Discussion Nvidia Blackwell in Q1-2025

Page 110 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Heartbreaker

Diamond Member
Apr 3, 2006
4,653
6,108
136
Daniel Owen makes the case that the 5080 is really a 5070 Ti tier class of GPU. Because if you look at generational improvements, it's the worst x80 series card ever:
View attachment 115890

While I agree this is a total garbage release, nothing quite says pathetic nerd, as loudly as making spread sheet to "prove" the name of the GPU is wrong.
 

adroc_thurston

Diamond Member
Jul 2, 2023
4,714
6,501
96
Same process node this time, right?
Maxwell was also 28nm HKMG.
So imagine Nvidia trying to release the GTX 500 series on the same node as the GTX 400 series.
well they did, both were 40nm.
The node shrink literally was the GTX 500 series
500 was a Fermi refresh that fixed every 40nm PDK goofup Fermi actual had.
Without it, you'd have the same garbage we see now with the 5000 series
There was no shrink, but it was also really incremental and launched like 9mo after the Fermi actual.
 

moonbogg

Lifer
Jan 8, 2011
10,667
3,197
136
While I agree this is a total garbage release, nothing quite says pathetic nerd, as loudly as making spread sheet to "prove" the name of the GPU is wrong.

He included prices too, which is what matters more. Nvidia could have offered more performance in that 80 card with so much room between it and the 5090. They just said "NAH".
 

moonbogg

Lifer
Jan 8, 2011
10,667
3,197
136
Maxwell was also 28nm HKMG.

well they did, both were 40nm.

500 was a Fermi refresh that fixed every 40nm PDK goofup Fermi actual had.

There was no shrink, but it was also really incremental and launched like 9mo after the Fermi actual.

OMG that wasn't a node shrink? It was so much more efficient and everyone loved it. I remembered it as a shrink. Wow.
 

adroc_thurston

Diamond Member
Jul 2, 2023
4,714
6,501
96
OMG that wasn't a node shrink? It was so much more efficient and everyone loved it. I remembered it as a shrink. Wow.
yeah, Maxwell was good.
In general NV has a pretty extensive history of sticking to older nodes and building meaner uarches on them.
Blackwell is their shiny new future SM (sm100-thingy) and it's this.
 
Reactions: Tlh97 and moonbogg

Mopetar

Diamond Member
Jan 31, 2011
8,201
7,027
136
While I agree this is a total garbage release, nothing quite says pathetic nerd, as loudly as making spread sheet to "prove" the name of the GPU is wrong.

He's not wrong though. If you look at the current branding compared to historic naming conventions, NVidia is selling lesser GPUs under names previously used to indicate greater chips.

If you're not picking up on it, imagine Ford releasing a new Mustang only it's actually a Fiesta or whatever their low-end model is these days. They've been watering down their own brand for a while now.
 

jpiniero

Lifer
Oct 1, 2010
15,633
6,110
136
He's not wrong though. If you look at the current branding compared to historic naming conventions, NVidia is selling lesser GPUs under names previously used to indicate greater chips.

The gap between the 2 die and the rest are only going to get bigger over time now.
 

Heartbreaker

Diamond Member
Apr 3, 2006
4,653
6,108
136
He's not wrong though. If you look at the current branding compared to historic naming conventions, NVidia is selling lesser GPUs under names previously used to indicate greater chips.

If you're not picking up on it, imagine Ford releasing a new Mustang only it's actually a Fiesta or whatever their low-end model is these days. They've been watering down their own brand for a while now.

I disagree. It would be like if Ford released a New Mustang that was just like the previous Mustang, because that is exactly what happened.
 

adroc_thurston

Diamond Member
Jul 2, 2023
4,714
6,501
96
NVidia is selling lesser GPUs under names previously used to indicate greater chips.
no? there's a small horde of 300-400mm^2 dies in x8 tier. Goes way back when.
Things like 780/ti, 3080 or 2080 are outliers.
People doing this whole nerd rage chart stuff should actually try and go back all the way to FX5800.
 

linkgoron

Platinum Member
Mar 9, 2005
2,493
1,131
136
OMG that wasn't a node shrink? It was so much more efficient and everyone loved it. I remembered it as a shrink. Wow.
As stated Maxwell wasn't a shrink, rdna wasn't (from Vega20) and iirc rdna2 wasn't as well. All actual improvements on the same node.

no? there's a small horde of 300-400mm^2 dies in x8 tier. Goes way back when.
Things like 780/ti, 3080 or 2080 are outliers.
People doing this whole nerd rage chart stuff should actually try and go back all the way to FX5800.
Iirc the 8 cards were almost always the big zero die (later refreshes might have had a different die) until the 680 which used the 4 die
 

adroc_thurston

Diamond Member
Jul 2, 2023
4,714
6,501
96
As stated Maxwell wasn't a shrink, rdna wasn't (from Vega20) and iirc rdna2 wasn't as well. All actual improvements on the same node
NV in general has a track record sticking iso node and going for a shrink later.
It was ATi that was always aggressive with the node transition.
Iirc the 8 cards were almost always the big zero die (later refreshes might have had a different die) until the 680 which used the 4 die
Yeah but all that matters is the final area commitment which has been 300-400mm^2 with some outliers here and there.
 

MrTeal

Diamond Member
Dec 7, 2003
3,749
2,137
136
The 5080 isn't a failure of a card because it's only a 378mm² die, it's a failure because after two years it brought no meaningful improvement over its predecessor.
If GB204 clocked in at 311mm² with 16GB VRAM but it was 30% above a 4080 Super at 4k, people would buy it hand over fist. Some people here would still whinge about die size, but overall no one would care.
 

Hans Gruber

Platinum Member
Dec 23, 2006
2,448
1,316
136
If the 5080 used the same amount of power as the 4080, that would be one thing. It doesn't and the power consumption is way beyond the 8% performance improvement. Add to it the performance regression considering the GDDR7 memory is 30gbps vs 22gbps of the 4080. It's about time Nvidia has a bust of a GPU launch. They have been executing consistently over more than a handful of GPU generations.

New generations of GPU's used to be released every 18 months. Now it's between 2.5-3years. You get 10% better performance with the 5080. AMD should make a run of 7900xtx on N4P. There is even room for Intel if they release their B770 card for the value play.
 
Reactions: igor_kavinski

gaav87

Senior member
Apr 27, 2024
452
794
96
For anyone still arguing about 5070ti and 5070 performance. 5080 has higher fp16 and higher pixel and texture fill rate than 4080.
5070ti and 5070 have everything lower.
Add lower cache bandwidth and worse cache latency. Add worse gddr7 latency (only better bandwidth).
I can see 5070 losing in 1080p vs 4070FE in some games not even 4070S coz its light-years away.
5070ti should be 5-10% behind 4070tiS in 4k and prolly more with lower resolution.

 

Meteor Late

Senior member
Dec 15, 2023
266
292
96
no? there's a small horde of 300-400mm^2 dies in x8 tier. Goes way back when.
Things like 780/ti, 3080 or 2080 are outliers.
People doing this whole nerd rage chart stuff should actually try and go back all the way to FX5800.

One has to compare with the flagship of the same generation. It is very disingenuous to compare mm2 of 70 or 80 class GPU with that of another generation if the flagship got much bigger.
For example, with another similar chungus GPU, RTX 2080 ti, we got RTX 2080 with 545 mm2, with proper bus width and cuda core count compared to flagship, that is to say it's nowhere near half everything. RTX 2070 is 445 mm2, still a bit more than RTX 5080. Like it's a total joke.
 

Ranulf

Platinum Member
Jul 18, 2001
2,653
1,898
136
Last edited:

Win2012R2

Senior member
Dec 5, 2024
647
609
96
Read this bit yesterday -


And it seems clear why new improved SM did not make any difference - they can't do FP32/INT at the same time now! AFAIK before Ampere/Ada they could do FP32+INT or FP32+FP32, so now the only thing that changed is that they can do INT+INT too, which as we can see does not do a lot. It started with fake cores, now fake frames...
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |