Will you buy a RTX 5090?

[JustBenching]

I'd still like to see how it behave when power limited to 300-350 watts, cause no way in hell im going to be using a 700w GPU.

 

[mtosev]

View attachment 382810
Well now, let's keep this safely tucked away until the wife is out for a few hours

Now I really need to see what I have lying around, that I can use to bypass the old GPU's waterblock for a few weeks...

Will you post a few pics of the card?

 

[Bomby569]

Will you post a few pics of the card?

nudes

 

[RJ]

I really don't want to but I guess I'll have to. All I wanted was a 4090 and I would be set for the foreseeable future, if this wasn't a paper launch and the 5080 wasn't as weak, it would have worked out perfectly. I don't upgrade too often, still on my 6yr old 2080ti.

But the 4090's aren't being made anymore and the ones in circulation are pretty steeply priced for 2yr old GPU's with expired warranties and of questionable status in terms of durability.

I'm getting a decent tax return and at my performance review at work I earned the full bonus that will pay out in March, a 5090 price won't be a problem that way but I am definitely not some wide eyed nvidia fan excited for it. This will be the last nvidia product I buy. Even 5 years down the road, if I don't lose interest in gaming, which seems like a good possibility, I'll buy intel or amd, no matter how poorly they perform compared to nvidia, out of spite.
If it hurts them so much to make geforce gpu's, they should stop making geforce gpu's, just go all out on big AI gpus and don't come back.

 

[Dragam1337]

I really don't want to but I guess I'll have to. All I wanted was a 4090 and I would be set for the foreseeable future, if this wasn't a paper launch and the 5080 wasn't as weak, it would have worked out perfectly. I don't upgrade too often, still on my 6yr old 2080ti.

But the 4090's aren't being made anymore and the ones in circulation are pretty steeply priced for 2yr old GPU's with expired warranties and of questionable status in terms of durability.

I'm getting a decent tax return and at my performance review at work I earned the full bonus that will pay out in March, a 5090 price won't be a problem that way but I am definitely not some wide eyed nvidia fan excited for it. This will be the last nvidia product I buy. Even 5 years down the road, if I don't lose interest in gaming, which seems like a good possibility, I'll buy intel or amd, no matter how poorly they perform compared to nvidia, out of spite.
If it hurts them so much to make geforce gpu's, they should stop making geforce gpu's, just go all out on big AI gpus and don't come back.

I reckon alot of people were feeling like you, just wanting 4090 performance with the 5080, for that 1000 usd price... but looks like it will be quite a few years before that happens !

 

[Dr. Dro]

I'd still like to see how it behave when power limited to 300-350 watts, cause no way in hell im going to be using a 700w GPU.

I'll let you know whenever I get mine...

 

[T4C Fantasy]

poll has been added

 

[3valatzy]

This generation (that is mostly upcoming) is not a good purchase variant.
If you can postpone the decision to buy, postpone it and wait.

Let the scalpers run by Nvidia "buy" the little stock it has "released".

 

[Dr. Dro]

I honestly don't get it. It's like people are actively finding reasons to be outraged. First the card is way too expensive, "ngreedia is a ripoff!", then "there's no stock!", so, you actually meant to buy it all along? It's like y'all want to keep the pitchfork lit no matter the cost. You'll groan that there aren't products for everyone day one like this is something completely unprecedented. Just as an example, Nintendo had 1,5 million preorders for the Super Famicom in Japan back in 1990, and only 300,000 units available at launch.

Orders take time to fulfill. Not everyone gets their unit on the very first day. Or even the very first month. They've been manufacturing cards for a while now. Everyone will get theirs. It takes a bit. Just be patient. Start getting antsy if you did not get the card that you ordered on launch day by March.

 

[xorbe]

I decided to hang tight, and check again in August. Maybe they will hit us with a 24GB 450W 384-bit 5080 Ti, which I would find more interesting than 16GB 5080 or 600W 5090. Realistically it's probably just going to be a long wait to the 6000 series for me.

 

[Dr. Dro]

I decided to hang tight, and check again in August. Maybe they will hit us with a 24GB 450W 384-bit 5080 Ti, which I would find more interesting than 16GB 5080 or 600W 5090. Realistically it's probably just going to be a long wait to the 6000 series for me.

Any 5080 Ti would realistically be a further cutdown GB202 as the 5080's GB203 configuration is already maxed and juiced. So you can probably expect power consumption to be quite high, at least in the same league. It was the only way they could bring any marketable numbers against the 4080 Super, seems like the improvement in Blackwell SM's raw throughput against Ada is relatively small, I reckon they realized that the performance gap between Ada and RDNA 3 was huge as it is, Blackwell widens it even more, with the 5090 at ~3x the average 4K RT performance of 7900 XTX and with AMD's upcoming card openly admitted to being a regression throughout, they opted to improve their already excellent architecture instead of coming up with something completely new.

Despite the fact that it is twice as powerful as the 4080 Super and the first GPU to reach and exceed the 100 TFLOPS mark, I see it as an iterative release, it basically refines and perfects Ada Lovelace. Most of the improvements on RTX 50 series GPUs will come from optimizing the existing floorplan, either through hardware or software. The new release 570 driver has major improvements in video playback performance and RTX Video Super Resolution, they slashed the power consumption for decoding and upscaling and improved performance at the same time, for example. This improvement also applies to earlier generation cards, 50 series will receive things like hardware flip metering and the ability to execute all sorts of instructions throughout all cores simultaneously, which should greatly improve throughput and maximize hardware efficiency.

 

[chrcoluk]

Surprised at poll results.

 

[Dragam1337]

Surprised at poll results.

Mostly people interested in it who are gonna vote.

I honestly don't get it. It's like people are actively finding reasons to be outraged. First the card is way too expensive, "ngreedia is a ripoff!", then "there's no stock!", so, you actually meant to buy it all along? It's like y'all want to keep the pitchfork lit no matter the cost. You'll groan that there aren't products for everyone day one like this is something completely unprecedented. Just as an example, Nintendo had 1,5 million preorders for the Super Famicom in Japan back in 1990, and only 300,000 units available at launch.

Orders take time to fulfill. Not everyone gets their unit on the very first day. Or even the very first month. They've been manufacturing cards for a while now. Everyone will get theirs. It takes a bit. Just be patient. Start getting antsy if you did not get the card that you ordered on launch day by March.

The 5090 is the biggest paper launch ever though, that is just a fact. Extremely limited stock.

 

[Outback Bronze]

The 5090 is the biggest paper launch ever though, that is just a fact. Extremely limited stock.

Yeah, this is a big problem atm. You think this has purposely been done? If so, this feels like some kind of money laundering just to push prices up. Its BS.

 

[Dragam1337]

Yeah, this is a big problem atm. You think this has purposely been done? If so, this feels like some kind of money laundering just to push prices up. Its BS.

I think that nvidia just don't give a rats arse about gamers, and sold all chips they could to ai customers / datacenters.

 

[Outback Bronze]

I think that nvidia just don't give a rats arse about gamers, and sold all chips they could to ai customers / datacenters.

If so, it's a sad state of affairs.

 

[Dragam1337]

If so, it's a sad state of affairs.

From their perspective it makes perfect sense, as the margins are alot higher in the other sectors.
But they are going to leave the backdoor open to competitors in the gaming sector, much like intel did in cpu space with their stagnation 10 years ago.

 

[Klemc]

I agree, that's the actual business... and gamers are fans so will buy anyway.

 

[RJ]

Why take a piece of wafer and sell it for $2000 when you can take roughly the same piece of wafer, add a few more components and sell it for $30000?
I get it. But then go all out, just sell it all to AI, leave consumer gpu market and don't let the door hit you on the way out. We'll manage without you.

 

[freeagent]

I agree, that's the actual business... and gamers are fans so will buy anyway.

Because we made them who they are, not data centers or AI.

 

[oxrufiioxo]

From their perspective it makes perfect sense, as the margins are alot higher in the other sectors.
But they are going to leave the backdoor open to competitors in the gaming sector, much like intel did in cpu space with their stagnation 10 years ago.

I mean if all their competitor is going to do is offer similar crap at a slight discount Nvidia has nothing to worry about, until they consistently prove me wrong in this regard I'm not holding my breath for that. Also intel stagnated because they hit a wall and didn't want to switch to tsmc which happened anyways they mismanaged their foundry business. Nvidia is stagnating not because they can't do better it's because they've chosen to give us a 70 class like product for 1000 usd.... The minute a competitor even shows half a pulse they will stomp on it with a refresh.

The other big difference with CPU is if you choose intel/amd you really aren't locked out of anything the same games play the same way on every cpu they just spit out more or less frames with a gpu if you leave Nvidia you are also leaving their ecosystem of features which most feel are superior to their competitors. So any company that want's to seriously compete is going to have to beat them at everything and I just don't see that happening.

Because we made them who they are, not data centers or AI.

Let's be real though if you have a product and you can sell it to one customer for 10000 and another for 2000 what are you going to do...... Anyone who picks 2000 is full of $h!+

The reality is it isn't even that close they are selling B100/200 as fast as they can make them.

 

[Dragam1337]

I mean if all their competitor is going to do is offer similar crap at a slight discount Nvidia has nothing to worry about, until they consistently prove me wrong in this regard I'm not holding my breath for that. Also intel stagnated because they hit a wall and didn't want to switch to tsmc which happened anyways they mismanaged their foundry business. Nvidia is stagnating not because they can't do better it's because they've chosen to give us a 70 class like product for 1000 usd.... The minute a competitor even shows half a pulse they will stomp on it with a refresh.

The other big difference with CPU is if you choose intel/amd you really aren't locked out of anything the same games play the same way on every cpu they just spit out more or less frames with a gpu if you leave Nvidia you are also leaving their ecosystem of features which most feel are superior to their competitors. So any company that want's to seriously compete is going to have to beat them at everything and I just don't see that happening.



Let's be real though if you have a product and you can sell it to one customer for 10000 and another for 2000 what are you going to do...... Anyone who picks 2000 is full of $h!+

The reality is it isn't even that close they are selling B100/200 as fast as they can make them.

Everything you wrote is true - another component is that nvidia are somewhat successful in making an apple-esque ecosystem, where you NEED nvidia for the juicy features.

That's why all the nvidia proprietary solutions were always going to be bad for consumers in the longterm - and nvidia are obviously banking on those features securing sales, even if the hardware being sold is VERY lacklustre.

 

[Outback Bronze]

From their perspective it makes perfect sense, as the margins are alot higher in the other sectors.

It would be nice if they could separate AI GPU's for commercial and gaming GPU's in different market segments with prices reflecting.

But the problem here is that DLSS uses those AI Tensor cores that gamers so called need I suppose...

 

[oxrufiioxo]

It would be nice if they could separate AI GPU's for commercial and gaming GPU's in different market segments with prices reflecting.

But the problem here is that DLSS uses those AI Tensor cores that gamers so called need I suppose...

They have done that in the past but it seems cheaper for them to focus on one architecture although they were using Hopper for AI and ADA for gaming and honestly it didn't make much difference


Although as bad as the 4080 was it was still worlds better than the 5080 at launch it at least beat the 3090 by a decent margin while being 300 usd cheaper....

 

[Dragam1337]

It would be nice if they could separate AI GPU's for commercial and gaming GPU's in different market segments with prices reflecting.

But the problem here is that DLSS uses those AI Tensor cores that gamers so called need I suppose...

Indeed ! They should have done that back in the day with turing, before they started making up uses for the rt and tensor cores - cores that would otherwise have been wasted on gaming gpus. Dlss could easily have been made from the start to run on raster cores - fsr does that after all.

They have done that in the past but it seems cheaper for them to focus on one architecture although they were using Hopper for AI and ADA for gaming and honestly it didn't make much difference


Although as bad as the 4080 was it was still worlds better than the 5080 at launch it at least beat the 3090 by a decent margin while being 300 usd cheaper....

I don't doubt that the development is cheaper that way, but the gaming gpu's could have been alot better for gaming, had that been the focus on the development, rather than AI. Im thinking we could potentially have had gpu's with the same gaming performance as a 5070, but at a 150w budget, like with 1070 prior to all the rt and tensor fluff being added.