Nvidia just lately scored an enormous win with its RTX 4070 Ti graphics card, which has me fairly optimistic about the way forward for Nvidia’s midrange choices to come back, however a brand new spec leak from a reasonably dependable Twitter leaker has me significantly questioning what the corporate is pondering.
For starters, it is essential to qualify spec leaks like this since Nvidia hasn’t introduced something but, and any “leaks” on-line should be taken with a grain of salt. However kopite7kimi (opens in new tab) has been pretty on level up to now, so these specs cannot be written off fully both.
RTX 4060 nonetheless makes use of PG190. AD107-400-A13072FP328G GDDR6 18Gbps115W24M L2I will attempt to stay impartial about any leaks sooner or later. 😁😁😁February 13, 2023
For starters, the RTX 4060 appears prefer it’ll be utilizing the AD107 GPU, which is a step down from the AD106 that we might count on to see within the RTX 4060, for the reason that RTX 3060 used the GA106 GPU. Even the RTX 3050 8GB used a reduce down GA106 GPU, so the AD107 GPU would look like a regression right here.
What’s extra, the purported RTX 4060 specs (in addition to a few of our back-of-the-napkin calculations) are almost equivalent to the RTX 4060 Cell specs for which we now have official numbers for comparability.
| Part | RTX 4060 Cell (Official) | RTX 4060 Desktop (Rumored) | RTX 3060 Desktop |
|---|---|---|---|
| GPU | AD107-GN21-X4 | AD107-400-A1 | GA106-300-A1 |
| CUDA cores | 3,072 | 3,072 | 3,584 |
| Tensor cores | 96 | 96 | 112 |
| Ray Tracing Cores | 24 | 24 | 28 |
| Base Clock Frequency | 1,545MHz | TBD | 1,320MHz |
| Enhance Clock Frequency | 1,890MHz | TBD | 1,777MHz |
| VRAM | 8GB GDDR6 | 8GB GDDR6 | 12GB GDDR6 |
| Reminiscence Clock | 2,000MHz | 2,250MHz | 1,875MHz |
| Efficient Reminiscence Velocity | 16Gbps | 18Gbps | 15Gbps |
| Reminiscence Bus Width | 128-bit | 128-bit | 192-bit |
| Reminiscence bandwidth | 256GB/s | 288GB/s | 360GB/s |
| TDP | 115W | 115W | 170W |
As you may see, the largest distinction within the leaked specs from the confirmed RTX 4060 Cell is the massive downgrade within the variety of streaming multiprocessors going off the CUDA core depend, from 3,584 within the RTX 3060 and three,072 within the RTX 4060 — a 14.28% lower. With out the lower, this falls light-years in need of the best graphics card in Nvidia’s lineup, the RTX 4090, and is available in at simply north half that of the RTX 4070 Ti, so the lower in core depend right here goes to restrict the RTX 4060’s potential.
There’s additionally the matter of the marginally larger efficient reminiscence clock over the RTX 4060 Cell, which provides the RTX 4060 desktop a 12.5% improve in reminiscence bandwidth, however in any other case it is roughly the identical on the reminiscence entrance as its cellular variant.
A key unknown proper now’s what the ultimate base and enhance clock speeds will probably be for the RTX 4060, however on condition that the Nvidia Lovelace structure is a 4nm course of versus Nvidia Ampere’s 8nm course of, we count on base clock speeds north of two,000MHz, with the enhance clock presumably upwards of two,500MHz if the roughly 56% quicker base clock and 48% quicker enhance clock gen-on-gen sample for the RTX 4090 and RTX 4080 maintain for the RTX 4060.
Two causes these specs have me fearful
First, let me begin with this: the lower in streaming multiprocessors and CUDA cores is not that huge of a deal. With 128 CUDA cores per SM, the leaked specs imply a lower from 28 SMs within the RTX 3060 to 24 SMs within the RTX 4060. A 14.28% drop in tensor cores and ray tracing cores is not nice, but additionally contemplate that these are fourth-gen tensor and third-gen ray tracing cores. They’re merely a lot, significantly better than Ampere’s third- and second-gen cores, respectively, so they’ll really carry out higher regardless of there being fewer of them.
What does fear me although is the reminiscence. 8GB VRAM at this stage is somewhat paltry, and whereas this technically needs to be a high-end 1080p graphics card, given how properly the RTX 4070 Ti handles 4K and the way properly the RTX 3060 Ti manages with 1440p gaming, we might hope that the RTX 4060 could be a stable candidate for the best 1440p graphics card. Given its specs although, I am uncertain.
The issue right here is that 8GB VRAM is simply advantageous for 1080p, for the reason that dimension of texture recordsdata that may rapidly replenish VRAM are a lot smaller than they’re at 1440p or 4K. And whereas the RTX 3060 Ti additionally had 8GB GDDR6 VRAM, it additionally had a a lot wider reminiscence bus (256-bit), giving it an efficient reminiscence bandwidth of 448.0 GB/s.
That is greater than sufficient to get 1440p textures effectively loaded up and processed regardless of the smaller VRAM pool, whereas the RTX 3060 had 50% extra VRAM (12GB) and a bigger reminiscence bus (192-bit), so it clocked in with a 360.0GB/s reminiscence bandwidth, additionally giving it some decent-ish 1440p efficiency. The RTX 4060, in the meantime, appears like it would have a a lot decrease reminiscence bandwidth, so bigger textures at 1440p are going to bottleneck a lot quicker, limiting 1440p efficiency and just about limiting this card to 1080p gaming.
There’s additionally the matter of it utilizing the identical base GPU because the RTX 4060 Cell. We’ve not been capable of take a look at the RTX 4060 Cell for ourselves but, however usually the cellular chips are a few tier to a tier-and-a-half decrease than their desktop counterparts by way of the GPU used (so the RTX 4090 desktop’s AD102 is a tier up from the RTX 4090 Cell’s AD103), whereas additionally being reduce down variations of these chips.
The RTX 4060 Cell is certainly a reduce down variant of the AD107, but when the RTX 4060 desktop card can also be an AD107, then you definately undoubtedly cannot count on an excessive amount of extra efficiency out of the desktop model than you’d get with the cellular GPU. That is not nice for desktop gaming, even at 1080p.
Nonetheless, there’s one huge cause these specs is likely to be excellent news for avid gamers
Whereas all of this is likely to be dangerous information for avid gamers by way of efficiency, there’s a sliver of hope right here, and that is worth. If Nvidia was going to attempt to make a RTX 4060 that’s aggressively priced (I am pondering lower than $300/£300), then the commerce off right here would not simply be justified, it would very properly be what avid gamers are calling for.
Most avid gamers are nonetheless gaming at 1080p, in line with the Steam Hardware Survey (opens in new tab) with a rising quantity upgrading to 1440p at most. Whereas the reminiscence constraints do not bode properly for 1440p gaming, getting an reasonably priced graphics card into the palms of avid gamers the place they’re at present at would undoubtedly be one thing to have a good time.
Lots of avid gamers are additionally nonetheless working GTX period GPUs, so the efficiency improve with an RTX 4060 will nonetheless be substantial sufficient that I am certain lots of avid gamers aren’t going to care about how significantly better it may be with an AD106 GPU, and ultimately, that is actually all that ought to matter — particularly if Nvidia can get the pricing on this card proper.
