Wednesday, January 9th 2019

AMD Announces the Radeon VII Graphics Card: Beats GeForce RTX 2080

AMD today announced the Radeon VII (Radeon Seven) graphics card, implementing the world's first GPU built on the 7 nm silicon fabrication process. Based on the 7 nm "Vega 20" silicon with 60 compute units (3,840 stream processors), and a 4096-bit HBM2 memory interface, the chip leverages 7 nm to dial up engine clock speeds to unprecedented levels (above 1.80 GHz possibly). CEO Lisa Su states that the Radeon VII performs competitively with NVIDIA's GeForce RTX 2080 graphics card. The card features a gamer-friendly triple-fan cooling solution with a design focus on low noise. AMD is using 16 GB of 4096-bit HBM2 memory. Available from February 7th, the Radeon VII will be priced at USD $699.

Update: We went hands on with the Radeon VII card at CES.
Add your own comment

153 Comments on AMD Announces the Radeon VII Graphics Card: Beats GeForce RTX 2080

#102
INSTG8R
My Custom Title
cucker tarlson said:
which gives them 16nm gen nvidia performance in enthusiast (vega 7 vs 1080ti) and probably nvidia's 12nm gen performance with 7nm navi if it can match the 2060/2070.
Equating process node to performance is a fools game.. Turing is HUGE! Makes my Vega look small...

Edit: I guess me equating die size is no better but point still stands.
Posted on Reply
#103
cucker tarlson
INSTG8R said:
Equating process node to performance is a fools game.. Turing is HUGE! Makes my Vega look small...

Edit: I guess me equating die size is no better but point still stands.
yes but turing has RT and tensor cores which vega 1/2 has none.
Posted on Reply
#104
INSTG8R
My Custom Title
cucker tarlson said:
yes but turing has RT and tensor cores which vega 1/2 has none.
No but it’s got lots of Compute units that are asleep most of the time...so are your tensor cores ;)
Posted on Reply
#105
prtskg
Excerpt from my fantasy mill -
TSMC - Lisa, our 7nm isn't running at full capacity. We'll give you a rebate if you want more production.
Lisa - Mark, cpu division is in testing phase so they won't require more 7nm capacity, right?
Mark - Yes Lisa. And Navi is too far out for more 7nm capacity.
Lisa - How about we bring Vega 20 to consumers and increase it's production? Even though the smallest Vega 20 has 16GB HBM, selling it won't be difficult seeing RTX prices.
Mark - Right you are Lisa. I don't know what AMD would have done without you. :respect:
Lisa - :rockout:
Mark - At CES announcement do say that you love gamers and more gamers will buy it.
Lisa - Really?
Mark - Yes :cool:
Posted on Reply
#106
bajs11
meh its just another Vega, high power consumption and expensive
and I was hoping for a cheaper alternative to the 2080...

why hbm though I mean it didn't give Vega any advantage over Pascal cards and is super expensive
Posted on Reply
#107
prtskg
bajs11 said:
meh its just another Vega, high power consumption and expensive
and I was hoping for a cheaper alternative to the 2080...

why hbm though I mean it didn't give Vega any advantage over Pascal cards and is super expensive
It's Radeon Instint 50 cards with fans. So AMD brought a professional card to consumer market with minimum investment (fans :p) to compete with 2080. Something is better than nothing, right?
Posted on Reply
#108
Sandbo
prtskg said:
It's Radeon Instint 50 cards with fans. So AMD brought a professional card to consumer market with minimum investment (fans :p) to compete with 2080. Something is better than nothing, right?
Exactly what I think, if they are able to sell lower bin chips at a price similar to a competing product, why not?
As the users this may not be a wonderful new product, but it is at least an option.
Posted on Reply
#111
FordGT90Concept
"I go fast!1!11!1!"
Vega 20 is 7nm where RTX 2080 is 12nm (rebranded 16nm). Chip size doesn't really tell us anything useful because of that.
Posted on Reply
#112
newtekie1
Semi-Retired Folder
16GB on this card makes no sense, make it cheaper and use 4GB stacks instead.

the54thvoid said:
Also, to point out the obvious. Isn't this AMD's most expensive consumer gfx card ever?
Not even close to the most expensive. The 295x2 was $1500.
Posted on Reply
#113
INSTG8R
My Custom Title
newtekie1 said:
16GB on this card makes no sense, make it cheaper and use 4GB stacks instead.



Not even close to the most expensive. The 295x2 was $1500.
Not my quote but let’s just clear that up shall we?

`Unfortunately, you can't scale down the HBM2 any further and still retain the 128 ROPs, so 16 GB is the smallest capacity AMD can offer, which is why the pricepoint on this is so close relative to the 2080.
`
Posted on Reply
#114
cucker tarlson
medi01 said:
331 mm² Vega 7 vs 545 mm² 2080.
It ain't that bad, if you think about it.


I recall 1060 was "much faster" than 480.
what ain't bad ?
a 550mm2 12nm card from nvidia runs at 230W with RT features, a 7nm 330mm2 Vega can barely match it at 300W.
The card is decent if you look at the performance only. technologically a 7nm 300W card with 1080Ti performance is taking amd another step behind nvidia.They chasing big pascals in efficiency and can barely match it at 7nm and with 4 stack hbm2.They'll be chasing 2080Ti in 2020 when nvidia drops 7nm cards with a power efficiency improvement similar to maxwell-pascal.
Posted on Reply
#115
INSTG8R
My Custom Title
cucker tarlson said:
what ain't bad ?
a 550mm2 12nm card from nvidia runs at 230W with RT features, a 7nm 330mm2 Vega can barely match it at 300W.
The card is decent if you look at the performance only. technologically a 7nm 300W card with 1080Ti performance is taking amd another step behind nvidia.They chasing big pascals in efficiency and can barely match it at 7nm and with 4 stack hbm2.They'll be chasing 2080Ti in 2020 when nvidia drops 7nm cards with a power efficiency improvement similar to maxwell-pascal.
Sorry and your getting your Vega power numbers from where? I can make my Vega run at 240W(that’s actually stock no adjustments) if I want to and lose zero performance doing it but do tell me more about your 300W assumptions...
Posted on Reply
#116
cucker tarlson
INSTG8R said:
Sorry and your getting your Vega power numbers from where? I can make my Vega run at 240W if I want to and lose zero performance doing it but do tell me more about your 300W assumptions...
on power save bios,where you lose performance too
sorry,but anyone can do that and it has nothing to do with what were talking about.I can run my 1080Ti at 70% power limit too and lose very little performance,so can every other amd/nvidia owner with their card. as if vega was anything special at undervolting :rolleyes: I could run my gtx 1080 with fans off when I undervolted,it never broke 60 degrees and it still got 90-95% of the stock performance.
I'm quoting every review of v64 ever made,only the ones that are conducted in objective manner,stock bios vs stock bios.
Are we going to debate that v64 is a power hog now ? :) I thought it was common knowledge since it got out.
Posted on Reply
#117
siluro818
Rowsol said:
HBM is expensive and 16GB is overkill. I would rather see it ship with 8GB and a lower price. Can someone point me to a benchmark of a game using more than 8GB of vram?
Assassin's Creed Odyssey uses 9+ GB VRAM with 4K/HDR/Ultra settings when utilizing Adaptive AA mode. Though on Vega 56/64 you can just enable HBCC and it works like a charm.
Posted on Reply
#118
Assimilator
INSTG8R said:
Not my quote but let’s just clear that up shall we?

`Unfortunately, you can't scale down the HBM2 any further and still retain the 128 ROPs, so 16 GB is the smallest capacity AMD can offer, which is why the pricepoint on this is so close relative to the 2080.
`
Very valid point that myself and others have overlooked - I imagine it was cheaper to just throw 16GB of expensive HBM2 on a stopgap card that probably won't be available in large numbers, than to redesign Vega 20 to have a different ROP:VRAM ratio.
Posted on Reply
#120
INSTG8R
My Custom Title
cucker tarlson said:
on power save bios,where you lose performance too
sorry,but anyone can do that and it has nothing to do with what were talking about.I can run my 1080Ti at 70% power limit too and lose very little performance,so can every other amd/nvidia owner with their card. as if vega was anything special at undervolting :rolleyes: I could run my gtx 1080 with fans off when I undervolted,it never broke 60 degrees and it still got 90-95% of the stock performance.
I'm quoting every review of v64 ever made,only the ones that are conducted in objective manner,stock bios vs stock bios.
Are we going to debate that v64 is a power hog now ? :) I thought it was common knowledge since it got out.
No I own one so I know exactly how much power it uses. I don’t go by outdated numbers you’re using and I’ve tested both BIOS many times and the performance difference is non existent. Oh and the “HOT” BIOS tops out a 276W so again your 300W is outdated. I’ve never even made ANY attempt at undervolting.Yeah I can easily push it over 300(Ran the “hot” BIOS this morning with the clocks at 1750/1000 50% power limit and hit 347W peak so nobodies saying it can’t eat up power. But please drop your 300W assumptions they’re wrong and your opinion not fact at this point.
Posted on Reply
#121
newtekie1
Semi-Retired Folder
INSTG8R said:
Not my quote but let’s just clear that up shall we?

`Unfortunately, you can't scale down the HBM2 any further and still retain the 128 ROPs, so 16 GB is the smallest capacity AMD can offer, which is why the pricepoint on this is so close relative to the 2080.
`
True, but do they need 128 ROPs in a desktop graphics card on this performance level?

Assimilator said:
Very valid point that myself and others have overlooked - I imagine it was cheaper to just throw 16GB of expensive HBM2 on a stopgap card that probably won't be available in large numbers, than to redesign Vega 20 to have a different ROP:VRAM ratio.
A redesign isn't really necessary, just remove two of the HBM2 stacks from the GPU and disable two of the memory controllers(which would also disable half of the ROPs if I'm not mistaken).
Posted on Reply
#122
INSTG8R
My Custom Title
newtekie1 said:
True, but do they need 128 ROPs in a desktop graphics card on this performance level?



A redesign isn't really necessary, just remove two of the HBM2 stacks from the GPU and disable two of the memory controllers(which would also disable half of the ROPs if I'm not mistaken).
Sadly the design is the design so their hands are tied outside of your proposed scenario that who knows they might even do that, it’s not an unreasonable idea.
Posted on Reply
#123
R0H1T
Wasn't less ROPs supposedly one of the biggest drawback of Fury & Vega? If you cut the ROPs in half then what would happen to the performance, surely none wants Vega64 respun?
Posted on Reply
#124
Gasaraki
MrGenius said:
2x the VRAM. Don't forget that. So you're getting A LOT more for your money actually.
R0H1T said:
Two 8 pins doesn't automatically mean 300W TDP, I remember the Polaris meltdown when the cards were pulling above 75W from the PCIe slot. This could just be to avoid that same brouhaha.
This a a Vega 20 card. The Radeon instinct card was already at 300W so... this has higher clocks...
Posted on Reply
#125
moproblems99
cucker tarlson said:
the problem will be 2020 when 7nm nvidia will have NO competition. NONE.
Since you seem to know the future, can we get together and discuss the lottery? If you aren't quite that talented in the future, can we start at the Super Bowl then? We can discuss how to split the profits.

Gasaraki said:
This a a Vega 20 card. The Radeon instinct card was already at 300W so... this has higher clocks...
Also on a smaller node, no?
Posted on Reply
Add your own comment