Thursday, September 26th 2024
NVIDIA GeForce RTX 5090 and RTX 5080 Specifications Surface, Showing Larger SKU Segmentation
Thanks to the renowned NVIDIA hardware leaker kopite7Kimi on X, we are getting information about the final versions of NVIDIA's first upcoming wave of GeForce RTX 50 series "Blackwell" graphics cards. The two leaked GPUs are the GeForce RTX 5090 and RTX 5080, which now feature a more significant gap between xx80 and xx90 SKUs. For starters, we have the highest-end GeForce RTX 5090. NVIDIA has decided to use the GB202-300-A1 die and enabled 21,760 FP32 CUDA cores on this top-end model. Accompanying the massive 170 SM GPU configuration, the RTX 5090 has 32 GB of GDDR7 memory on a 512-bit bus, with each GDDR7 die running at 28 Gbps. This translates to 1,568 GB/s memory bandwidth. All of this is confined to a 600 W TGP.
When it comes to the GeForce RTX 5080, NVIDIA has decided to further separate its xx80 and xx90 SKUs. The RTX 5080 has 10,752 FP32 CUDA cores paired with 16 GB of GDDR7 memory on a 256-bit bus. With GDDR7 running at 28 Gbps, the memory bandwidth is also halved at 784 GB/s. This SKU uses a GB203-400-A1 die, which is designed to run within a 400 W TGP power envelope. For reference, the RTX 4090 has 68% more CUDA cores than the RTX 4080. The rumored RTX 5090 has around 102% more CUDA cores than the rumored RTX 5080, which means that NVIDIA is separating its top SKUs even more. We are curious to see at what price point NVIDIA places its upcoming GPUs so that we can compare generational updates and the difference between xx80 and xx90 models and their widened gaps.
Sources:
kopite7kimi (RTX 5090), kopite7kimi (RTX 5080)
When it comes to the GeForce RTX 5080, NVIDIA has decided to further separate its xx80 and xx90 SKUs. The RTX 5080 has 10,752 FP32 CUDA cores paired with 16 GB of GDDR7 memory on a 256-bit bus. With GDDR7 running at 28 Gbps, the memory bandwidth is also halved at 784 GB/s. This SKU uses a GB203-400-A1 die, which is designed to run within a 400 W TGP power envelope. For reference, the RTX 4090 has 68% more CUDA cores than the RTX 4080. The rumored RTX 5090 has around 102% more CUDA cores than the rumored RTX 5080, which means that NVIDIA is separating its top SKUs even more. We are curious to see at what price point NVIDIA places its upcoming GPUs so that we can compare generational updates and the difference between xx80 and xx90 models and their widened gaps.
181 Comments on NVIDIA GeForce RTX 5090 and RTX 5080 Specifications Surface, Showing Larger SKU Segmentation
Zero hope of going out of 8GB hell on the mid range again.
Since AMD wont compete at this performance range, Ti version may not even come. But who knows, maybe nvidia gets bored along the way. However, I think it would be a mistake for nvidia to put a product between these two. With this performance gap they can force people to buy the expensive card instead of going with xx80 class.
170 SM and 512-bit @ 32Gbps GDDR7 bus seems too good to be true knowing that they were going for 162 SM and 448-bit @ 28Gbps not that long ago... but if priced around $1800 it will sell really well!
if the 5080 manages to beat the 4090, then AMD may compete with it. Otherwise, no.
Ironically that sounds like something you'd hear the strippers yell out in Sleeping Dogs along with something like $20 dollah hollah!! That game is satirical gift that keeps giving...and the intro theme will continue to bang and love you long time.
5080 Super with ~90 SM and slightly better VRAM speed is a likely continuation (provided the 84 SM 5080 is a real thing).
But if the leak is true the gap between the 5080 -> 5090 will be even bigger than 4080 -> 4090 ! ~2x vs ~68%
Also the 5090 having a 512-bit bus would mean 2x more Memory Bandwidth vs 5080 too! Whereas the 4090 "only" has 40% more bandwidth than the 4080, that's a big difference knowing that 4K and RT/PT are definitely Memory Bandwidth bound.
And in case you are wondering, it only cost me $718.9743 bazillion, just a drop in my proverbial (cash) bucket.....
I'll make back at least 100x that selling them off to scalpers, hoarders and other assorted ilk of the interwebz (but not you good folks here though) :D
Yea yea I know ya'll are sooooo disappointed, but hey, money talks & bullshit walks, so just deal with it....
1st come 1st served as they say !
I also wouldn't be surprised at all with a 3000USD price tag.
Nope. No way. Nvidia didn't become hottest share in market by selling gaming paraphernalia.
I expect Nvidia to clearly release RTX 5090, and to lesser extent RTX 5080 to home AI acceleration crowd. Data center revenue is up more than 150% compared to last year, and now represents vast majority of Nvidia's income, thanks to AI craze. "Gaming" cards will no longer be just "gaming", as is clear from the change of RTX logo that now includes this:
And I don't believe they have included "Powering Advanced AI" just because DLSS is technically what we already call AI (machine learned upscaling), or that you could accelerate upcoming NPCs with ChatGPT - like abilities of conversation.
These cards will be offered as "cheap" accelerators for smaller neural network servers and as tools for artists and content creators.
And their prices will reflect that they are tools for generating income, not playthings. I expect a 100% price hike, just like in the good old cryptomadness days.
But don't despair, lower end RTX 40x0 will still be offered alongside Blackwell for all those who still like to play with their computer.
That's said I don't see AI enthusiasts buying the 5080 if the specs are right. 16GB of VRAM is too skimpy and it's core count is skimpy. Those people would be better off scooping up 4090s with 24GB of VRAM. 16GB of VRAM can barely fit the FLUX model in it but with 24GB you can do FLUX plus at least one LORA.
Eh, if that doesn't change we ain't seeing full RT, not now, not ever.
electrolytesnewest generation AI acceleration magical cores.RDNA 4 high-end chips were having issues so they cancelled all of them and are going to sell Mainstream GPUs instead to make more money, but RDNA 5 is supposed to fix all those issues and be a real competitor to RTX 5090.
I don't actually own a hat though so that's not much of a promise but I just can't see it happening. If they have already designed midrange GPUs with a midrange memory bus width then what would hold them back?
I understand the issues with a big chip and competing there but to stop at a 5070 competitor just doesn't make sense to me.
I have no doubt it will exceed my 4090 and my 3090.
I am just sad it'll cost $2000 on top of the cost of the new 15900k and the new motherboard for it.