• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Vega countdown has appeared

CSI_PC said:
And as I said the Tegra releases much later than the dGPU it is based upon, along with commitments to HPC projects that require the hardware to start to be implented in 2017.
That said, I am amazed your downplaying the Q3-Q4 Xavier sampling status in the hands of manufacturers, this means the design must not only be taped out but in production albeit more limited for that solution Q3-Q4 and now you want to say it means production is 2018.
Anyway ask yourself this, how long has it taken the Pascal Drive PX2 to be truly available compared to all of the Pascal lineup (safe to say nearly all Pascal has launched-available before full production of the Pascal Drive PX2).
The 10nm was always a rumour started by only a few sources that went into overdrive, remember Nvidia introduced Pascal as a stepping point to Volta.

Do you really think Nvidia and IBM managed to finalise with adjusted contractual spec performance (this update was November) based upon thin air, or that Nvidia has just now released news in their CES opening speech the Volta Tegra now 30 TOPS DL at 30W rather than their original aim of 20 TOPS DL at 20W?
Maybe we should take that to Nvidia thread, but I think some are in for a shock if they really think Volta is next year.
Reminds me how even just before launch of Pascal sites were adamant it did not exist or had massive problems.
The strategy will be a re-run of Pascal, why do you think they went straight for the P100 and largest die as the 1st Pascal GPU.
1st Volta will be the V100 for HPC and like P100 for only core clients and much later as an individual GPU, btw we were offered a full DGX-1 type solution months ago and ready for delivery 1 week later, but the perception is that P100 is not in production because you cannot buy the card individually or at more general Nvidia associated stores.
And probably like before followed promptly by GV104 in the consumer space - that said I would not expect a consumer GV104 until late Q3 at the earliest.

Cheers
Click to expand...

Sampling to manufacturers, then sure. As far as I was aware it was only demonstrated last year, not sampled outside of Nvidia.

I said 2018 in regards to 10nm if they were testing already last year.

Interesting to see the spec changing though with Volta Tegra.
I remember the DGX was available for quite some time before Paxwell dropped. Very interesting regarding P100 availability and quite telling indeed.. if not the most telling news of all. Almost definitely this year then and probably closer than later..

So really then, we don't even know what the 10nm chip is called yet and it's pretty irrelevant with Volta on the cards coming so soon. dGPU remains to be seen though, I have a feeling Vega will determine this.

Cheers!
 
N4CR said:
Sampling to manufacturers, then sure. As far as I was aware it was only demonstrated last year, not sampled outside of Nvidia.

I said 2018 in regards to 10nm if they were testing already last year.

Interesting to see the spec changing though with Volta Tegra.
I remember the DGX was available for quite some time before Paxwell dropped. Very interesting regarding P100 availability and quite telling indeed.. if not the most telling news of all. Almost definitely this year then and probably closer than later..

So really then, we don't even know what the 10nm chip is called yet and it's pretty irrelevant with Volta on the cards coming so soon. dGPU remains to be seen though, I have a feeling Vega will determine this.

Cheers!
Click to expand...
Yeah.
I added an edit showing they confirmed Volta Xavier as 16nm at GTC Europe.
You are right that Nvidia is sort of in-between the cycle for node-fabrication size with their launch cycle, but then that is also applicable to AMD (Navi would be the 1st to be a node shrink for them after 14nm) and so takes some pressure off Nvidia.
I could see the Xavier eventually going from 16nm to 10nm as that is where the most pressure is for Nvidia, that is a strategy Nvidia has done in the past where the Tegra Maxwell was 20nm compared to the dGPU at 28nm.
Will be interesting to see if AMD has to consider 10nm for their semi-custom embedded/small-mobile solutions as well.
Cheers
 
I was curious how much longer but ve.ga doesn't appear to load...

Not sure if I really care about a "preview" since AMD has been misleading 100% of the time on performance benchmarks in recent years. If this is a new arch then core count and MHz probably doesn't mean anything.

I need a countdown to the [H] review.
 
CSI_PC said:
Yeah.
I added an edit showing they confirmed Volta Xavier as 16nm at GTC Europe.
You are right that Nvidia is sort of in-between the cycle for node-fabrication size with their launch cycle, but then that is also applicable to AMD (Navi would be the 1st to be a node shrink for them after 14nm) and so takes some pressure off Nvidia.
I could see the Xavier eventually going from 16nm to 10nm as that is where the most pressure is for Nvidia, that is a strategy Nvidia has done in the past where the Tegra Maxwell was 20nm compared to the dGPU at 28nm.
Will be interesting to see if AMD has to consider 10nm for their semi-custom embedded/small-mobile solutions as well.
Cheers
Click to expand...
According to rumors Vega 20 was moving to 10nm. If that's the case it would probably be the first AMD design on 10nm as opposed to Navi. Risk production likely less of an issue there for higher performing APU products in the server market.
 
N4CR said:
The FUD is that they're taped out. There would be more news about it if so, if not very shortly. It would have to also be soonish for dGPU if it's any time this year. I wouldn't bet Volta is not this year if they already demonstrated it in September publicly. dGPU volta? Who knows. It was a year before we saw anything from paxwell in public.
Just Volta is not what it was supposed to be (10nm) which is what I thought was still the plan. Thing is we have not seen any public confirmation that dGPU Volta is 16nm. We may see a split where mobile is OEM only 'volta 16nm' and dGPU is a later 10nm, this has happened in the industry before. I doubt it, but it's possible.

Sampling Q3-4 means on market Q4 earliest, more likely Q1.

If they were already testing and sampling 10nm last year in the same time, then that can't be so far behind either, 2018...? And in what?
Click to expand...


Volta is going to be on 16nm that is a given, if their SOC Volta is planned for this, the rest of their chips are too.

PS there was no news on Pascal tape out (other than one guy at B3D, and I stated its was likely taped out a while back because of certain ways nV employees were talking about (at least for me), when AMD stated they had months of lead, and the rumors started about Pascal being late , not only that, it really fooled everyone when nV announced Pascal and when they launched Pascal. nV isn't like AMD who runs their mouth off, six or more months prior to launch and then gets the short end of the stick.
 
Last edited:
Anarchist4000 said:
According to rumors Vega 20 was moving to 10nm. If that's the case it would probably be the first AMD design on 10nm as opposed to Navi. Risk production likely less of an issue there for higher performing APU products in the server market.
Click to expand...


Vega 20 was never moved to 10nm, it was always planned for 10nm, since the first time the rumors of it being there at least.

Risk production is based on complexity and die size, and higher performing parts usually are greater on both.......

If the node is still immature that is when ya see small chips done first, like a low end GPU. CPU+ APU are a bit more complex and usually a bit bigger then the value GPU's
 
Last edited:
razor1 said:
Vega 20o was never moved to 10nm, it was always planned for 10nm, since the first time the rumors of it being there at least.

Risk production is based on complexity and die size, and higher performing parts usually are greater on both.......

If the node is still immature that is when ya see small chips done first, like a low end GPU. CPU+ APU are a bit more complex and usually a bit bigger then the value GPU's
Click to expand...
Not a problem for a high performance server APU with a potentially $1k+ price tag. Being separate dice they may or may not be that large. Two small ones probably work in that APU market where sharing between them isn't as relevant.
 
Anarchist4000 said:
Not a problem for a high performance server APU with a potentially $1k+ price tag. Being separate dice they may or may not be that large. Two small ones probably work in that APU market where sharing between them isn't as relevant.
Click to expand...


Still is a problem, because of the complexity the likely hood of even getting one good part because of node immaturity is even higher. If a wafer costs 6-8k which is about what a new 10nm wafer would be in risk production. How much do you think 1k+ parts are going to be viable?
 
and the ve.ga site is down, anyone have another link for their CES presentation?
 
razor1 said:
Still is a problem, because of the complexity the likely hood of even getting one good part because of node immaturity is even higher. If a wafer costs 6-8k which is about what a new 10nm wafer would be in risk production. How much do you think 1k+ parts are going to be viable?
Click to expand...
With the shrink a ~200mm2 GPU die may be reasonable. That's not what I would call large, especially with their apparent ability to reroute their mesh around certain defects or clock CUs independently of each other.
 
Anarchist4000 said:
With the shrink a ~200mm2 GPU die may be reasonable. That's not what I would call large, especially with their apparent ability to reroute their mesh around certain defects or clock CUs independently of each other.
Click to expand...

And what about the CPU portion of the APU?

You end up with around 300nm with both right? when have we seen risk production with midrange GPU's which are around that size?

Hmm, never, usually they are around 150nm or less, the smallest of the a gen.....
 
well they should have a link to the architecture preview, which is what I'm more interested in :) that will give me a lot more insight on Vega's performance, the slide desks, most of the things they have talked about, Polaris already has at least for gaming concerns, Vega has some other things that will help in HPC type situations over Polaris though.
 
Something that wasn't revealed from slides (or misdirection on the slides) is that the taped out chip was paired with 2 HBM stacks and not 4 even though the mockups on the slides were 4 stacks.
 
yeah even their architecture preview, gives very little, hopefully in another 7 hours or so will be more.

Archiecture preview, there has been some change over Polaris that "might" give it better throughput but its really iffy and if it does not by much.
 
what did I say about Vega going to be larger than Pascal GP 102 die size?

Its die size is in the neighborhood of 520 – 530 mm²

Expect the ratio of fixed function units to be higher than Fiji.

2-1080.3975960926.jpg
 
Guru3d said enterprise first. You guys will be waiting a long time for the consumer version. By the time it comes out we'll be in Volta territory. I think AMD graphics is officially in AMD vs. Intel processor territory.
 
lolfail9001 said:
Sorry, are you blind? Pictures visibly show the ES of Vega in the sea of 1080s. Also, AotS bench is 12-01, it is like a month old.
Click to expand...

It was a Vega ES faster than a stock 1080 back then.. it's nearly 3/4 of a year old.. this is the point I'm trying to make. Thank you. It will depend on the title but it seems it should be ~1080 speed or better depending. Maybe not closer to Pitan but around that spread as drivers mature and developers come onboard.
 
TaintedSquirrel said:
No one cares about Doom anymore. The fact that they're focusing on it at this point just makes me think they're using it (Vulkan) as a crutch.
Click to expand...

AMD is stuck in a time where iD Software engines were all people benchmarked on. The landscape is no longer the same.
 
Anyone know if Zen is on the stairs, floor, or drumstick in the video here?

 
razor1 said:
what did I say about Vega going to be larger than Pascal GP 102 die size?

Its die size is in the neighborhood of 520 – 530 mm²

Expect the ratio of fixed function units to be higher than Fiji.
Click to expand...
Yep.
And fits with exactly what I was saying a little while ago that Vega would be a large die and a cut core model would be for consumer and the full core for HPC, which IMO is big Vega and as I mentioned at the time same as what Nvidia did with GP102.
Also within the size some of us (including you) discussed.

Cheers
 
You must log in or register to reply here.
Back
Top