New RTX 4060 leaked specs have me worried about Nvidia’s mid-range offering
Nvidia’s recent big win with its RTX 4070 Ti graphics card has me quite optimistic about the future of Nvidia’s mid-range offerings, but a new spec leak from a fairly reliable Twitter leaker has me seriously questioning the company’s thinking.
For starters, identifying a spec leak like this is important, as Nvidia hasn’t announced anything yet, and any “leaks” online need to be taken with a grain of salt.but kopite7 Kimi (opens in a new tab) It’s always been important in the past, so those specs can’t be completely undone either.
RTX 4060 still uses PG190. AD107-400-A13072FP328G GDDR6 18Gbps115W24M L2I will try to be neutral about any future leaks. 😁😁😁February 13, 2023
For starters, it looks like the RTX 4060 will use the AD107 GPU, which is a step down from the AD106 we’d expect to see in the RTX 4060, since the RTX 3060 uses the GA106 GPU. Even the RTX 3050 8GB uses a scaled down GA106 GPU, so the AD107 GPU seems to be a regression here.
What’s more, the alleged RTX 4060 specs (and some of our sloppy calculations) are nearly identical to the RTX 4060 Mobile specs for which we have official numbers to compare.
|Element||RTX 4060 Mobile (Official)||RTX 4060 Desktop (rumored)||RTX 3060 Desktop|
|Ray Tracing Core||twenty four||twenty four||28|
|base clock frequency||1,545 MHz||to be determined||1,320 MHz|
|boost clock frequency||1,890 MHz||to be determined||1,777 MHz|
|video memory||8GB GDDR6||8GB GDDR6||12GB GDDR6|
|memory clock||2,000 MHz||2,250 MHz||1,875 MHz|
|effective memory speed||16Gbps||18Gbps||15Gbps|
|memory bus width||128 bits||128 bits||192 bits|
|thermal design plan||115W||115W||170W|
As you can see, the biggest difference in the confirmed leaked specs for the RTX 4060 Mobile is the massive drop in the streaming multiprocessor count from the CUDA core count, from 3,584 for the RTX 3060 to 3,072 for the RTX 4060 — a 14.28% reduction. That’s light-years away from the best graphics card in Nvidia’s lineup, the RTX 4090, if not reduced, and half as fast as the RTX 4070 Ti, so the reduction in core count here will limit the RTX 4060’s potential.
There’s also the issue that the RTX 4060 Mobile has a slightly higher effective memory clock, which gives the RTX 4060 Desktop a 12.5% increase in memory bandwidth, but is about the same as its mobile counterpart in terms of memory.
A key unknown right now is what the final base and boost clock speeds of the RTX 4060 will be, but given that the Nvidia Lovelace architecture is a 4nm process compared to Nvidia Ampere’s 8nm process, we’d expect a base clock speed of over 2,000MHz, and if The RTX 4090 and RTX 4080 have roughly 56% faster base clocks and 48% faster boost clock generation modes than the RTX 4060, and the boost clock could exceed 2,500MHz.
Two reasons why these specs worry me
First, let me start with this: Streaming multiprocessors and CUDA core reduction are not a big deal. With 128 CUDA cores per SM, the leaked specs imply a reduction from 28 SMs in the RTX 3060 to 24 SMs in the RTX 4060. The 14.28% reduction in tensor cores and ray tracing cores is not good, but also consider these 4th generation tensor and 3rd generation ray tracing cores. These are much better than Ampere’s 3rd and 2nd generation cores respectively, so even though there are fewer of them, they will actually perform better.
But it’s the memory that worries me. 8GB of VRAM is pretty paltry at this stage, although this technically Should be a high-end 1080p graphics card, and given how well the RTX 4070 Ti handles 4K and the RTX 3060 Ti manages 1440p gaming, we expect the RTX 4060 to be a solid candidate for the best 1440p graphics card. I’m skeptical, though, given its specs.
The problem here is that 8GB of VRAM is just fine for 1080p, because the size of the texture files that can quickly fill VRAM is much smaller than at 1440p or 4K. While the RTX 3060 Ti also has 8GB of GDDR6 VRAM, it also has a wider memory bus (256 bits), giving it an effective memory bandwidth of 448.0 GB/s.
Although the VRAM pool is smaller, this is more than enough to efficiently load and process 1440p textures, while the RTX 3060 has 50% more VRAM (12GB) and a larger memory bus (192 bits), so it clocks in at 360.0GB/s memory bandwidth, giving it some decent 1440p performance as well. Meanwhile, the RTX 4060 looks like it has a much lower memory bandwidth, so larger textures at 1440p will hit the bottleneck sooner, limiting 1440p performance and pretty much limiting this card to 1080p gaming.
There’s also the issue that it uses the same base GPU as the RTX 4060 Mobile. We haven’t been able to test the RTX 4060 Mobile in person yet, but typically mobile chips are one to half a tier lower than desktop chips in terms of GPUs used (so the RTX 4090 desktop’s AD102 is an upgrade from the RTX 4090 Mobile’s AD103 ), which are also scaled-down versions of these chips.
The RTX 4060 Mobile is definitely a cut down version of the AD107, but if the RTX 4060 desktop card is also the AD107, then you definitely can’t expect the desktop version to perform much better than the mobile card. That’s not great for desktop gaming, even at 1080p.
Still, these specs could be good news for gamers, and there’s one big reason
While all of this may be bad news for gamers in terms of performance, there is a silver lining here and that is the price. If Nvidia is going to try and make an aggressively priced RTX 4060 (under $300/£300 I think), then the trade-off here is not only reasonable, but likely one that gamers are asking for.
Most gamers are still gaming at 1080p, according to Steam Hardware Survey (opens in a new tab) More and more people are upgrading to 1440p at most. While memory limitations don’t bode well for 1440p gaming, getting an affordable graphics card into the hands of gamers where they are currently is definitely something to celebrate.
A lot of gamers are still running GTX era GPUs so the RTX 4060 is still a big enough performance upgrade that I’m sure a lot of gamers won’t care how good it is Can With the AD106 GPU, in the end, that’s what matters — especially if Nvidia can price the card correctly.