Alleged leak shows GeForce RTX 4090 GPU at 3.0 GHz with 20K points in TimeSpy Extreme

assuming 4090 consume 450W, same as 3090Ti, that means ~75% uplift in efficiency

at 350W TDP the 4090 will be ~75% faster than 3090, about the same efficiency jump as Maxwell to Pascal

 

Highly doubtful imo.

450Watts is extreme, air cooling at 55°C is impossible. Maybe fans ran at 100% but who does that at home?
You want a quiet GPU with highest possible efficiency.

Also generally speaking about the 4090 I think 450 watts still going to be a problem. This card will be satisfying only under liquid cooling environments, or for miners who don't care about noise.

 

It's a benchmark run... obviously he is running with 100% fanspeed.

I can do the same with my gpu at 480 watts... let it sit with 100% fanspeed for a couple of minutes before starting the run, so it gets down to like 25c, then it tops off at 60c towards the end of the run.

And no, 450 watts wont be a problem, if you have a sufficient case and sufficient airflow, which you ought to have if you are forking out for a 3090 / 4090...

 

those CPU temps look...odd....

 

Especially odd, considering it's not a cpu being tested...

You do realize that it's purely a gpu benchmark right ? So a water cooled cpu will stay in the 30's...

 

Then the two of us have a different definition about a quiet environment.
A sufficient airflow within the case doesn't affect the temps of the GPU that much, it is however required to get rid of the excess heat so it doesn't get stuck within the case thus to avoid overheat.

 

You can have a very good AND quiet airflow in your case, if you got a big case with plenty of quality noctua fans running at like 1000 rpm.

And obviously the gpu will need to have a big cooler aswell, which it will have, like the 3090 ti has a quite beefed up cooler vs the 3090.

 

I agree with @Dragam1337 here as I've been on this quest to find a case/case fan combination paired with GPU undervolt to achieve the perfect balance between temps and performance. I've also opted not to go for CPU AIO due to more noise sources. One important factor here is definitely having a large case and also go for 140mm fans.

 

Though only go for 140mm if there are full openings for 140mm - even if you can mount 140mm, sometimes a part of the fan will be blocked off, which will cause noise, and obviously also interfere with airflow - in that case it's actually better to go with a 120mm fan

 

As someone with a custom loop, thats nonsensical bullshit.

You will not get idle temps on the cpu with it not being pushed with a 350w-450w pushing that much heat in a custom loop. That coolant is gonna heat up alot, and you can see a 5-10c uplight in idle temps.

 

That's 3,000 more points than my 3090. Not bad. With water and an OC add another 2-3k points.

 

3090 supposed to be in 10k range so thats 2x.

 

That was it running in power save mode. It will do 10,500+

 

100% this dude just doesnt realize that these are the timespy EXTREME results, not timespy...

This is mine btw

https://www.3dmark.com/spy/26622034

So yeah, nearly 2x.

 

Timespy extreme score is 4k resolution, normal timespy isn't directly comparable.

 

If the performance gap between the 4000 and 3000 series is this big, then Nvidia needs to drop the prices of current cards by 50%!...

 

Or up the prices 50% for the 40 series.

 

Or lock a bunch of new value-add features to the 4000 series and only increase prices by 25% lol