Nvidia RTX 50 GB102 Blackwell early internal performance gains only in range of Pascal to Turing and not 2.6x from AD102, increased SM counts on the anvil

[Redaktion]

Nvidia's next gen RTX 50 Blackwell with purported 144 SMs isn't reaching ambitious performance gains in internal testing, according to latest rumors. Apparently, Blackwell's performance gains vis-à-vis AD102 are in the range of what we've seen from Pascal to Turing, but Nvidia wants more. This could imply that Nvidia might go for higher SMs counts with GB102 and possibly even a multi-chip module architecture similar to the purported GB100 datacenter GPU.

https://www.notebookcheck.net/Nvidia-RTX-50-GB102-Blackwell-early-internal-performance-gains-only-in-range-of-Pascal-to-Turing-and-not-2-6x-from-AD102-increased-SM-counts-on-the-anvil.716261.0.html

[Dude]

What is SM? It's good practice to include descriptions of all acronyms at their first use. Shader M
odule? Super Man? Surface Mount? Sticky Mud?

[Neenyah]

Streaming Multiprocessors. I agree with this though:

It's good practice to include descriptions of all acronyms at their first use.

[Light]

What is SM? It's good practice to include descriptions of all acronyms at their first use. Shader M
odule? Super Man? Surface Mount? Sticky Mud?

I'll agree to that.

[Mr Majestyk]

Ah always good to hear rumors about rumors possibly being false rumors. Anybody that believes Blackwell was going to be 2.6x faster than Lovelace I have a nice Bridge to sell you in Sydney Harbour.

Only another 18 months of fake news to go before Blackwell actually ships

[Anonymousgg]

I'm mostly interested to see if VRAM continues to creep up, for machine learning workloads.

Maybe we're due for another 8K gaming push.

[CZ]

Theres enough of unused Lovelace to make another much better version of each card save the 4090.