Will you buy a RTX 5090?

[Tyler-98-W68]

Passmark is a great tool to compare system performance, however its relatively poor when it comes to comparing GPU performance. Too virtually everything else is way faster on a 5090 vs a 5090.

Also your 4080 sucks because my 5070 gets a higher 3d rating than yours

PassMark Software - Display Baseline ID# 2625091

www.passmark.com

See my point, comparing passmark cpu results is relevant, it's not for comparing gpus.

Further to the point, using your numbers my 5070 ti is better than a 5090....but it's not

PassMark Software - Display Baseline ID# 2625014

www.passmark.com

Then again my 5090 is scoring higher than 39000

46994 to be exact

PassMark Software - Display Baseline ID# 2709937

www.passmark.com

So again passmark 3d score is not a reliable method to compare GPUs. Accept the fact a 5090 crushes a 4080 super in everything

I did a poor score with a 5090....but it was on my 10 year old system

PassMark Software - Display Baseline ID# 2712400

www.passmark.com

Also this is the reason

https://www.reddit.com/r/hardware/comments/1iz8v8p/passmark_discovered_the_real_reason_the_50905080

 

[Pets4Ever]

Passmark is a great tool to compare system performance, however its relatively poor when it comes to comparing GPU performance. Too virtually everything else is way faster on a 5090 vs a 5090.

Also your 4080 sucks because my 5070 gets a higher 3d rating than yours

PassMark Software - Display Baseline ID# 2625091

www.passmark.com

See my point, comparing passmark cpu results is relevant, it's not for comparing gpus.

Further to the point, using your numbers my 5070 ti is better than a 5090....but it's not

PassMark Software - Display Baseline ID# 2625014

www.passmark.com

Then again my 5090 is scoring higher than 39000

46994 to be exact

PassMark Software - Display Baseline ID# 2709937

www.passmark.com

So again passmark 3d score is not a reliable method to compare GPUs. Accept the fact a 5090 crushes a 4080 super in everything

The numbers I quoted are the average of many systems from Passmark, they're not measurements of mine.

Some cards will be higher, some will be lower, that's how averages work. Obviously the 5090 is more powerful, it better be for drawing around double the wattage under load! The point is that the cost/performance increase is not acceptable in my opinion. For that kind of cost I'd want to see at least a 100% increase in average performance.

 

[Tyler-98-W68]

The numbers I quoted are averages of many systems from Passmark, they're not measurements of mine.

Some cards will be higher, some will be lower, that's how averages work. Obviously the 5090 is more powerful, it better be for drawing around double the wattage under load! The point is that the cost/performance increase is not acceptable in my opinion. For that kind of cost I'd want to see at least a 100% increase in average performance.

https://www.reddit.com/r/hardware/comments/1iz8v8p/passmark_discovered_the_real_reason_the_50905080

That's the reason, why does almost every other bench mark show the 5090 almost 50% faster than a 4080 super? Sure it's not worth it for most people, but you cherry picked a single benchmark to try and throw shade on it.

As for efficiency it's actually not as bad as you make it out to be.

NVIDIA GeForce RTX 5090 Founders Edition Review - The New Flagship

NVIDIA's GeForce RTX 5090 is the fastest graphics card ever released. It comes with a whopping 32 GB VRAM, and support for multi-frame generation, which achieves hundreds of FPS easily. We also managed to disassemble the card, and our review includes performance testing without upscaling or...

www.techpowerup.com

0.2w more per frame???

 

[Pets4Ever]

https://www.reddit.com/r/hardware/comments/1iz8v8p/passmark_discovered_the_real_reason_the_50905080

That's the reason, why does almost every other bench mark show the 5090 almost 50% faster than a 4080 super? Sure it's not worth it for most people, but you cherry picked a single benchmark to try and throw shade on it.

As for efficiency it's actually not as bad as you make it out to be.

NVIDIA GeForce RTX 5090 Founders Edition Review - The New Flagship

NVIDIA's GeForce RTX 5090 is the fastest graphics card ever released. It comes with a whopping 32 GB VRAM, and support for multi-frame generation, which achieves hundreds of FPS easily. We also managed to disassemble the card, and our review includes performance testing without upscaling or...

www.techpowerup.com

0.2w more per frame???

One benchmark...that was run on thousands of systems.

But since you insist, look at this chart from TPU:

NVIDIA GeForce RTX 5090 Founders Edition Review - The New Flagship

NVIDIA's GeForce RTX 5090 is the fastest graphics card ever released. It comes with a whopping 32 GB VRAM, and support for multi-frame generation, which achieves hundreds of FPS easily. We also managed to disassemble the card, and our review includes performance testing without upscaling or...

www.techpowerup.com

25% difference @ 1080p
33% difference @ 1440p
42% difference @ 2160p

The conclusion is the same, the difference does not justify a tripling of price and a near-doubling of TDP for me.


You need to get a hold of yourself. I didn't say noone should buy it. I said for me the math doesn't fly and I'm waiting.

NVIDIA GeForce RTX 5090 Founders Edition Review - The New Flagship

NVIDIA's GeForce RTX 5090 is the fastest graphics card ever released. It comes with a whopping 32 GB VRAM, and support for multi-frame generation, which achieves hundreds of FPS easily. We also managed to disassemble the card, and our review includes performance testing without upscaling or...

www.techpowerup.com

 

[Tyler-98-W68]

Yes for your math using a highly inaccurate benchmark, sorry I had to point that out to you.

Not sure what your issue is, in THAT benchmark only It doesn't show a large increase in performance, which again doesn't happen in any other real world testing or benchmark.

Why did you ignore the efficiency numbers (right from TPU) so the 5090 is so much "worse" from a power consumption standpoint, but it's only 0.2w per frame more power hungry than a 4080 super.....

If you are not happy with only a 25% difference at 1080p good for you, why does it grow to 42% at 4k, which is where most people who buy a 5090 would be gaming at.

 

[Pets4Ever]

Yes for your math using a highly inaccurate benchmark, sorry I had to point that out to you.

Not sure what your issue is, in THAT benchmark only It doesn't show a large increase in performance, which again doesn't happen in any other real world testing or benchmark.

Why did you ignore the efficiency numbers (right from TPU) so the 5090 is so much "worse" from a power consumption standpoint, but it's only 0.2w per frame more power hungry than a 4080 super.....

If you are not happy with only a 25% difference at 1080p good for you, why does it grow to 42% at 4k, which is where most people who buy a 5090 would be gaming at.

The real world testing from TPU shows better numbers, but not enough better for it to matter.

Efficiency has nothing to do with my original post which is cost/performance ratio = no buy from me, a direct response to the thread topic. I only mentioned TDP because 600 watts is another downside on top of cost. It's a downside because you'd need a bigger PSU, might need to rework your case cooling, and you're upping your electic bill.

The bold part highlights the absolute and utter ridiculousness of this conversation. I never said that no one should buy a 5090, or that there was no reason whatsoever to buy one. I specifically stated that I wasn't going to buy it in response to the topic of the thread. It is completely irrelevant if some other guy with a 4k monitor needs one. I'm never going to buy a 4k monitor for gaming so that I can pay $3000 to get the same framerates I have now at 1440p

 

[iameatingjam]

Passmark is a great tool to compare system performance, however its relatively poor when it comes to comparing GPU performance.

True. My run to run variance is insane with passmark GPUs. I actually like passmark, mostly because its fast and I'm inpatient, and its my primary tool for comparing bios settings when I adjust for things like cpu and ram.

But its unreliable when it comes to GPUs, at least in my experience.

I'm still not buying 50 series though. Well... not unless I have an unexpected failure. Actually I'm not sure what I'd do in that situation.

0.2w more per frame???

Well yeah, seems to be like blackwell is ada refresh 2. So less efficiency makes sense as other ways are used to increase performance, higher clocks, more vram channels, just a bigger die, etc. GDRR7 is more efficient.... but I guess thats not enough to make up for the rest of it.

 

[Tyler-98-W68]

@Pets4Ever

This is what you originally posted, and what I took issue with.

4080 Super
Cost: $1000 (what I paid)
Passmark DX11: 302 FPS
Passmark DX12: 133 FPS
G3D Rating: 34215

5090
Cost: ~$3000
Passmark DX11: 344 FPS
Passmark DX12: 184 FPS
G3D Rating: 39748

Conclusion:


5090 is 13.9% faster overall and costs ~300% more

You made a claim that a 5090 is only 13.9% faster than a 4080 Super yet costs 300% more.

Your source was (as I have shown) incredibly flawed and does not represent every other test showing the performance of a 5090.

So your whole argument on performance was that a 5090 only is 13.9% faster in Passmark's 3D test? And we should ignore every other test/benchmark of a 5090? Like ignoring TPU's tests showing a 5090 is between 40-50fps faster at 1080p 1440p and 4K than a 4080 Super? Why only focus on a single benchmark which is known to be flawed?

Drink from AMD's trough, drink from Nvidia's trough or whoever you like, it makes no difference to me, but at least when you try to make bold statements to justify why you don't want to buy something, use an accurate representation of facts, not a cherry picked flawed benchmark

Your power consumption numbers also make no sense saying how inefficient the 5090 is when there is a 0.2W difference.

So a 4080 super is really efficient at doing what it does, but the 5090 is really bad??



So what if the 4080 Super was drawing 600w, and giving 5090 performance? Would that make it a bad card?

 

[webmanau]

The numbers I quoted are the average of many systems from Passmark, they're not measurements of mine.

Some cards will be higher, some will be lower, that's how averages work. Obviously the 5090 is more powerful, it better be for drawing around double the wattage under load! The point is that the cost/performance increase is not acceptable in my opinion. For that kind of cost I'd want to see at least a 100% increase in average performance.

Forget the flawed tool. There is no place in the real world where a 5090 is only 10% faster than a 4080 Super unless you’re dropping both from a tall building.

One benchmark...that was run on thousands of systems.

But since you insist, look at this chart from TPU:

NVIDIA GeForce RTX 5090 Founders Edition Review - The New Flagship

NVIDIA's GeForce RTX 5090 is the fastest graphics card ever released. It comes with a whopping 32 GB VRAM, and support for multi-frame generation, which achieves hundreds of FPS easily. We also managed to disassemble the card, and our review includes performance testing without upscaling or...

www.techpowerup.com

25% difference @ 1080p
33% difference @ 1440p
42% difference @ 2160p

The conclusion is the same, the difference does not justify a tripling of price and a near-doubling of TDP for me.


You need to get a hold of yourself. I didn't say noone should buy it. I said for me the math doesn't fly and I'm waiting.

NVIDIA GeForce RTX 5090 Founders Edition Review - The New Flagship

NVIDIA's GeForce RTX 5090 is the fastest graphics card ever released. It comes with a whopping 32 GB VRAM, and support for multi-frame generation, which achieves hundreds of FPS easily. We also managed to disassemble the card, and our review includes performance testing without upscaling or...

www.techpowerup.com

Your 42% difference is really a 72% increase in frame rate and that doesn’t do it justice either. The more demanding games tend to have a bigger delta and the 5000 series generally overclocks very well.

 

[Godrilla]

Forget the flawed tool. There is no place in the real world where a 5090 is only 10% faster than a 4080 Super unless you’re dropping both from a tall building.


Your 42% difference is really a 72% increase in frame rate and that doesn’t do it justice either. The more demanding games tend to have a bigger delta and the 5000 series generally overclocks very well.

Derbauer did a shunt mod for the astral lc 5090 oc to rtx 6000 performance. I'm personally getting around 35% delta form air 5090 pny oc from Suprim liquid 4090 both at 3ghz.

 

[webmanau]

Derbauer did a shunt mod for the astral lc 5090 oc to rtx 6000 performance. I'm personally getting around 35% delta form air 5090 pny oc from Suprim liquid 4090 both at 3ghz.

Sounds about right. I don’t get why so many defend the 4000 series when it does not need defending but it’s also just not as fast as a 5090. When has the price increase for the top model EVER been in line with the performance increase? In the days, I ran two 80class (whatever the second best was back then) GPUs in SLI because the price wasn’t much more than a single 90class GPU and performance was better, ignoring micro stutters etc.

 

[Pets4Ever]

The 5090 is not worth $3000 period, end of story. All of this other debate is essentially pointless beside that fact.

 

[Tyler-98-W68]

The 5090 is not worth $3000 period, end of story. All of this other debate is essentially pointless beside that fact.

That's a hell of a cop out when you can't be bothered to admit you were wrong and got holes poked through your lame attempt to justify not buying a 5090. It's ok there are 1000s of people who will enjoy them instead.

Make sure to keep quoting passmark 3d tests to further your cause

 

[Pets4Ever]

That's a hell of a cop out when you can't be bothered to admit you were wrong and got holes poked through your lame attempt to justify not buying a 5090. It's ok there are 1000s of people who will enjoy them instead.

Make sure to keep quoting passmark 3d tests to further your cause

I think you'll find I stopped mentioning them a long time ago and it's you who can't let go. It's almost like you know your GPU was a ripoff and hearing it made you butthurt.

 

[webmanau]

The 5090 is not worth $3000 period, end of story. All of this other debate is essentially pointless beside that fact.

FOR YOU it may not be. Clearly others disagree. Our MSRP is about $4k AUD and that's what I spent because it was worth it FOR ME but I simply didn't want to spend more.
Trying to get someone else to agree to your point is pointless. It's not hard and you could have worked that out yourself.

 

[Pets4Ever]

FOR YOU it may not be. Clearly others disagree. Our MSRP is about $4k AUD and that's what I spent because it was worth it FOR ME but I simply didn't want to spend more.
Trying to get someone else to agree to your point is pointless. It's not hard and you could have worked that out yourself.

That is what the thread topic is about, the decision of the individual posters to buy or not.

Nowhere did I try to advise others what to do

 

[Vayra86]

That is what the thread topic is about, the decision of the individual posters to buy or not.

Nowhere did I try to advise others what to do

I hate to break it to you, but since pretty much Ampere, buying top end GPUs has been a lucrative business. And before that too, if you were mining with them. And EVEN prior to crypto, if you were a smart buyer and timely seller. Today though, you can just float your next purchase on your old one because the GPUs barely lose value.

This also holds true in the entire high end segment. Mid range, not so much anymore. There is zero movement in the midrange, at best a VRAM upgrade, but nothing else of substance, while the price does go up. But on the higher end cards, there is also some actual progress on the silicon in raw performance.

All it takes is basic math, common sense and a starting budget. The last part is where many fall short, and the first two as well, looking at topics in general.

FOR YOU it may not be. Clearly others disagree. Our MSRP is about $4k AUD and that's what I spent because it was worth it FOR ME but I simply didn't want to spend more.
Trying to get someone else to agree to your point is pointless. It's not hard and you could have worked that out yourself.

You didn't 'want' to spend more? I'd argue you couldn't spend more

 

[webmanau]

You didn't 'want' to spend more? I'd argue you couldn't spend more

What makes you think that? I could have bought it for $6k or $5k AUD. Just didn’t want to.

 

[Vayra86]

What makes you think that? I could have bought it for $6k or $5k AUD. Just didn’t want to.

You have the top end GPU available, there's nothing above it. You can choose to pay way more than the asking price, but what sane person does that, now?
So effectively, you've just adjusted your 'acceptable price range' to what it needed to be to buy this product. You've turned that around in your statement above, but that ain't how it works right.

And that's fine, btw. But if the only 5090 was 5k, I think your initial post would have contained that number instead.

 

[Godrilla]

The 5090 is not worth $3000 period, end of story. All of this other debate is essentially pointless beside that fact.

If you want to make a difference best practice to to bring light to the new lower outliered pricing to potential buyers instead of focusing on yesterday's pricing and bashing those that bought it at a higher rate. Instead of causing resentment you'll actually be contributing to the conversation and put pressure on the retailers still holding inflated pricing. This is my core value hence why I always post a great deal even if I purchased the 5090 myself at $3200 2 months ago! It's called efficient trolling where you make a difference per comment instead of defending your position to nowhere.
Happy 4th of July everyone!

 

[webmanau]

.

And that's fine, btw. But if the only 5090 was 5k, I think your initial post would have contained that number instead.

No, it's more like this: I fell asleep during 'launch' night and in the morning the cards were sold out and immediately scalper offers popped up, as in actual retailers asking for $5k plus for entry level cards. Yes, if that had been MSRP I probably would have paid that but instead I held onto my old card until prices went back to where they were that night.
That said, if someone paid $6k to get the card immediately then that's their choice and I'm not belittling them for it.

 

[833psz]

It's all a matter of perspective.

I had a Sapphire Pulse 7900 XTX in a Bykski block flashed with Asrock Aqua vBIOS and I loved the card for almost two years. However, it left a bit to be desired on 4K (in the few games I play). Anything AI is also a little bit more difficult on AMD (or a lot more difficult if on Windows). Plus unsolvable AMD driver crashes became a HUGE annoyance. Crash to black screen and after a reboot, the dGPU is disabled in Device Manager. Super annoying. Whether it's an AMD issue or a Windows issue, it happens at varying frequency on every Radeon system I've ever used.

I find it entertaining hearing all the whining about nVidia drivers lately, because a big reason I forked out for my PNY 5090 is to get away from AMD drivers! Anyone whining about nVidia drivers is spoiled.

I got what I paid for the 7900 XTX and used it for 2 years for free... I put that money towards the 5090 and I regret nothing. Sure, it was expensive. Sure, I wish it was cheaper. But I can afford it, the math worked out for me and I couldn't be happier!

To each their own!

 

[Godrilla]

5090 fell below MSRP in UK

 

[Hyderz]

5090 fell below MSRP in UK View attachment 406743

Don’t think 5090 will sell much more in uk… living cost is so high at the moment in uk… I don’t even live there but from the news and social media a lot of the ppl living are struggling the ones that can afford I think will skip this gen or buying the 9070xt or 5070ti

 

[Dr. Dro]

Any water blocks around for it? Would like to get it on water.

Cheers.

Sorry, just saw this now (for some reason notifications not working on this thread) - in addition to the Alphacool that was recommended, Bykski also has them. $160 on Aliexpress.