Nvidia's next gaming architecture rumoured to be called Lovelace

[HEXUS]

And the big GPU of the RTX 40 family could boast as many as 18,432 CUDA cores.

Read more.

[QuorTek]

how is it that AMD don't need so many cores and still be up there?

[Tabbykatze]

So the Samsung 5nm has me a little bit at odds, they failed their 7nm EUV and not sure they've even gotten it fully on track to being a fulfilled competitive node.

So to jump to 5nm Samsung? Hmm, i'm a little unsure.

[3dcandy]

So the Samsung 5nm has me a little bit at odds, they failed their 7nm EUV and not sure they've even gotten it fully on track to being a fulfilled competitive node.

So to jump to 5nm Samsung? Hmm, i'm a little unsure.

Don't fall into the trap of comparing nodes. Samsungs 8nm is roughly comparable to TSMC's 7nm and whilst the yields started off bad I'm hearing that they are improving quite well. Intel have failed to get their 10nm node working well so far so whilst I'm sure that TSMC are ahead Samsung appear to be on track if a little later than expected

[Zhaoman]

how is it that AMD don't need so many cores and still be up there?

Didn't Nvidia update their 'cores' to do 'up to' two instructions per clock so they're now counting them as double? I.e. the 10496 CUDA 'cores' in the 3090 is actually 5248 physical cores which is comparable to the 4352 in the 2080Ti. Nothing wrong with that but I think in terms of physical cores AMD and Nvidia are still comparable.

[CAT-THE-FIFTH]

Hopefully its more Ada Lovelace and not more Linda Lovelace!

[Tabbykatze]

Don't fall into the trap of comparing nodes. Samsungs 8nm is roughly comparable to TSMC's 7nm and whilst the yields started off bad I'm hearing that they are improving quite well. Intel have failed to get their 10nm node working well so far so whilst I'm sure that TSMC are ahead Samsung appear to be on track if a little later than expected

I am in no way comparing nodes, I am stating the inadequacy of Samsungs EUV 7nm to be suitable for Nvidias originally purported release being on a next gen node and instead settling back on an expensive, leaky, low yield, multi-DUV-patterning 10nm derivative. It is "roughly" comparable, but that's like saying a granny smith apple and a cooking apple "roughly" taste the same. What I'm getting at is whether 5nm will be ready considering we're still not really seeing much coming out around Samsungs 7nm offering and this Lovelace architecture will either be end of 2021 or H1 2022 meaning that ain't a lot of time to go through the motions. TSMCs 5nm also seems reasonably smaller and more dense than Samsungs 5nm. But as of yet, to affirm, although both TSMC and Samsung are in volume production of 5nm we only have real products from TSMC of 5nm (Apple silicon) whereas there is only a purported Exynos chip incoming from Samsung.

Intel was not mentioned, not interested in adding more slaps down than I already have to their multi-year potentially full decade of consistent failure of 10/7nm. Nor does it matter in this context, Nvidia wouldn't be able to purchase fab line from Intel for their GPUs which is the topic of this thread.

[3dcandy]

I am in no way comparing nodes, I am stating the inadequacy of Samsungs EUV 7nm to be suitable for Nvidias originally purported release being on a next gen node and instead settling back on an expensive, leaky, low yield, multi-DUV-patterning 10nm derivative. It is "roughly" comparable, but that's like saying a granny smith apple and a cooking apple "roughly" taste the same. What I'm getting at is whether 5nm will be ready considering we're still not really seeing much coming out around Samsungs 7nm offering and this Lovelace architecture will either be end of 2021 or H1 2022 meaning that ain't a lot of time to go through the motions. TSMCs 5nm also seems reasonably smaller and more dense than Samsungs 5nm. But as of yet, to affirm, although both TSMC and Samsung are in volume production of 5nm we only have real products from TSMC of 5nm (Apple silicon) whereas there is only a purported Exynos chip incoming from Samsung.

Intel was not mentioned, not interested in adding more slaps down than I already have to their multi-year potentially full decade of consistent failure of 10/7nm. Nor does it matter in this context, Nvidia wouldn't be able to purchase fab line from Intel for their GPUs which is the topic of this thread.

I did wonder - thought you wouldn't but hey.
Samsungs 7nm was deemed too risky by a fair few people for nvidia - but the problem they had (NV) is they screwed Samsung over prices....so kinda tough really. TSMC have no 7nm capacity anyhoo - and Apple have ALL the 5nm capacity at moment
The EUV side seems to be the issue. Crack it and yields are great, don't and they are pretty pants

Intel is not an issue now - but if reports are to be believed they are looking into getting extra fab capacity. Then they WILL be part of this thread, especially as it appears that they will want fab capacity for the Xe graphic cards

[Rubarb]

Hopefully its more Ada Lovelace and not more Linda Lovelace!

Yes hopefully the supply won't suck as hard

[fend_oblivion]

Wonder how AMD will respond to this

[ohmaheid]

Wonder how AMD will respond to this

Toulouse-Lautrec ??

[3dcandy]

Toulouse-Lautrec ??

naming a graphics card after a disabled alcoholic painter probably not the best idea to be fair

[ik9000]

Toulouse-Lautrec ??

naming a graphics card after a disabled alcoholic painter probably not the best idea to be fair

Van Gogh. You'll have to part with a body part to be able to get one...

[3dcandy]

Van Gogh. You'll have to part with a body part to be able to get one...

See much better! I'd have gone with Rene Magritte. This is not a graphics card

[Solex]

Makes you think which NVIDIA department came up with this name.

[scottismo]

So it was just me who thought of "Lovelace" the Penguin then...