Another day and another morsel from @Kopite7kimi. This time the serial leaker provides an update on core clocks for Nvidia's upcoming RTX 4070 mid-range GPU, a GPU which could one day feature among our Best Graphics Cards for Gaming. Core frequencies are now estimated to be well above anything Nvidia has ever done before, with a base clock of 2310 MHz, a rated boost clock of 2610 MHz, and a maximum boost clock of 2800 MHz via GPU Boost 4.0 — if there's enough power and temperature headroom to spare.
While this isn't the official word from Nvidia, Kopite has a good track record and the specificity of the numbers suggests inside information, perhaps from one of the graphics card vendors. It also looks like Nvidia might be taking a page or two from AMD's RDNA 2 design book.
base 2310boost 2610max >2800TGP 285WAugust 10, 2022
These core clocks are significantly faster than anything we've seen from Nvidia before, and is very reminiscent of the frequency jump we saw with the Pascal-based GTX 1000 series in comparison to the 900 series Maxwell architecture. Back then, Nvidia delivered about a 500 MHz core clock improvement across the board, but the past three generations haven't moved the needle much. That looks ready to change with Ada and TSMC 5nm.
The RTX 4070's rumored clock speeds put it well ahead of the previous generation RTX 3070, with 50% higher boost frequencies. Combine this with the new L2 cache and SM improvements rumored to be on the Ada architecture, and we could be looking at an 80% to 100% performance improvement over the RTX 3070 in some workloads — at least theoretically. Kopite has also tweeted that the RTX 4070 Ti is expected to surpass the performance of the RTX 3090 Ti, which would be in line with these figures.
| GPU Model Number | PG141-SKU331 |
| CUDA Core Count | 7680 FP32 |
| Base Clock | 2310MHz |
| Boost Clock | 2610MHz |
| Maximum GPU Boost 4.0 Clock | 2800MHz |
| Memory Configuration | 12G 21Gbps GDDR6X |
| Target Board Power (TGP) | 285W |
| 3DMark TimeSpy Extreme Score | 11,000 Points |
These frequencies will also surpass AMD's current RX 6000 series RDNA2 GPUs, which feature high clock frequencies of around 2.2–2.5 GHz, sometimes more on certain models. In real-world applications, the clock speed alone could also improve the 40-series' performance retention at lower resolutions.
With the RDNA 2 versus Ampere battles of the past two years, we've seen plenty of cases where AMD's comparable (in pricing at least) GPUs are able to surpass the performance of their Nvidia's RTX 30-series counterparts. That's largely thanks to the higher clock speeds and large Infinity Cache on AMD's RDNA 2 GPUs. This was especially true at lower resolutions, where the cache hit rates are higher.
AMD has talked about designing RDNA 2 specifically to high higher clock speeds without killing efficiency, compared to the previous RDNA architecture. All of the rumors and leaks coming out suggest that Nvidia is likely planning to take a similar approach with the Ada architecture. Ada GPUs would also be helped by higher core counts, where rumors indicate core counts could shoot up by 50–80%.
Kopite also re-confirmed the RTX 4070's power target of 285W that we reported on yesterday. The GPU's approximate Time Spy Extreme score of 11,000 points remains in place, despite the drop in power consumption. That would put the RTX 4070 about 20% ahead of the RTX 3080 performance, and roughly on par with the RTX 3090 Ti.
