Uncategorized

NVIDIA RTX 5090 Specs: Over 24000 Cores, 32Gbps GDDR7 Memory, and 1536 GB/s Bandwidth

The GeForce RTX 5090 will be NVIDIA’s next gaming flagship, doubling down on memory, cache, and cores. Multiple sources have shared their findings on the Blackwell top gun, which we’ll discuss in this post. RedGamingTech claims that the RTX 5090 will feature 204 SMs for a total of 26,112 cores. On the memory side, we’re looking at 32Gbps GDDR7 memory paired with a 384-bit bus. The L2 cache will be consolidated to 96MB as well. Plans of using a 512-bit bus seem to have been shelved.

According to @kopite7kimi, the next-gen Blackwell “GB202” gaming die will have a 12 (GPCs) x 8 (TPCs) configuration. This would result in a massive GPU featuring 12 GPCs (Graphics Processing Clusters), each with 8 TPCs (Texture Processing Clusters). A TPC comprises 2 SMs (Streaming Multi-Processors), each of which contains 128 FP32 cores.
The above numbers net a core count of 24,576 (192 SMs) for the fully enabled GB202 die. Like its predecessor, the GeForce RTX 5090 should contain a few disabled SMs to ensure proper yields. Optimistically, we can expect a core count of 24,046 or less.

The GeForce RTX 5090 with 32Gbps GDDR7 memory and a 384-bit bus would produce an external memory bandwidth of 1,536 GB/s, up from 1,008 GB/s on the RTX 4090. A 50% increase in memory throughput backed by a ~60% gain in shader density would easily net generational uplifts of up to 2x or more. It’s still too early to finalize a number, but it’d be fair to say that the RTX 5090 should deliver the same kind of upgrade as its predecessor.

Graphics Card NameNVIDIA GeForce RTX 4090NVIDIA GeForce RTX 4080NVIDIA GeForce RTX 4070 TiNVIDIA GeForce RTX 4070NVIDIA GeForce RTX 4060 TiNVIDIA GeForce RTX 4060
DieAD102AD103AD104AD104AD106AD107
FP32 Cores1638497287680588843523072
TMUs/ROPs512/176320/112240/80184/64136/48TBD
Tensor/RT Cores512/128304/76240/60184/46136/34TBD
L2 Cache72 MB64 MB48 MB36 MB32 MB24 MB
Base Clock2230 MHz2210 MHz2310 MHz1920 MHz2310 MHz1830 MHz
Boost Clock2520 MHz2510 MHz2610 MHz2475 MHz2535 MHz2460 MHz
VRAM24 GB GDDR6X16 GB GDDR6X12 GB GDDR6X12 GB GDDR6X8-16 GB GDDR68 GB GDDR6
Bus Width384-bit256-bit192-bit192-bit128-bit128-bit
Memory Clock21.0 Gbps23.0 Gbps21.0 Gbps21.0 Gbps18.0 Gbps17.0 Gbps
Bandwidth1008 GB/s736 GB/s504 GB/s504 GB/s288 GB/s
(554 GB/s Effective)
272 GB/s
(453 GB/s Effective)
TBP450W320W285W200W160/165W115W
MSRP$1599$1199$799$599$399-$499$299
Launch12th Oct 202216th Nov 20225th Jan 202313th April 202324th May/ Jul 2023Jul 2023

An RTX 5080 featuring a bus width of 192-bit and 32Gbps GDDR7 memory would net bandwidth of 768GB/s, a small upgrade from the 717GB/s offered by the RTX 4080. Neither the RTX 4080 nor the 4080 Super utilize a 192-bit bus. With Ada, NVIDIA is using larger L2 cache buffers to improve on-die hit rates, and when they’re not being stingy with DRAM, it mostly works.

The GeForce RTX 5090 Founders Edition will allegedly use the same heatsink as the recently leaked RTX 4090 Ti. This massive thermal solution is four slots wide with four fans and a complex, side-mounted PCB. The final design will likely be toned down as there’s no need for such a large shroud.

We already know that the RTX 4090 runs well on cheap, aftermarket coolers. There’s no reason for the RTX 5090 to accommodate a similar-sized heatsink even though it’s expected to be fabbed on TSMC’s 3nm-class process technology.

Source: Kimi.

Areej Syed

Processors, PC gaming, and the past. I have been writing about computer hardware for over seven years with more than 5000 published articles. Started off during engineering college and haven't stopped since. Mass Effect, Dragon Age, Divinity, Torment, Baldur's Gate and so much more... Contact: areejs12@hardwaretimes.com.
Back to top button