Nvidia GeForce RTX 3080 Founders Edition (Ampere)

[1/2Genius1/2Wit]

That depends if you consider 20-25% similar or not, I don't really.

I said similar *performance per dollar*, not overall performance. C'mon, don't straw-man me

[Zhaoman]

Thanks, interesting link. Unless I’m reading it incorrectly though, the 3080 averaged 48fps in their test @1440p, versus 37 for my 1080Ti. So the 3080 is giving roughly a 30% boost? I know FS is more cpu-limited than most modern games, but that still seems a decent increase to me. Especially after watching the slideshow of New York last night!

Yeah but I thought excalibur's point was that you'd get better value by getting a second hand 2080Ti because they show similar performance at 1080p and 1440p. It's only at 4k that 3080 pulls ahead significantly. Of course everyone's judgement of value is different but in general I think it's clear the 3080 is a little disappointing in 1080p and 1440p.

[kalniel]

I think it's clear the 3080 is a little disappointing in 1080p and 1440p.

I don't think it's clear at all. It's not the 3080 that is 'disappointing' at 1080p but rather there's no point going for a mega fast GPU if you are being held back by a CPU. So if anything you should be disappointed in the CPU. If frame rates in the 100s are disappointing that is.

[CAT-THE-FIFTH]

The RTX3080 consumes more power under load than my entire system. Its one of the reasons why I got an HD5850 instead of a GTX460 or GTX470 for my Shuttle XPC at the time.

[1/2Genius1/2Wit]

I think it's clear the 3080 is a little disappointing in 1080p and 1440p.

I don't think it's clear at all. It's not the 3080 that is 'disappointing' at 1080p but rather there's no point going for a mega fast GPU if you are being held back by a CPU. So if anything you should be disappointed in the CPU. If frame rates in the 100s are disappointing that is.

There's a lot of assumptions that it's the CPU causing this, but that's not definitive. Ampere performs best when the balance of its workload is FP32 - that's more likely at higher resolutions and detail settings where it will spend more time on pixel shaders than on geometry. It may be that the CPU is the limit in some instances, but in others it's clear that the card simply performs better (relatively speaking) when it has more shading work to do. 1080p (and 1440p to a lesser extent) just aren't optimal workloads for the architecture.

[[GSV]Trig]

Meh, just got the "Notification" email from NVidia for the FE 3080, clicked on the link straight away, out of stock...

[cptwhite_uk]

Trig, same, Jebaited. I wasn't that bothered anyway to be honest, so I find the whole thing quite amusing.

[Zhaoman]

I don't think it's clear at all. It's not the 3080 that is 'disappointing' at 1080p but rather there's no point going for a mega fast GPU if you are being held back by a CPU. So if anything you should be disappointed in the CPU. If frame rates in the 100s are disappointing that is.

Yeah what 1/2Genius said above.
If it was as simple as a CPU bottleneck I'm sure reviewers would have pointed that out quite clearly. The 3080's 'doubled' CUDA cores are not able to realise their potential when not given FP32-heavy workloads. So 1080p and 1440p performance seems to scale differently compared to 4k. This is not a CPU bottleneck in most cases. (But in FS2020 it probably was!)

[ik9000]

Getting excited for the second-hand market about now..

I see a good replacement for my RX580 8Gb on the horizon...

in the meantime a certain river continues to sell the 2080ti for well over £1k. LOL.

[dannyboy75]

Meh, just got the "Notification" email from NVidia for the FE 3080, clicked on the link straight away, out of stock...

Same here. Stock, my arse.

I did see this when I went onto nVidia's site though, which got my attention for a moment. Of course, the link didn't work, and when I went to the retailer's website the card was listed at normal price.

[[GSV]Trig]

Yeah I noticed that link, went to something else...

[ik9000]

The RTX3080 consumes more power under load than my entire system. Its one of the reasons why I got an HD5850 instead of a GTX460 or GTX470 for my Shuttle XPC at the time.

the 460 was very good on power draw. It was the 470 and 480 where things jumped up.

[Taniniver]

I've got a question about your testing methodology -

You mention your test setup uses a Fractal Design Define 7 Clear Tempered Glass case. I notice that case has a flexible design of the roof section and can be either closed or open.

The founders edition has a unique fan design where one fan will blow air directly toward the roof of the case instead of hitting the PCB and being mostly deflected.

The case setup may therefore have some impact on the cooling performance in comparison to partner cards that use a more traditional design with fans (mostly) having their airflow stopped by the PCB.

Could you tell us if the vented panel on the case is open or closed, and what setup you are using in terms of case fans, e.g. any exhaust fans in the roof?

[CAT-THE-FIFTH]

the 460 was very good on power draw. It was the 470 and 480 where things jumped up.

Even a GTX460 consumed more power than an HD5850 which was faster!!

[Shante]

Such a great card but why did it have to be an awful launch

[[GSV]Trig]

I've got a question about your testing methodology -

You mention your test setup uses a Fractal Design Define 7 Clear Tempered Glass case. I notice that case has a flexible design of the roof section and can be either closed or open.

The founders edition has a unique fan design where one fan will blow air directly toward the roof of the case instead of hitting the PCB and being mostly deflected.

The case setup may therefore have some impact on the cooling performance in comparison to partner cards that use a more traditional design with fans (mostly) having their airflow stopped by the PCB.

Could you tell us if the vented panel on the case is open or closed, and what setup you are using in terms of case fans, e.g. any exhaust fans in the roof?

Valid point, I've seen YT reviewers running on an open test bench to show just the cooler working rather than airflow and fans in a case etc, still, if your tests are all ran under the same conditions , and your monitoring ambient then you have a baseline for everything.