Nvidia GeForce RTX 4090 Reviews

[eastmen]

NVIDIA produced over 100,000 RTX 4090 units thus far

Since the NVIDIA GeForce, RTX 4090 graphics card came out about a month ago, at least 100,000 of them have been sold, according to data....

www.guru3d.com

That is great numbers for Nvidia I think.

Personally way to expensive for me (my wife would kill me ) but in all honesty its not super expensive in terms of how hobby's go. Traxxas just released a new truggy that is $1k for the car and then you are looking at another $500 or so for battery and charger.

I just bought $150 shoes for running and I go through 4 pairs a year or so. So I don't think the price is crazy for someone who is really into the hobby.

 

[Scott_Arm]

Unfortunately one thing that the gpu shortages during the mining craze proved is that gamers will spend way way way above msrp on gpus. Nvidia and AMD would be insane not to increase their prices. They can only produce so many of a given gpu, and if they can sell them for hundreds more, they should take the money. I can see price cutting down the road, but for now $$$$$$$$$$.

 

[eastmen]

Jayztwocents has a 4090 from evga apparently

 

[davis.anthony]

I can imagine they had some 4090 designs in the testing phase before they decided to part ways.

As a GPU collector I would love to get my hands on one!

The GPU power connector on the back end of the PCB also needs to make a comeback!

 

[trinibwoy]

More accurately it can rasterise 33B tris/s and transform/cull 198B/s

Was the transform and cull rate per TPC ever confirmed at 1/clk? I recall back in Fermi/Kepler days it was 0.33 triangles/clk.

 

[Qesa]

Was the transform and cull rate per TPC ever confirmed at 1/clk? I recall back in Fermi/Kepler days it was 0.33 triangles/clk.

It's varied depending on the generation of polymorph engine. I was using 0.5/clock but that's as of turing since I haven't seen microbenchmarks of anything newer.

 

[trinibwoy]

It's varied depending on the generation of polymorph engine. I was using 0.5/clock but that's as of turing since I haven't seen microbenchmarks of anything newer.

Got it. That would only be ~88B tris/sec on the 4090 though (64 TPCs).

 

[TopSpoiler]

1-7x NVIDIA GeForce RTX 4090 GPU Scaling

NVIDIA's GeForce RTX 4090 24GB is an incredible GPU, but sometimes one card just isn't enough. How much more performance can you get if you used seven cards rather than just one?

www.pugetsystems.com

 

[Remij]

So Nvidia still really hasn't said anything about the connector issue, huh?

Mine seems to be fine and I've really been putting this card through its paces, overclocked since day 1, and mine seems to be good... but I'd like a bit of assurance and closure on this issue.

 

[Jawed]

Fully Connected 16-Pin Connector On The NVIDIA GeForce RTX 4090 Ends Up Melting Too

A fully connected 16-Pin '12VHPWR' connector on the NVIDIA GeForce RTX 4090 has melted & we also get our first Founders Edition case.

wccftech.com

Nice summary (which I renumbered due to sloppy publication):

1 - Fully plugged-in connector can melt
2 - Partially plugged-in connector can melt
3 - ATX 3.0 PSU can also make connector melt
4 - Not an issue solely related to 12VHPWR adapter (16-pin to 4 x 8-pin)
5 - The cable can come loose after some time by itself (due to bends & resistance)
6 - PCI-SIG planned revision for December 6th
7 - NVIDIA working on plug revision (rumor for now)

 

[digitalwanderer]

So basically we know that something seems to be making some of them melt some of the time?

 

[Remij]

Stupid Nvidia. I mean seriously, how have they not made a comment on this yet?

They can still be testing and not knowing what's going on... but still COMMUNICATE THAT to us...

The longer this goes on... the worse it will be for them.

 

[Cyan]

the performance is outworldly

 

[DegustatoR]

NVIDIA confirms there are two manufacturers of 12VHPWR power adapter cables - VideoCardz.com

There are two manufacturers of NVIDIA’s 12VHPWR connector, and you should probably avoid one Igor’sLAB is working tirelessly to get to the bottom of the 12VHPWR ‘melting’ problem reported by many users already. The latest development is a discovery that there are two manufacturers of the NVIDIA...

videocardz.com

 

[BRiT]

 

[Clukos]

So absolutely nothing in the end, 0.1% failure rate...

However, the design of these connectors needs to be improved to provide better feedback.

 

[eastmen]

So absolutely nothing in the end, 0.1% failure rate...

However, the design of these connectors needs to be improved to provide better feedback.

It's 0.1% failure rate so far. That is the real problem. If it is the connectors not fully socketing then the longer these cables are sitting in cases with pressure or even gravity working on it , they can fail at higher rates.

They really need to figure out some other way to make these plugs. These cables are so big and bulky. I know usb c can't push the same amounts of power but they seem to be rising and it would be great if something close to usb size could be used

 

[Kyyla]

According to physics the trivial way to decrease heating of the conductors without changing the physical size is to reduce the current by increasesing voltage. That is what USB did.

 

[Kaotik]

According to physics the trivial way to decrease heating of the conductors without changing the physical size is to reduce the current by increasesing voltage. That is what USB did.

With the push for everything to use 12V in PC I don't see putting GPUs (pretty much only thing consuming such ridiculous amounts of power) on 48V happening

 

[Kyyla]

With the push for everything to use 12V in PC I don't see putting GPUs (pretty much only thing consuming such ridiculous amounts of power) on 48V happening

Yet my entire house's electrical connection is 75 amps and this little trinket is supposed to withstand 50.