Discussion Nvidia Blackwell in Q1-2025

[Sunaiac]

Why did they include an $800 5080 option in their perf/$ chart? This makes no sense. . .

Well he needed to have it above the 7900XTX I guess
I won't stay on the subject longer, but I'll just add that he's always been nothing more than an nvidia advertiser for me. His +/- are always trololol, he's always had a selection of games and drivers and scenes that are kinda biased (the gtx 680 vs 7970 was some sight to behold), and he pretty much always show more difference in favor of a certain brand than the average of tests, which is over 20 years statistically unprobable.

Anyway for the 5080, before even talking about the 7900XTX, let's focus on the fact the 4080S is basically the same at the same price

 

[blckgrffn]

Well he needed to have it above the 7900XTX I guess
I won't stay on the subject longer, but I'll just add that he's always been nothing more than an nvidia advertiser for me. His +/- are always trololol, he's always had a selection of games and drivers and scenes that are kinda biased (the gtx 680 vs 7970 was some sight to behold), and he pretty much always show more difference in favor of a certain brand than the average of tests, which is over 20 years statistically unprobable.

Anyway for the 5080, before even talking about the 7900XTX, let's focus on the fact the 4080S is basically the same at the same price

Noted.

Worse on the 4080S front - they were easily available last fall (late November into mid December) here in the states for $900-$1k, so we are seeing a regression in price, too, IMO. The 4080S seemed prime to move to ~$899 territory, and as inflated as everything is at least it was a powerful a card and that represented a nice little slide down from $1200 at "launch" for the 4080.

This is just a MSRP reset? Super lame. And since the gulf is even wider to the 90 series, my guess is like @moonbogg prognosticated there will be people "forced" to settle for the 5080 as the 5090 is a bridge to far and the 4090 is unobtainium.

"yay"

 

[MrTeal]

Why did they include an $800 5080 option in their perf/$ chart? This makes no sense. . .

I like it. "This is where the 5080 would place if it was priced correctly."

 

[DAPUNISHER]

Blackwell has a CPU overhead issue. The Delta Force results seriously highlight it with the 5080.

 

[poke01]

Blackwell has a CPU overhead issue. The Delta Force results seriously highlight it with the 5080.

it’s slower in RT than a 4080 Super, which indicates those new RT cores are not so new and if new they are bad.

This architecture has many issues

 

[adroc_thurston]

it’s slower in RT than a 4080 Super, which indicates those new RT cores are not so new and if new they are bad.

They are new but the L1 cache bandwidth regression is severe enough to hamper RT workloads I guess.

This architecture has many issues

No, but some glaring ones indeed.

 

[NTMBK]

I'm looking forward to the Chips and Cheese deep dive on this one. What the hell happened to this architecture?

 

[CakeMonster]

Blackwell has a CPU overhead issue. The Delta Force results seriously highlight it with the 5080.

There is some truth to that, while I think the new chips are unimpressive a lot of people will look at a ton of graphs and not realize that a good amount of those graphs are held back by the CPU. I agree with 60% of the criticism I guess

Oh, and 4nm also holds back performance as well, but they could have made 5080 and lower slightly larger to get to 20-30% performance increase, but they cheaped out.

 

[DAPUNISHER]

I'm looking forward to the Chips and Cheese deep dive on this one. What the hell happened to this architecture?

After you read it please provide me with the C&C deep dive for dummies version. 🤪

 

[RnR_au]

Quote
Also, the way AMD is handling Radeon lately makes me wonder if their discrete GPU brand will still be around in two or three years (c)

Not happening.

Are we sure they are not just sourcing Anandtech forum discussion in regards to AMD make a new graphics brand to relaunch their consumer gpu operations?

 

[Dave3000]

It looks like I'm good with my RTX 4090 for a few more years.

 

[Thunder 57]

After you read it please provide me with the C&C deep dive for dummies version. 🤪

That's how I feel about GPU reviews especially on C&C. I understand much of their CPU stuff but when it comes to GPU's I'm much less knowledgeable. They recently published how most GPU L1 caches are either "read only" or "write through". Write through I get, but how can you have a "read only" cache?

 

[adroc_thurston]

but how can you have a "read only" cache?

writes go to gmem from round go.
Good old antique GPU stuff before they became all CPU-like and compute and programmable and blahblah.

 

[Timmah!]

Dual-pumping ALUs make very little difference if they're not increasing caches and registers accordingly. Turing introduced FP32 + INT32 and that's where the bulk of performance increase per-SM happened. Ampere then made it FP32/INT32+FP32 and it did almost nothing, and now it's FP32/INT32+FP32/INT32 and it did nothing.


The truth is dual-pumping ALUs seems to be super cheap transistors and area-wise, but performance gains are just as small.
AMD saw the same when they introduced dual-pumped ALUs in RDNA3.

Did nothing almost nothing where? Cause in Octane Render its likely the result why 3090 was almost 2x as fast as 2080Ti.

 

[adroc_thurston]

Did nothing almost nothing where? Cause in Octane Render its likely the result why 3090 was almost 2x as fast as 2080Ti.

It also had 2x as much L1 bandwidth. Hope that helps.

 

[linkgoron]

WOW what a disappointing 5080. Just a slight bump higher than the 4080. Can't believe TPU gave it a "highly recommended" badge, they are a joke. Well, they also said that the the 600 watt 5090 is "Highly energy-efficient" so what can I say.

Can't believe that the geniuses at AMD thought that this was the correct gen not do to a big die.

 

[Meteor Late]

Lisa Su is really the biggest Nvidia hero, always being saved by her, must feel nice.

 

[Thunder 57]

WOW what a disappointing 5080. Just a slight bump higher than the 4080. Can't believe TPU gave it a "highly recommended" badge, they are a joke. Well, they also said that the the 600 watt 5090 is "Highly energy-efficient" so what can I say.

Can't believe that the geniuses at AMD thought that this was the correct gen not do to a big die.

They probably didn't know. Too bad, because they could've struck back hard.

 

[IEC]

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:

 

[adroc_thurston]

Daniel Owen makes the case that the 5080 is really a 5070 Ti tier class of GPU.

what is this cope.
'tier' is arbitrary cope by people who do not understand semicon at all.

Blackwell problem is the uarch is a nothingburger.
Like RDNA3 was a miss, but at least it delivered something.

 

[jpiniero]

Galax GeForce RTX 5080 1-Click OC Review

The Galax GeForce RTX 5080 1-Click OC comes at no price increase over NVIDIA MSRP: it sells for $1000. That makes it an excellent alternative to the NVIDIA Founders Edition. You even get an improved cooling solution that runs quieter than NVIDIA's reference card.

www.techpowerup.com

TPU was able to get their 5080 to boost over 3 Ghz. So it looks like there is no actual regression on clock speeds in the design. It's just that (due to not using N4P?) that the 5090 cannot do it at 600 W.

 

[IEC]

what is this cope.
'tier' is arbitrary cope by people who do not understand semicon at all.

Blackwell problem is the uarch is a nothingburger.
Like RDNA3 was a miss, but at least it delivered something.

Nothing you said addresses the point. As usual, you lack intelligence, you do not address the content of the post, and your narcissistic insistence on denigrating other posters is duly noted.

The point still stands. 5080 is the worst x80 series card in value and generational uplift... ever.

 

[Mopetar]

50 series is a bit of a disappointment. I expected more.

This comment is a bit of an understatement. I expected more.

That’s why some people refer to him as Jayz2braincells.

Who'd he borrow the other one from?

W1zzard is a troglodyte. As is Jarred. "Driver issues in games and apps". Less than 10% faster than a 4080 Super? Still 3.5 stars!

I was contemplating what would get them to give it anything under 3 stars and wondered if it would take a card that summons JHH to piss in their mouths, but I figured that would be an instant 6/5 for them.

After you read it please provide me with the C&C deep dive for dummies version. 🤪

I'll save you the trouble of waiting. "Not enough chip, too much cheese."

 

[adroc_thurston]

5080 is the worst x80 series card in value and generational uplift... ever.

Because the uarch is a nothingburger. It's a new SM that does ~nothing.
Remember Maxwell? Now that was an SM that did a lot of something.
No DPFP part but man did they deliver on PPA bumps across the stack.

 

[moonbogg]

Same process node this time, right? So imagine Nvidia trying to release the GTX 500 series on the same node as the GTX 400 series. The node shrink literally was the GTX 500 series. Without it, you'd have the same garbage we see now with the 5000 series. Thing is though, Nvidia could have offered more value with better prices and more ram. To that, they said "no".