AMD Confirms Ryzen And Vega Launch Schedule, Developing 7nm Zen Products

Status
Not open for further replies.

3ogdy

Distinguished
So where is that confirmed launch schedule? We knew it was coming in March. The only thing new to me in this article is this section:
"Su referred to the next-generation products as "Zen 2" and "Zen 3." The company previously announced that it would provide a Zen+ variant as the successor to the original Zen core, but hasn't specified what lithography it will use for the "+" products. It's possible that Zen+ will be a tock-equivalent, while Zen 2 and 3 represent future shrinks. "
That is the entire article, the rest is just another Young & Restless episode - reiterating what we already knew over and over again. Is Tom's paying the writers per-article? How about getting paid for the actual content of the article, the substance?
 

spdragoo

Expert
Ambassador
"Lisa Su, AMD CEO, announced during the company's Q4 earning call that Ryzen CPUs would debut in early March, while Vega GPUs will come to market in Q2. AMD has a busy 2017 launch schedule that also includes Naples server chips in Q2 and the Zen core-based Raven Ridge APUs in the second half of the year."

Very first sentence in the article. Can't get much clearer to read or easier to find than that...& if you were expecting a "hour X on day Y" announcement, that's your problem.
 

WFang

Reputable
Dec 10, 2014
130
0
4,680
I had been hoping that the speculations that came from the wording of that Gamescon slide was true... But I guess with a debut in 'early March', it was not to be (unless they are pulling a quick one on us, but that seems dangerous to do in an investor meeting). Anyway, the rumor that it would be on the market immediately prior to, or during the February gamescon event is now dead, and I get to wait another week or two.

It also seems like they have walked back the statements from last year where they said to launch to enthusiasts first.. now apparently it will first be available to integrators, then oem's. I'm assuming 'integrators' mean e.g. Dell, Alien, etc. and OEM is 'anyone and us'??
 

InvalidError

Titan
Moderator
Intel's 10nm products are supposed to ship this year. If AMD competes against that on 7nm, that won't be before Intel has had a roughly year-long head-start.

In any case, what I'm most curious to know about Ryzen is pricing: if prices fail to provide a significant improvement in bang-per-buck, then most people will have little reason to consider AMD over Intel. Performance from six years ago (i7-2600) shouldn't still cost $250+ today. Especially in a physically IGP-less form as is supposed to be the case with the first wave of Ryzen parts.
 

TJ Hooker

Titan
Ambassador

Don't be silly, clearly no one was expecting a release hour. Hoping for a release day however is not unreasonable. We got a scheduled release date for Polaris about a month out, IIRC.
 

bentonsl_2010

Distinguished
Apr 12, 2010
68
0
18,630
I have liked intel cpu for a long time now but I hope to god AMD punches intel in the junk with this launch. I'm so sick of the BS pricing intel is getting away with nowadays with the crap performance improvements.
 

InvalidError

Titan
Moderator

Don't set yourself up for disappointment: Intel's performance per core is stagnating because there are practical limits to how much instruction-level parallelism can be extracted from software regardless of instruction set or hardware architecture. The only reason AMD could make such a large leap forward is because AMD was that far behind. Now that AMD caught up, don't expect AMD to make another similar leap forward again.

The only way performance will go up by a significant margin again is by writing well optimized multi-threaded software to leverage extra cores and threads in future CPUs.
 
Perhaps it was a poor choice of words, or perhaps it was words specifically designed to get us to examine and read the article. I had to go back and re-read it to catch the generalized release schedule. I missed it for all the extra information that came after it, especially when looking for a hard date and/or parts lineup with MSRPs.
 

_TheD0ct0r_

Commendable
Apr 19, 2016
114
0
1,710


I literally just said. CPUs in March, GPUs later in Q2
 
Hi en the way and statement made at the first demo in early june I worry it's going to be a paper launch. If not a paper launch worse ryzen vapor ware. A working demo nearly a year ago is getting a bit on my last nerve I guess.
 
InvalidError,
True. I keep trying to tell people Intel isn't dragging their feet. They are in fact REFINING the CPU.

The ONLY way now to make a notable jump in IPC is to break backwards compatibility with legacy code which is what is RUMORED they are doing. Even then, would we gain 20%? No idea.

The future is more about the SOFTWARE side of things in terms of performance. Such as VULKAN and OPENCL (CL not GL).
 
Okay how many think AMD will push the ZEN back into at least April raise your hands. Okay, 1,2,3,4,5,6,7,8,9,10, one sec I have to take my shoes off....

Seriously, I really hope this is a good stable launch, there are still many die hard AMD fans that have hated to buy Intel but had no choice seeing to the lack of a competitive AMD product. APUs... not the biggest hit with customers. If these are decent they will sell well even if they aren't as fast as the Intel's 2011v3 models.

 

nyannyan

Reputable
Jul 11, 2016
24
1
4,520
Honestly? I want Ryzen to be out as soon as possible to see what Intel does price-wise, because I'm basically sold on Intel for the simple reason that I don't want the uncertainty that is Windows 10 (and DX12 isn't really compelling since nothing I'm interested in uses only that, and Vulkan is more of an open standard). The newest I'm considering is Skylake. Also, wrt IPC, even if AMD can't do much better it seems they're doing it at a lower TDP.
 

bit_user

Polypheme
Ambassador
Looking at that slide, I hope AMD will settle for being #2 in "Visual Computing", as well. Even if Vega takes the crown at launch, it won't be long before Nvidia snatches it back. It's pretty appalling that they'll have left Pascal unanswered for an entire year.
 

bit_user

Polypheme
Ambassador
Not exactly.

Intel decided to keep the performance segment on 14 nm, for another generation. It's called Coffee Lake.

It's true that Cannon Lake will target the low-power segment @ 10 nm.
 

Xajel

Distinguished
Oct 22, 2006
167
8
18,685
I'm wondering how Ryzen will be configured in laptops.. I mean a high performance laptop will have a TDP of 35-45W.. can this still have 8c/16t ?

I know AMD will not make any APU with more than 4C/8T, so probably we wont see it anytime soon, at least in the first Ryzen based APU's, but maybe we might see it with 7nm.. Intel is already planning a 6 cores laptop CPU's later in 2018 which slightly align with AMD's 7nm plans (late 2018/ early 2019)
 

InvalidError

Titan
Moderator

With the amount of refinement that has gone into the instruction decoders over the past 20 years to keep x86 competitive against all other high-performance ISAs that have come and many gone in both performance and power efficiency, I would be surprised if switching ISAs at this point would yield much in the way of performance improvements that aren't achievable by further architecture and compiler tweaking.

Once you fully embrace multi-threaded software design though, there is no point in going to extreme lengths to extract the absolute maximum IPC per thread and that's why GPU, HPC and some server-oriented architectures use a larger number of much simpler cores with many more hardware threads instead to keep the cores' execution units busy without deep out-of-order execution, speculative execution and other expensive tricks. Making those simpler cores even simpler by ditching backwards compatibility, now that is something ditching ISA compatibility would be worth doing for.
 


TBH, It's the approach too.... Those who took the Intel approach back in the day (Z80, 808x) had to look forward to approximately four clock cycles per instruction. Motorola, with at least the 680x family (MOS's 65xx family by extension), if not the 68k family too, (I could be wrong on the 68k family,) was able to average one clock cycle... of course this was/is an average as there were instructions that did require more clock cycles to complete. (especially dealing with 16bit data on an 8bit data bus or CPU) So, a 4MHz i8086 would complete an average instruction at roughly the same time as a 1MHz MC6809.
 

MWP0004

Respectable
Oct 26, 2016
491
0
1,960
Could someone explain the difference in 14nm, 10nm, and 7 nm chips to me in simple terms? I see this often, and haven't found a simplistic explanation yet.
 

problematiq

Reputable
Dec 8, 2015
443
0
4,810


ELI5 I would say, it's the size that they draw the transistors for processors. It's not a 100% true but it's the general idea. The smaller you can draw things, the more you can fit in the same area. Also the less power you need. And power=heat.
 
Status
Not open for further replies.