Zotac GeForce RTX 3070 Ti AMP Holo review
Gigabyte GeForce RTX 3070 Ti Gaming OC review
Palit GeForce RTX 3070 Ti GamingPRO review
GeForce RTX 3070 Ti founder edition review
MSI GeForce RTX 3070 Ti Suprim X review
Asus ROG Gladius III (wired and wireless) mouse review
GeForce RTX 3080 Ti founder edition review
Palit GeForce RTX 3080 Ti GameRock OC review
Gigabyte GeForce RTX 3080 Ti Gaming OC review
MSI GeForce RTX 3080 Ti SUPRIM X review
AMD Vega 10 and Vega 20 Slides Revisited
You can't say that AMD hasn't got little media attention the past few months. A set of slides surfaced at the usual suspect, this round a set with VEGA 10 and VEGA 20 slides.
These slides (which are not new btw) give some more insight on VEGA,. The VEGA information from then compared to now does seem to match up fairly closely. I do have to state though that the slides still do look slightly very odd, the logos used are dated and they do not match up with what is the style that AMD uses in their slides. Also the Radeon card pasted into the slides simply seems like a photoshop hack job ? Who knows though.
In the first slide you'll see that Vega 10 gets 64 NCU or Next-Gen Compute Units as AMD (x64 = 4096 shader processors). TDP is set at 225 W. The HBM2 memory comes in at 2 stacks at 16GB for a total of 521 GB/s, this value is off as we think it should be 8GB. Then again it might be a different SKU. Performance is rated at 12 TFLOPS for single precision and 750 GFLOPS for double precision. Also spot Vega 10 x2 which seems to be a dual chip solution coming in at 300W with likely lower clocks.
Vega 20 is mentioned in the enxt slide to be released second half of 2018. It is mentioned to get the same 64CUs as in Vega 10 and is fabbed at 7nm and features PCIe 4.0. TDP has a wide range of 150W to 300W, with I presume several products all at different core configurations and clocks using the same base die.
The last slide shows a server road-map for future products and mentions architecture codenames. As with previously released slides, Navi 10 will arrive in 2019.
Again here I have to say, the slides still look like a hackjob, but the Vega info does have some merit on them.
Preview: AMD VEGA GPU Architecture - 01/05/2017 04:46 PM
It is time to talk through a few mysteries from the AMD VEGA graphics processor architecture, join us in a preview architecture breakdown of what is to be expected with AMD VEGA graphics processors. R...
AMD Vega to likely get added to new Mac Pro - 12/08/2016 03:37 PM
Yesterday some info appeared in some forums, as it seems a beta version of MacOS Sierra 10.12.2 shows references to several new AMD video cards. Vega 10 is spotted inside the driver as well as Polari...
AMD Vega 10 - Vega 11 and Vega 20 GPUs Mentioned - 09/20/2016 03:24 PM
In an investors event AMDs CTO mentioned AMD Vega 10 - Vega 20 and Vega 11 GPUs. Not just that, some specs are shared as well....
AMD VEGA (Radeon RX 490) Launch Imminent ? - 08/15/2016 10:31 AM
As you all know, AMD is yet to release the enthusiast SKUs for their graphics card line. The Radeon RX 490 series will be based on a GPU called VEGA 10. We already know that AMD Radeon Technologies G...
AMD VEGA10 now scheduled for October launch - 05/11/2016 01:21 PM
The past few days a few rumors surfaced on the web, the GeForce GTX 1080 seems to be performing seriously good and as such AMD's board partners are in a bit of a fright. It seems only mainstream prod...
Evildead666
Senior Member
Posts: 1306
Joined: 2003-09-14
Senior Member
Posts: 1306
Joined: 2003-09-14
#5379167 Posted on: 01/07/2017 01:27 PM
It looks like Vega 10 is Gaming GPU (1/16 DP FP) and Vega 20 is Server GPU (1/2 DP FP).
They may have the same number of CU's, but they are different architecture.
Quite excited about the Dual GPU.
We're getting close to unified memory
I wonder if they'd still need a bridge chip?
It looks like Vega 10 is Gaming GPU (1/16 DP FP) and Vega 20 is Server GPU (1/2 DP FP).
They may have the same number of CU's, but they are different architecture.
Quite excited about the Dual GPU.
We're getting close to unified memory
I wonder if they'd still need a bridge chip?
Raplapla
Member
Posts: 30
Joined: 2016-12-16
Member
Posts: 30
Joined: 2016-12-16
#5379189 Posted on: 01/07/2017 04:25 PM
Power consumption is too high… At least I won’t be tempted to "upgrade" my GTX 1060.
Power consumption is too high… At least I won’t be tempted to "upgrade" my GTX 1060.
cowie
Senior Member
Posts: 13273
Joined: 2005-03-30
Senior Member
Posts: 13273
Joined: 2005-03-30
#5379195 Posted on: 01/07/2017 04:50 PM
I hope if they try to be the "best single card" with a Frankenstein x2 chip gpu it fails very very very hard.
I am sick of them setting back the advancement of r&d with x2 chip carss.
then they will sit and do nothing for what will seem like forever
I hope if they try to be the "best single card" with a Frankenstein x2 chip gpu it fails very very very hard.
I am sick of them setting back the advancement of r&d with x2 chip carss.
then they will sit and do nothing for what will seem like forever
Loophole35
Senior Member
Posts: 9793
Joined: 2011-09-21
Senior Member
Posts: 9793
Joined: 2011-09-21
#5379197 Posted on: 01/07/2017 05:10 PM
I wish we could make dual GPU cards illegal.
I wish we could make dual GPU cards illegal.
Click here to post a comment for this news story on the message forum.
Senior Member
Posts: 5642
Joined: 2003-09-15
Quote "Performance is rated at 12 TFLOPS for single precision and 750 GFLOPS for double precision".
It's SP performance does suit the enterprise market.