• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Nvidia Blackwell in Q1-2025

Page 88 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
CastleBravo said:
Just be careful pairing it with an OC 14900k or you might exceed the limit of a standard NEMA 5-15 outlet and/or start tripping breakers if your monitor and some other stuff are on the same circuit.
Click to expand...
Who would pair a 5090 with that junk, 9800X3D any day and if it’s a workstation then it’s the 9950X
 
There are some benchmarks here:

It seems ~40% faster overall and is a pretty good step up, but the power and heat output is much higher too.
 
poke01 said:
okay scores for raster. meh for price, should have been N3E and had higher clocks. RT is impressive tho.

50% raster hould be minimum increase for a 90 class card. This is a total skip generation for 30/40 RTX users.


View attachment 115410
View attachment 115412
Click to expand...
20% higher raw performance using a 23% larger die with 33% higher TDP for 25% more cost (MSRP). No wonder why Jensen is strictly pushing AI-enhanced performance.
 
Pretty bad if this is accurate. 50 series could be Nvidia’s worse new gen in as long as I can remember with those numbers.

Really looking forward to temp and acoustics for the FE in the reviews tomorrow. 575W is crazy.
 
Blackwell paper launching has to mean a late ramp of something.
Either validation took longer than hoped, it needed an additional stepping, or GDDR7 ramp is behind schedule.
And if 5090 is really only a 20% bump in actual games. it is officially NV's first major perf miss since Fermi1.
8TPC/GPC screams like the culprit, NV has not changed the compute layering since Pascal, and the caching below L2 since Turing.
A Hopper style change is needed to improve scaling on very wide compute engines.

Nice synthetics perf, especially in Steel Nomad but this could be explained by membw or perfect compute scaling.
 
nvidia should skip straight to N2 with next gen. should give them a similar node jump to the one they got with Ada but would be so expensive though.
they can decrease the cost by putting only the GPCs on N2 and the rest on a cheaper node.
 
msj10 said:
nvidia should skip straight to N2 with next gen
Click to expand...
They'd want to use A16 with backside power delivery that should benefit GPUs, but this will be crazy expensive for even 2027, consumer stuff will get N3P if we lucky - next year, which will be essential to forget 50 series and wait 2 more years for N2 class to be come reasonably priced, if that ever happens.
 
branch_suggestion said:
Blackwell paper launching has to mean a late ramp of something.
Either validation took longer than hoped, it needed an additional stepping, or GDDR7 ramp is behind schedule.
Click to expand...
Previous rumors mentioned a delay in delivering the final BIOS for the cards, apparently AIBs received the firmware late in December. Whether that's the actual cause or just the last piece of the domino is unknown atm. What we know for sure is this delay now gets amplified by the holiday season in China and Taiwan.
 
Win2012R2 said:
Fermi did not cost as much!
Click to expand...
It is all relative.
Win2012R2 said:
Chinese New Year!
Click to expand...
Blackwell should've ramped well before CNY, should be good supply by now.
coercitiv said:
Previous rumors mentioned a delay in delivering the final BIOS for the cards, apparently AIBs received the firmware late in December. Whether that's the actual cause or just the last piece of the domino is unknown atm. What we know for sure is this delay now gets amplified by the holiday season in China and Taiwan.
Click to expand...
So NV isn't making many FEs then. Still board design and mfg takes more time than BIOS to sort out, usually cookie cutter stuff outside of AIB specific bloatware.
But yes, being this late is going to actually hurt them, assuming AMD has good volume and seemingly a good advantage in cost structure.
Should see a bit of mss correction this gen, just far less than if AMD actually made the 13 die wonder.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top