Nvidia to Use AD107 Die for RTX 4060 Mobile and Desktop

For as long as we can remember, the desktop-to-mobile GPU conversion math was pretty simple: The big chip goes in the desktop, and the smaller version goes in the laptop. That’s what Nvidia has been doing for years and is still doing. For example, its mobile RTX 4090 chip is the RTX 4080 desktop SKU, aka AD103. It’s not the full RTX 4090 AD102 die for thermal reasons, not to mention it would be ridiculously expensive and massive. Regardless, that’s been the pattern previously. Now Nvidia looks to be abandoning that strategy for its upcoming x60 series GPUs. For the first time we can recall, it will use the same die in both mobile and desktop versions of the RTX 4060. This could be good news for laptop gamers but bad news for desktop owners.

News of Nvidia’s plans comes from a reliable source, albeit on Twitter, so the usual dose of salt is necessary. Notorious leaker @kopite7kimi is reporting the upcoming desktop RTX 4060 will use the PG190 PCB with AD107. We already know the specs of this chip because Nvidia “launched” it as the RTX 4060 mobile at CES several weeks ago. Its specs include 3,072 CUDA cores, 96 texture mapping units, and 32 ROPs. It also has 96 tensor cores and 24 ray tracing cores as well. Its memory configuration is 8GB of GDDR6 running at 16Gb/s on a 128-bit bus. Its TDP is 115W, so we imagine the desktop version will be quite small.

RTX 4060 still uses PG190.
AD107-400-A1
3072FP32
8G GDDR6 18Gbps
115W
24M L2

I will try to remain neutral about any leaks in the future. 😁😁😁

— kopite7kimi (@kopite7kimi) February 13, 2023

We will have to wait for benchmarks, obviously, but these specs are a departure from previous x60 GPUs. In almost every regard, the RTX 4060 has less of everything compared with the RTX 3060 that it’s replacing. It has fewer CUDA cores, a narrower memory bus along with less memory bandwidth, and a much lower TDP. These are different architectures, so it’s not an apples-to-apples comparison. However, there have traditionally been certain specs that are locked down for specific categories of GPUs. For example, both the RTX 3060 and 2060 have 192-bit memory buses. The RTX 4060’s is 128-bit instead. There’s also a version of the RTX 3060 with 12GB of memory, so the 4060’s 8GB is another downgrade.

x60 GPUs have always been the most popular Nvidia cards due to their price-to-performance ratio. The big boy x90 and x80 cards grab headlines, but the x60 cards are what sell in volume. On the latest Steam hardware survey, x60 cards occupy seven out of the top 10 results. Of course, a big part of their popularity is their pricing, but it remains to be seen what the RTX 4060 will cost. For this generation, Nvidia has bumped up pricing across the board. If it comes out at $499 or something ridiculous, that’ll have a huge impact on its sales.

For now, Nvidia has yet to launch the RTX 4060. However, the mobile version is about to be released in laptops. It’s unclear when Nvidia will launch the desktop version and for how much, but we’ll find out soon. It’ll be interesting to see how it fares against its predecessors, especially the RTX 3070.

Now read:

• First Nvidia GeForce RTX 4090 Mobile Benchmarks Show Stunning Improvements
• CES 2023: Nvidia Launches RTX 4070 Ti, Ada Lovelace Mobile, GeForce Now With RTX 4080
• Chrome and Edge Browsers to Get Nvidia RTX AI Video Upscaling (Updated)

sourse ExtremeTechExtremeTech https://ift.tt/a8W4GqR