problem for AMD. AMD needs a win, or at least a push, on single-core performance AND multithreading.

Bottom line: I think you're right, it's "right size" for right now. But in a year or two (and certainly 3) its going to feel ancient like 990FX does now.

Etheoma · January 10, 2017, 3:16pm

@wendell in the demo before the horizon event where they showed off Zen vs Broadwell E both at 3Ghz they showed off Naples and said that it should be arriving HL1 2017... which basically means Q2 2017 and there was a officially posted slide somewhere which slated it for HL1 2017 as well I think that might have been at the Horizon event.

And the consumer APU's are slated for release in Q3 2017, which I thought was weird as well you would expect the APU's to come before the server stuff but w.e.

wendell · January 10, 2017, 3:19pm

I will bet that apus come before server stuff. It seems unlikely. If the server stuff was that close to ready, it should enter mass production 90 days ahead.. that means the platform would be fialized or at least close to finalized? Enough for a workstation graphics card demo? Given they are working with OEMs and given that I saw workable intel 3467 or whatever boards last year at computex (rough, but working) I would think.. that tells me they are more than 6 months out which is August, at best, for Naples. I would love to be wrong about that though.

Etheoma · January 10, 2017, 3:24pm

@wendell I'll try and source that information... although it's gunna be hard as the Horizon event overshadowed the earlier event. But I know it wasn't just a leak or rumor.

Oh and there were benchmarks of Naples showing up in databases as early as September 2016.

Etheoma · January 10, 2017, 3:39pm

Here http://www.amd.com/en-us/press-releases/Pages/zen-processor-core-2016aug18.aspx it was showed off in that event

"AMD also conducted the first public demonstration of its upcoming 32-core, 64-thread “Zen”-based server processor, codenamed “Naples,” in a dual processor server running the Windows® Server operating system."

That was at the earlier event at Hotchip... Ah sorry I can't find the specific statment when they said that it was in HL1 but I am 99.8% sure it was an AMD statement not just a leak. And I really can't be asked to watched the Horizon Hotchips and GDC talks to find where it was.

And that was back in August and they already had it stable enough to run windows... ok not a high bar but considering that was 5 months ago not that bad. And also considering we have just under 6 months till the end of HL1 yeah it's totally possible that it will come out this year.

Azulath · January 10, 2017, 5:46pm

Could AMD have meant the AM4 socket and/or Zen architecture when talking about the 4y lifespan? Meaning that when Zen2 is released it is going to fit onto AM4 mainboards.

Etheoma · January 10, 2017, 6:19pm

Zen+ will fit into AM4 boards like Piledriver went with the AM3 boards.

Pholostan · January 10, 2017, 7:26pm

End of Q2 is what I've heard about server Zen stuffs. Probably about the same time as Vega is supposed to be around. If it isn't delayed, and we can wait for Vega some more...

Etheoma · January 10, 2017, 7:42pm

I don't think Vega will be delayed, they are already late really they can't hold off any longer on that, I can see Naples being delayed but not Vega it's too important on too many levels.

If Vega is delayed that does not only delay the consumer GPU's but the enterprise also, where as a Naples delay only delays Enterprise stuff.

But Naples and Vega launching simultaneously does make the most sense given that AMD will want system builders to use Naples and Vega together, so I think there aim at least is to launch them pretty close together, whether they manage it is a different matter.

Pholostan · January 10, 2017, 8:06pm

I really, really hope they don't get delayed at all. I hope for them to be early, to beat their roadmap.