AMD VEGA confirmed for 2017 H1

fs123 · 7 Jan 2017 at 11:55

Wige said:
AMD's Raja Koduri "The P in PC should stand for performance"

Interesting that Raja says they are still figuring out the clockspeed limits, etc.

That means Doom could have been running on a low clocked Vega and may well be faster at launch. Potentially could rival Titan XP if that is the case.

Mauller · 7 Jan 2017 at 12:09

fs123 said:
Interesting that Raja says they are still figuring out the clockspeed limits, etc.

That means Doom could have been running on a low clocked Vega and may well be faster at launch. Potentially could rival Titan XP if that is the case.

it is an A0 engineering sample, most likely running conservative clocks. The demos are just to show it working considering the parts are only supposed to be first run, month old parts.

JediFragger · 7 Jan 2017 at 12:09

fs123 said:
Interesting that Raja says they are still figuring out the clockspeed limits, etc.

That means Doom could have been running on a low clocked Vega and may well be faster at launch. Potentially could rival Titan XP if that is the case.

It would certainly be in their interest to downplay the gameplay demo to throw nvidia a bit of a dummy

Doom112 · 7 Jan 2017 at 12:45

Mauller said:
it is an A0 engineering sample, most likely running conservative clocks. The demos are just to show it working considering the parts are only supposed to be first run, month old parts.

Polaris 11 was also a engineering sample, which was compared with GTX 950 ,however, the final product turned out worse than GTX 950 in term of performance and efficiency.

Rroff · 7 Jan 2017 at 12:50

Mauller said:
it is an A0 engineering sample, most likely running conservative clocks. The demos are just to show it working considering the parts are only supposed to be first run, month old parts.

It will be interesting to see how it turns out - early Polaris sampling was apparently quite positive compared to the actual volume production which was way down on what 14nm is really capable of.

D.P. · 7 Jan 2017 at 12:55

Mauller said:
it is an A0 engineering sample, most likely running conservative clocks. The demos are just to show it working considering the parts are only supposed to be first run, month old parts.

Likely the opposite, it will be one of the fastest clicked samples they have. As they get new samples in they will be fixing problems that cause failures and trying to work out a clock speed that can be produced in enough volume. That could be lower than the engineering samples they have now.

Nutella33 · 7 Jan 2017 at 13:13

Its already came out that they were using F3 stepping during the demo's but there's an F4 that is 3.6 boosting to 4ghz.

Just depends if that's the production run or not.

fs123 · 7 Jan 2017 at 13:13

D.P. said:
Likely the opposite, it will be one of the fastest clicked samples they have. As they get new samples in they will be fixing problems that cause failures and trying to work out a clock speed that can be produced in enough volume. That could be lower than the engineering samples they have now.

The first ZEN demo they showed ran at 3GHz base and now they have them running at over 3.4GHz/4GHz so if Vega samples are following the same path then expect faster clocks at launch compared to what they are showing now.

Of course it all depends on how confident they are with what they are showing. If they showed a golden sample that runs faster than what the median yield is, then they run the risk of the launch product performing slower than what they are showing.

I think they showed a sample that can beat a 1080 or match it but is conservative enough to not reveal anything major to Nvidia. One interesting observation is that the CES Vega demo appears to be the same performance as the leaks in November/December while the Zen demo has improved significantly.

Mauller · 7 Jan 2017 at 13:29

Nutella33 said:
Its already came out that they were using F3 stepping during the demo's but there's an F4 that is 3.6 boosting to 4ghz.

Just depends if that's the production run or not.

I was talking about Vega, not the summit ridge processors used with them.

D.P. · 7 Jan 2017 at 13:36

fs123 said:
The first ZEN demo they showed ran at 3GHz base and now they have them running at over 3.4GHz/4GHz so if Vega samples are following the same path then expect faster clocks at launch compared to what they are showing now.

Of course it all depends on how confident they are with what they are showing. If they showed a golden sample that runs faster than what the median yield is, then they run the risk of the launch product performing slower than what they are showing.

I think they showed a sample that can beat a 1080 or match it but is conservative enough to not reveal anything major to Nvidia. One interesting observation is that the CES Vega demo appears to be the same performance as the leaks in November/December while the Zen demo has improved significantly.

Look what happened to Polaris, AMD could get clocks and efficiency as good as their demo samples at CES.

Mauller · 7 Jan 2017 at 13:45

D.P. said:
Look what happened to Polaris, AMD could get clocks and efficiency as good as their demo samples at CES.

They never divulged information about polaris at CES2016 in relation to clocks. The demos they showed were efficiency and HDR demo's.

Mtom · 7 Jan 2017 at 13:53

fs123 said:
Interesting that Raja says they are still figuring out the clockspeed limits, etc.

That means Doom could have been running on a low clocked Vega and may well be faster at launch. Potentially could rival Titan XP if that is the case.

This is obvious with an ES chip, without proper cooling, etc.
That's what i wrote yesterday.

LoadsaMoney · 7 Jan 2017 at 13:58

It should beat the TXP, not rival it, as they are saying its better than VOLTA.

Baboonanza · 7 Jan 2017 at 14:01

If there is one thing I learned from the epic Polaris thread it's that all this speculation is pointless as none of us know didly.

JediFragger · 7 Jan 2017 at 14:07

LoadsaMoney said:
It should beat the TXP, not rival it, as they are saying its better than VOLTA.

Yeah, it's not like that statement will blow up in AMD's face....

opethdisciple · 7 Jan 2017 at 14:18

We know from PC's that footprint is more important than speed when it comes to RAM.

Mr Raja is saying that with Vega this will be different. (Of course this is GPU VRAM not RAM but maybe similarities)

Time will tell if this is actually a new reality or just marketing spiel.

LoadsaMoney · 7 Jan 2017 at 14:19

JediFragger said:
Yeah, it's not like that statement will blow up in AMD's face....

akarypid · 7 Jan 2017 at 14:20

Baboonanza said:
If there is one thing I learned from the epic Polaris thread it's that all this speculation is pointless as none of us know didly.

...but it's so FUN!

opethdisciple · 7 Jan 2017 at 14:25

I guess what I like about AMD is that they are more adoptive and supporting of open source standards etc... compared to Nvidia.

akarypid · 7 Jan 2017 at 14:36

AthlonXP1800 said:
AMD had showed off a preview of Vega architecture at CES 2017 so it seemed to me many of you all has short term memory loss completely forgetten or didn't recalled what AMD did at CES 2016? Well AMD was showed off a preview of Polaris architecture at CES 2016!

http://www.guru3d.com/articles-page...-january-2016-amd-polaris-architecture,1.html

Vega will expect to launch in June 2017, Polaris was launched in June 2016 and Fury X was launched in June 2015.

Don't expect lots of Vega 10 cards shipped at launch, HBM2 is very large chip than HBM1 and it still very difficult to mass produced with interposer, 1000 Fury X cards was shipped worldwide at launch and people cant get one weeks after launch then they went to bought 980 Ti instead. The same thing will happen to Vega 10 at launch too and people will buy 1080 Ti after learned RTG shipped so few Vega 10 cards worldwide worldwide.

All very true, but I think the availability issue you mentioned last is not going to be repeated.

Around May/June AMD was running on fumes (financially I mean). It was before they issued shares and called back bonds to restructure their debt. This made them very conservative with launch day quantities.

I believe that this year they will be much less conservative. Come May they should have lots more breathing room financially thanks to CPU sales and the 480 selling decently. If Vega is competitive (and they will now where it stands in the 1080 -> Titan range) then I'm pretty sure they won't hold back on manufacturing output.