No one has said any different. The discussion is on profit margins and not volume
The large GP102 does that don't fully work end up as we Tie so the yields are not an issue.
-
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.
Poll: ** The AMD VEGA Thread **
- Thread starter TNA
- Start date
- Status
No one has said any different. The discussion is on profit margins and not volume
The large GP102 does that don't fully work end up as we Tie so the yields are not an issue.
Fair enough. But then AND are very clear the Vega won't have certified drivers so may not have been a problem.
yeah, so means anything from 484 - 500 mm^2
might not be fully certified, but i would think they will keep a focus on rendering accuracy. which means you still want a stable code base.
Well it does really well in the complex rendering like Medical and Energy tests, beating the Quadro P5000 there, and poor Titan Xp falling flat.
So they did clearly optimise as much as possible so far, without going the extra mile of specific certification and professional performance. They'll likely leave that for their Radeon Pro Vega card.
Which will make an interesting comparison for sure. Optimised vs Application Certified.
Which then goes back to the question of staff the FE is for. If it is supposedly not a gaming card but a pro card then it doesn't really do better than a Quadro P4000 which at least has certified drivers, can take correctly find and costs $800.
GM review telling people not to panic, his theory is the RX vega will be higher clocked, and priced lower than a gtx 1080 but will have 1080 performance, so as long as you ok with the extra power consumption the RX vega variant can be an alternative to a 1080.
As i mentioned before, when the new features are considered code stable, they get added along with the performance enhancements they bring. IT already does well in many areas, beating the P5000 as well, its just a matter of waiting and seeing how the vega RX driver compares. to see if the features really are enabled or not.
Yeah but it's not a perfect square, I haven't got the sharpest eye, but to me that looks like a rectangle, and I've just cut out the die on the PCper photo and it measured on my side 347x266, now yes they are saying 484 because Raja mentioned "a perfect square" and the closest rounded number that is below 500mm2 is 22x22, but he was talking about the package (die+hbm+interposer) and it is a perfect square at 47.3 x 47.3
i think he just means the area is nearest to the closes square number under 500mm^2, so 484mm^2. doesnt matter how you make that number, could be using a rectangle, as long as the area is near it etc.
Caporegime
- Joined
- 8 Jul 2003
- Posts
- 30,063
- Location
- In a house
But freesync is free, you are not buying into anything when you get one. It is there if you happen to have an AMD card, if not then you don't get that feature.
I have a Freesync monitor, I have no problems buying an Nvidia card if AMD fail to deliver with Vega. I do not feel locked in. But had I paid the huge premium with G-Sync, I would maybe feel locked it. Not that that is a bad thing these days with the lack of competition from AMD to be honest.
Double the TFLOPs, and 50% smaller process, 4x the VRAM that's faster than Fiji, and is 50% slower than TitanXp, and losing to its own previous engineering sample demos.
+1
This is my reaction also. But this is not a gaming card, so I am still not judging until the RX Vega cards come out.
Now if somebody has got the time to kill and the tools, there's a "simple" way that would allow us to be able to measure the die size pretty closely, by using these 2 photos :
FE : https://www.pcper.com/image/view/83430?return=node/68039
Fury Nitro : http://imgur.com/NTc1zCG
And scale down the FE to match the same size as the Nitro PCIE connection.
FE : https://www.pcper.com/image/view/83430?return=node/68039
Fury Nitro : http://imgur.com/NTc1zCG
And scale down the FE to match the same size as the Nitro PCIE connection.
Associate
- Joined
- 19 Jun 2017
- Posts
- 1,029
the statement can also imply the perfect square closest to 512... its not very clear.. the tweets go like this
WCCFguy: remember raja confirming a die size of 500 mm2.. [makes a reference to 512 mm2 as measured by pcper]
Raja: Yes I said so.. actually its the closest perfect square
so now its open to interpretation whether he is talking about the perfect square closest to 500 or 512...
in either case the packing density can be wildly different from rx 580.. given the power draw.. so would rather prefer to wait for official estimates on transistor count
Which imo we should have already had that info, why the hell are they hiding so much stuff ? I guess it generates buzz.
- Status