NVIDIA GeForce RTX 5090 Founders Edition

[Contra]

last version is 2.61.0

This version doesn't seem to support rtx50 - no chip/memory type info

 

[desnikraz87]

This version doesn't seem to support rtx50 - no chip/memory type info

So can i get the version 2.61.2 ?

 

[Contra]

So can i get the version 2.61.2 ?

Get 2.62 released today

 

[hardwarefan]

NVIDIA is betting on ray tracing and Blackwell comes with several improvements here. Our results show a several percent smaller performance loss from enabling RT, which definitely helps...

I have an issue with this statement in the conclusion. As I go to the "ray tracing" section, I observe the performance loss from ray tracing is nearly identical from that of ray tracing with the 4090, which would directly contradict this statement:

 

[hardwarefan]

I have one more question…
Does the RTX5070 have the same fps as the RTX4090 as they have announced?

If I may:

If, and only if, you use it with multi-frame gen.
Without multi-frame gen, by looking at the specs, and judging also from the comparison of the 5080 to the 4080, it will likely only match the 4070 super.

PD: What a horrendously misleading claim by jensen. Bordering on false advertising.
It should have said:
Multi-frame gen is so powerful, it can even make the 5070 match the 4090.

 

[b1k3rdude]

Moose Muffins!

 

[lexluthermiester]

View attachment 383293

I did just google that term. Had no idea there were that many references. For the record, I got the phrase from my grandfather. It was his alternate, less harsh and more polite way of saying "bullsh**". Similar to cow-patties but of the moose variety.

 

[SRS]

"at the SKU's baseline price of USD $1,999"

Ostensible price of $1999.

Anyone who paid attention to how the pricing for the 4090 went knows how this will go. A tiny number of FEs will be available, sporadically, and everyone else will have to buy 3rd-party cards with tiny overclocks for quite a bit higher price. The threads and trackers for "hope" over getting an FE 4090 is something gamers (and home AI enthusiasts) shouldn't forget nor should tolerate.

Only $400? What a steal.

 

[x4it3n]

I have an issue with this statement in the conclusion. As I go to the "ray tracing" section, I observe the performance loss from ray tracing is nearly identical from that of ray tracing with the 4090, which would directly contradict this statement:

View attachment 383157View attachment 383158

=> 100 / 76 = 31.57%
=> 60.1 / 44.7 = 34.45%


I think you meant to compare 4K RT OFF vs 4K RT ON right ?

=> 100 / 74 = 35.13%
=> 100 / 76 = 31.57%

In that case yes, RT is not better on RTX 50s in terms of IPC (at least as of now, maybe it will get better in several months with newer drivers).

 

[hardwarefan]

I think you meant to compare 4K RT OFF vs 4K RT ON right ?

Yes, I meant to compare that.
But I also wanted to show the graph for Alan Wake 2 as well, as a concrete example of the loss of performance from ray tracing for the 5090 vs 4090. Its only one example, sure, but it can be seen to be fairly representative of the ray tracing results of the 5090 vs 4090.

Yes, I can't discard the possibility it may get better with new drivers.
But for now at least, rtx 50 series doesn't seem to have significant improvements in ray tracing loss of performance vs 40 series.

 

[x4it3n]

Yes, I meant to compare that.
But I also wanted to show the graph for Alan Wake 2 as well, as a concrete example of the loss of performance from ray tracing for the 5090 vs 4090. Its only one example, sure, but it can be seen to be fairly representative of the ray tracing results of the 5090 vs 4090.

Yes, I can't discard the possibility it may get better with new drivers.
But for now at least, rtx 50 series doesn't seem to have significant improvements in ray tracing loss of performance vs 40 series.

I think Blackwell needs a lot more RT/PT improvement(s), Lovelace has gotten better over the past 2 years too, so I wouldn't be surprised if we see more and more games getting closer to 40-50% more performance vs 4090 due to some enhanced RT capabilities + much higher bandwidth use.

We all wish Nvidia went all-in and with a 3nm node for raw performance like they did with Lovelace when it released but they're all betting on AI and huge margins so unless AMD or Intel can really compete, it's not going to happen.

 

[b1k3rdude]

I did just google that term. Had no idea there were that many references. For the record, I got the phrase from my grandfather. It was his alternate, less harsh and more polite way of saying "bullsh**". Similar to cow-patties but of the moose variety.

Well you could use a brit term " Load of Bollox " :-D

 

[Artas1984]

In a years timelap i will be waiting for the GeForce RTX 5090 Ultra release in response to the upcoming Radeon 9080 XT eventuality. Back in 2003 i was siting on my FX 5200 and could only dream of FX 5900 Ultra or 9800XT. Eventually in 2005 i've got that FX 5900 (for fun) and it got beaten in Resident Evil 4 by even a low life such as GeForce 6200, not to mention how bad it was smashed by my GF 6600GT. I am talking about 2X to 3X improvement in FPS from FX 5900 to 6600GT. You don't get to see improvements like that these days from high end to mid range.

 

[x4it3n]

In a years timelap i will be waiting for the GeForce RTX 5090 Ultra release in response to the upcoming Radeon 9080 XT eventuality. Back in 2003 i was siting on my FX 5200 and could only dream of FX 5900 Ultra or 9800XT. Eventually in 2005 i've got that FX 5900 (for fun) and it got beaten in Resident Evil 4 by even a low life such as GeForce 6200, not to mention how bad it was smashed by my GF 6600GT. I am talking about 2X to 3X improvement in FPS from FX 5900 to 6600GT. You don't get to see improvements like that these days from high end to mid range.

I hope to be wrong, but I really doubt we will see a 9080 XT... AMD said they were aiming at Mainstream GPUs with RDNA 4 to gain market share, and we still don't know much about UDNA/RDNA5.
Nvidia having 90% Market share, AMD have to make RDNA 4 & UDNA a real bargain. But to get some really huge market share gains (including high-end GPUs), AMD would have to beat the 5090 in raw performance with UDNA next year while having a very compelling performance/price ratio. It would be amazing if it happened but yeah, wait and see!

 

[jnv11]

Based on the following video with der8auer discovering a massive safety hazard where one wire in the 12V-2x6 power cable was carrying over 22 amps causing temperatures to rocket to at least 150 degrees Celsius, I think that TechPowerUp needs to retest its RTX 5090 Founders Edition and possibly revoke the Editor's Choice award if TechPowerUp can replicate the safety hazards:

Buildzoid made a video about this issue below:

After viewing der8auer's video, Buildzoid made a follow-up video below:

 

[vgm]

@W1zzard Did you get access to architecture comparison with Ampere, Ada and Blackwell. I came across this architecture pdf file when I was searching for CUDA 12.X documentation and SDK for 4070 Super.
By the looks of it, compute wise Blackwell is just AI optimized card with heavy focus on FP4 performance reaching upto 3000 TOPS on 5090. I'm unsure if its actual info or just misinformation through clever manipulation of SEO.

Blackwell Architecture PDF

 

[W1zzard]

By the looks of it

page 12, some changes to the shader architecture, page 13, gddr7 memory is huge, page 16 changes to rt cores, some require game support but could be good once materialized, page 27 quite a bit of changes to the codecs, page 35, hardware units for dlss 4, and yes, fp4 and int4 on the tensor cores

 

[vgm]

page 12, some changes to the shader architecture, page 13, gddr7 memory is huge, page 16 changes to rt cores, some require game support but could be good once materialized, page 27 quite a bit of changes to the codecs, page 35, hardware units for dlss 4, and yes, fp4 and int4 on the tensor cores

FP32 performance remains marginal for 80 and 70 series Blackwell when compared against Ada and feels barely an upgrade if someone already has a Super version. Even die size for both 4080/5080 is similar and 5070 non Ti has a smaller die size and less dense than 4070. FP32/BF16 was essentially the same that's what confused me.

 

[Gameslove]

"Editor Choice" must be removed, due latest news about RTX 5090...

 

[jnv11]

A new YouTube video alleges that RTX 50 series is bad with DirectX 9:

Could someone cross check to see if the RTX 50 series is bad with DirectX 9? Such a test would be valuable to determine if the YouTuber's allegation is real or if there is something wrong with the YouTuber's setup like a driver installation issue.

We already have had the RTX 50 series remove 32-bit support for CUDA, OpenCL, and PhysX.