RX Vega M GH is a 24 CU / 1536 Shader part, but for some reason still has 64 ROPs.
http://www.eurogamer.net/articles/digitalfoundry-2018-core-i7-g-with-radeon-rx-vega-spec-analysis
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD CES 2018
- Thread starter shankly1985
- Start date
RX Vega M GH is a 24 CU / 1536 Shader part, but for some reason still has 64 ROPs.
http://www.eurogamer.net/articles/digitalfoundry-2018-core-i7-g-with-radeon-rx-vega-spec-analysis
Caporegime
- Joined
- 9 Nov 2009
- Posts
- 25,813
- Location
- Planet Earth
Hmm,it might explain why despite only having 1536 shaders Intel says the top SKU will be around GTX1060 Max-Q level performance,and why the cut down version,is more a GTX1050/GTX1050TI competitor.
Edit!!
I wonder if that highlights how ROP limited the AMD GPUs are - I wonder if the Vega shrink might increase the ROP to CU count?? Even Vega64 has 64 ROPS.
Spec seems strange. Both the 100 watt chips and two top end 65 watt use the exact same hardware and clock speed.
Is it just higher quality silicon that Intel feel is enough to produced a new part?
Can someone go show @shankly1985 the way here, he seems lost 
- Joined
- 25 Nov 2011
- Posts
- 20,680
- Location
- The KOP
Can someone go show @shankly1985 the way here, he seems lost
Lol Dont be a grass lad.
Was on phone I have no idea what happened. I was generally being serious I though Nvidia Green talk was in AMD CES thread. Silly me!"
Associate
- Joined
- 3 Feb 2006
- Posts
- 1,403
- Location
- Blackburn - Glastonbury - Portsmouth
in the vid he goes on about the nuc with 65w is not overclockable but the 100w chip on the nuc is, so does that need two chip types then ? one locked down and one not?
It's a new part, Quad Memory controllers for four stacks of HBM along with 1/2 DP support and possibility of an AMD equivalent of Tensor cores.
the 100w is the K version of the cpu...while the 65w is locked.
Mobile K parts seems strange. Not sure these will be fully multiplier unlocked.
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,335
- Location
- Billericay, UK
Good news for miners, bad news for gamers if AMD starts dedicating die space to their own FPGA cores as it's not like AMD has the money to develop two separate lines of products (one for consumers and professionals and one for HPC) like Nvidia can.
bits and pieces here and there
Zen+ coming soon on a 12nm process... then a phase 2 Threadripper set later on in the year
for 2019 we've got Zen2 coming on a 7nm process, then Zen3 in 2020
there's quite a bit of Vega stuff there as well being expanded in 2018... but no solid details on the desktop GPU's yet really
Zen+ coming soon on a 12nm process... then a phase 2 Threadripper set later on in the year
for 2019 we've got Zen2 coming on a 7nm process, then Zen3 in 2020
there's quite a bit of Vega stuff there as well being expanded in 2018... but no solid details on the desktop GPU's yet really
Are there any links to any of this? its almost as if its all a bit hush hush, someone at PR isn't doing their job well because no one even knows any of this is going on.
I'm subscribed to all of them and am not getting any notifications, usually i do about 3 minutes after they uploaded content.
Be nice if 12nm was the "real deal" for them - seems so many AMD products lately have suffered from either not being able to access the node they really needed or nodes that haven't quite hit the performance profile needed i.e. Fiji would have been a completely different story on TSMC 16nm and seemed pretty much made for sub 28nm and IMO GF 14nm has really held back Zen and Vega products which would work well on something like TSMC 12FF.