• 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.

NVIDIA ‘Ampere’ 8nm Graphics Cards

humbug said:
They are the same GA102 Die, TSMC 7nm, i think the 3070 will be on Samsung
Click to expand...

Samsung also has a 7NM process,so it probably is Samsung 7NM. Samsung 5NM is also a variant of their 7NM process.

The RTX3080 only have 30W lower board power for much less shaders,and so much less VRAM,and also the lowish clockspeeds,is telling me yields might not be great on the GA102.
 
BTW,another thing is Samsung recently has invested in a ton of EUV equipment for their 7NM/5NM process nodes,and recently Qualcomm decided to not use Samsung. This means they must have a large customer,and its obvious that is Nvidia.

BigBANGtheory said:
ah ok that makes sense the silicon vs the actual processor being separate
Click to expand...

The GP107 dies in the GTX1050 and GTX1050TI used Samsung 14NM. Interestingly enough they were less efficient than the TSMC made GP106 and GP104. Polaris and Vega used a GF licensed variant of the Samsung node,so they no doubt lost some effiency on that node too.
 
CAT-THE-FIFTH said:
Samsung also has a 7NM process,so it probably is Samsung 7NM. Samsung 5NM is also a variant of their 7NM process.

The RTX3080 only have 30W lower board power for much less shaders,and so much less VRAM,and also the lowish clockspeeds,is telling me yields might not be great on the GA102.
Click to expand...

I thought Samsung was 8nm, not that it matters its all from their 10nm including 5nm and that's all from 12nm which was from 16nm.........
 
The trouble with FinFets, you can name them whatever the #### you like and if you don't give it a name that's as good or better than your competitors you're in trouble.

This is why you get Samsung's 5nm which is actually worse than TSMC's 7nm, remember that, its all ####### :)
 
humbug said:
I thought Samsung was 8nm, not that it matters its all from their 10nm including 5nm and that's all from 12nm which was from 16nm.........
Click to expand...

Its all rebranded as you say. Also as TSMC 7NM is yielding very well,having the RTX3070 cut down so much,whilst retaining most of the TGP of the full fat GA102 with 2.4X the amount of VRAM,would make the design look very poor. I suspect the RTX3090 will use the top tier dies,which you will find in the Quadros. I suspect its just a rebranded Quadro which some software blocks.
 
five8five said:
Isn't the biggest RDNA2 card rumoured to be 300W+?
Click to expand...

Hard to say,but AMD at a design level is behind Nvidia in performance/watt anyway. 7NM RDNA1 is around 12NM Turing level efficiency. So Turing shrunk down to a better node,would probably be better than AMD still. But if you look at the GA100 on TSMC,it also has very high board power,just like the GA102 apparently has.The TU102 board power is well under 300W in its full form with 24GB of GDDR6. The RTX3090 on a better node,has a 25% higher board power!
 
Ive added the new rumour to the specs of the existing Turing cards:

OK2YdfF.png


Ok the 3080 may have lost a Gb of ram against the 2080 Ti but the transfer speed of that ram is significantly higher and its clocked higher.

I think looking at that though, the 3070 will be around the level of the 2080 Super, maybe a bit higher. IF that is $500, then its spot on where I thought the new curve would land.
 
five8five said:
Isn't the biggest RDNA2 card rumoured to be 300W+?
Click to expand...

That assumption was made by people assuming Navi 2X was some sort of code for two times Navi.. IE 80 CU's and that it would be some 600 Watt monster if not for 50% PPW improvement, so 300 Watts.

The truth is Navi 10 is a 200 Watt GPU, X2 (80CU's) is 400 Watts.

But we know the XBox Series X is 175 Watts with 52 CU's and an 8 core 16 thread Zen 2 in it.

If it is 80 CU (i don't think it is) but if it is... Nvidia are in trouble.
 
8GB for the 3070 and 10GB for the 3080?

Laughable.

Chris Boardman will be getting my dosh this time around.
 
Kelt said:
8GB for the 3070 and 10GB for the 3080?

Laughable.

Chris Boardman will be getting my dosh this time around.
Click to expand...

8GB, it is pathetic... i said this in another thread,i don't need more GPU muscle necessarily, i have plenty of that, its my 8GB buffer that's holding me back.

A card that's faster than mine which has 8GB is frankly useless too me, as i said, i already have the GPU muscle i need, i can't turn the res up because when i do texture streaming spills over my 8GB buffer and chokes on my NVMe drive.
 
TheRealDeal said:
Cycling in the winter no thanks, would rather be sat next to a new Nvidia space heater.

If the rain is anything to go by up here yesterday you would be better with a sailing boat.
Click to expand...

I find it OK as long as it isn't raining (too much).

And I like riding in the dark, gives me an excuse to use some of the lights I've purchased over the years.

I'll be using my current MTB for winter use anyhow, and if I get this Boardman, that will be my summer ride.
 
humbug said:
That assumption was made by people assuming Navi 2X was some sort of code for two times Navi.. IE 80 CU's and that it would be some 600 Watt monster if not for 50% PPW improvement, so 300 Watts.

The truth is Navi 10 is a 200 Watt GPU, X2 (80CU's) is 400 Watts.

But we know the XBox Series X is 175 Watts with 52 CU's and an 8 core 16 thread Zen 2 in it.

If it is 80 CU (i don't think it is) but if it is... Nvidia are in trouble.
Click to expand...
The rumour I'm talking about is it hasn't scaled very well at top end requiring a lot more power.
 
You must log in or register to reply here.
Back
Top Bottom