THe RTX 5000 series is gonna be a flop and heres why.

[Guldo]

If the leaks are true nvidia is coming out with a 16 pin power connector wich is devestating news for anyone that prefers nvidia over amd because that means these cards are going to be too big to fit in anyones case and over heating will also become an issue.

Now i can understand why amd is making no high end cards next gen because why even bother.

This definitly has to do with diminishing returns of shrinking the node as they arent getting the same uplift they used to so they have to make up the slack by making the cards bigger.

[lsdninja]

Początkowo opublikowane przez Kenshiro:

need to find some way to cool the card without increasing the size of the heatsink.

Integrated AIOs like that AMD Fury card a while back XD

[Phénomènes Mystiques]

GPU cooling blocks.

[Guldo]

Początkowo opublikowane przez Wynters:

GPU cooling blocks. :citizensurprised:

the way things are going liquid nitrogen cooling may become standard lol. possibly a new type of AIO that uses it.

[PopinFRESH]

You are absolutely clueless. Smaller process notes doesn’t just equate to “more power”. Generally smaller process nodes result in less loss / power waste for a number of reasons. GPU power changes are a design choice balancing power:performance:die size. Hence why the same die on a smaller process node (e.g. a die shrink) generally use less power.

The actual graphics card for the 30 and 40 series is significantly smaller than earlier cards going back to the GTX 200 generation. My liquid cooled 3090 is tiny in comparison to my 2080, 1070, 970, 780 Ti etc. There are still a plethora of options across the product segments that don’t have large heatsinks on them.

“Is there a way to ditch the big heatsinks…” yeah, it is called a custom water cooled loop.

[Guldo]

Początkowo opublikowane przez PopinFRESH:

You are absolutely clueless. Smaller process notes doesn’t just equate to “more power”. Generally smaller process nodes result in less loss / power waste for a number of reasons. GPU power changes are a design choice balancing power:performance:die size. Hence why the same die on a smaller process node (e.g. a die shrink) generally use less power.

The actual graphics card for the 30 and 40 series is significantly smaller than earlier cards going back to the GTX 200 generation. My liquid cooled 3090 is tiny in comparison to my 2080, 1070, 970, 780 Ti etc. There are still a plethora of options across the product segments that don’t have large heatsinks on them.

“Is there a way to ditch the big heatsinks…” yeah, it is called a custom water cooled loop.

not everyone build a custom water cooled loop but i will take the L on everything else.

[KalGimpa]

Początkowo opublikowane przez PopinFRESH:

You are absolutely clueless. Smaller process notes doesn’t just equate to “more power”. Generally smaller process nodes result in less loss / power waste for a number of reasons. GPU power changes are a design choice balancing power:performance:die size. Hence why the same die on a smaller process node (e.g. a die shrink) generally use less power.

The actual graphics card for the 30 and 40 series is significantly smaller than earlier cards going back to the GTX 200 generation. My liquid cooled 3090 is tiny in comparison to my 2080, 1070, 970, 780 Ti etc. There are still a plethora of options across the product segments that don’t have large heatsinks on them.

“Is there a way to ditch the big heatsinks…” yeah, it is called a custom water cooled loop.

i would not have a problem with it if it was just the cards getting a little longer or wider

it the amount of heat that has to be disposed of

at a certain point it gets comical to me that they are so proud of how small they get

but we need a giant air cooler or an entire water system to keep them cool enough to use

these things are covering three fliipin slots on my motherboard

useless now because of the gpu

we need either a redesign of the mb

which i kind of would like to see with only nvme

or something needs to be compromised to bring these back to a reasonable size to case ratio

[Guldo]

Początkowo opublikowane przez KalCuey:

Początkowo opublikowane przez PopinFRESH:

You are absolutely clueless. Smaller process notes doesn’t just equate to “more power”. Generally smaller process nodes result in less loss / power waste for a number of reasons. GPU power changes are a design choice balancing power:performance:die size. Hence why the same die on a smaller process node (e.g. a die shrink) generally use less power.

The actual graphics card for the 30 and 40 series is significantly smaller than earlier cards going back to the GTX 200 generation. My liquid cooled 3090 is tiny in comparison to my 2080, 1070, 970, 780 Ti etc. There are still a plethora of options across the product segments that don’t have large heatsinks on them.

“Is there a way to ditch the big heatsinks…” yeah, it is called a custom water cooled loop.

i would not have a problem with it if it was just the cards getting a little longer or wider

it the amount of heat that has to be disposed of

at a certain point it gets comical to me that they are so proud of how small they get

but we need a giant air cooler or an entire water system to keep them cool enough to use

these things are covering three fliipin slots on my motherboard

useless now because of the gpu

we need either a redesign of the mb

which i kind of would like to see with only nvme

or something needs to be compromised to bring these back to a reasonable size to case ratio

expecting everyone to build a custom water cooler is expecting too much but i think hes right on everything else. i noticed he carefuly made sure he didnt say they dont generate more heat because that would have been a lie lol. if they didnt they wouldnt be making the heatsinks bigger every generation lol.

[PopinFRESH]

Heat is power. So if you are talking about “smaller process node makes more heat” then no you are wrong. Again, the power is a design choice; e.g. Thermal Design Power (TDP). Two different cards that have a continuous power draw of 300W and are on similarly efficient process nodes are going to have largely the same heat rejection load.

[Guldo]

Początkowo opublikowane przez PopinFRESH:

Heat is power. So if you are talking about “smaller process node makes more heat” then no you are wrong. Again, the power is a design choice; e.g. Thermal Design Power (TDP). Two different cards that have a continuous power draw of 300W and are on similarly efficient process nodes are going to have largely the same heat rejection load.

you already explained that bub and no i didnt say smaller proccess node = more heat. its a fact the 4090 generates more heat than the 3090 or why have a bigger heat sink.

The 4090 gets 1.5% hotter than the 3090 despite the 4090 having a bigger heat sink.

Despite its outstanding performance, the RTX 4090 maintains a slight edge in thermal efficiency, running just 1.5% hotter than the RTX 3090 in Shadow of the Tomb Raider, with temperatures of 64°C and 63°C, respectively. Even with a 96% usage rate, the RTX 4090 falls below its 450W TDP at 415.5W.

[spicy little monkey]

Początkowo opublikowane przez Kenshiro:

the way things are going liquid nitrogen cooling may become standard lol. possibly a new type of AIO that uses it.

Written by someone who doesn't know how ln2 works.

[Guydodge]

what does the 16pin power connector have to do with the size of the card you buy a
angled cable problem solved..as far as your case being to small or overheating
issues thats on you build smarter.proper fan placement/direction and cooler options
make or break your build and as far as card temps go the 4090 is far superior
to the 3090 in every way including tempatures.i own both (evga 3090 ultra and
a asus tuf 4090 oc) its a much cooler card and needs a much less aggressive fan curve.
this post sounds like a chicken little youtube vid strictly made for hits.

they have been making large cases for many years now bought the GT501 in
2020 and it can handle the 4090 no problem and the 5090 should i upgrade
larger case= better air flow (as long as its not glass front) cards have been
getting bigger every generation since ive been building.build smarter !!!

[Phénomènes Mystiques]

AI integration will cause more troubles than not...

[r.linder]

As PopinFresh said, clueless.

The 16-pin connector already exists, they're just updating it with the Gen6 version which goes up to 675 watts, the current one allows for up to 600 and the 4090 doesn't even make full use of it, there's no reason to because the performance gain it gets from being overclocked to that point is marginal at best, the extra headroom is more likely for stability rather than to raise TGP.

The RTX 4090 can be undervolted to around 270W, a massive ~40% reduction in power consumption, but it only loses around ~8% of its performance in doing so. Most of the 40 series is pretty efficient like that, the 4060 was ~20% faster than the 3060 but had a power consumption reduction of around 30%. The 4060 Ti was less impressive with ~10% more performance but around 20% less power consumption. The 4070 had a 30~35% gain over the 3070 with 10% less power, and the 4080 was 30~40% faster than the 3080 for basically the same power consumption.

Most of the 40 series has been more power efficient compared to their older counterparts from the 30 series. 50 series can easily follow the same trend, or maintain the same power draw like the 4080 has.

Most cases have no issues fitting these GPUs either way, and if you can afford a 4090 or 5090, you can afford a bigger case.

[PopinFRESH]

Początkowo opublikowane przez r.linder:

...
Most cases have no issues fitting these GPUs either way, and if you can afford a 4090 or 5090, you can afford a bigger case.

Or even if you want a smaller case...

Lian-Li (DAN) A4-H20 mini-ITX that fits 3-slot full length GPUs like the 4090...
Fractal Design Terra mini that fits 3-slot full length GPUs like the 4090...
Cooler Master NCORE that fits 3-slot full length GPUs like the 4090...

Also, it is still unclear if NVIDIA will actually be moving to an MCM design with Blackwell or not. There has been conflicting information and rumors. If they do that will certainly have a significant impact on power characteristics.

[Sigma957]

They should do away with square power connection points. Make them round like the microphone connectors with locking clips so there is no unintentional slippage or loose linkage.