This month NVIDIA unveiled three new GeForce RTX 3000 series graphics solutions for gaming and mobile workstation laptops and promised we’d see more than 70 different laptops featuring the new GPUs.
But if you want to know what kind of performance to expect from the first mobile GPUs featuring NVIDIA “Ampere” graphics, it turns out you need to know a lot more than the model number, because PC makers can choose from a range of different configurations for each model in order to prioritize power consumption, performance, or some mix of the two.
According to details obtained by German news site ComputerBase, there are more than 30 different RTX 3000 GPU options available to PC makers.
In previous years you could get a rough idea of whether you were getting a low-power or a high-performance NVIDIA mobile GPU by looking to see if it was a Max-Q (energy efficient for thinner and lighter laptops) or Max-P (highest performance for laptops with heavy duty cooling) version of the processor.
This year NVIDIA isn’t using those labels anymore. And the move sort of makes sense, because there aren’t just two versions of each GPU anymore. According to ComputerBase, there are at least:
- 11 versions of the RTX 3060 mobile GPU with 60 watt to 115 watt TGP (total graphics power)
- 6 versions of the RTX 3070 mobile GPU with 80 watt to 125 watt TGP
- 11 versions of the RTX 3080 mobile GPU with 80 watt to 150 watt TGP
Unfortunately that makes it very difficult to know exactly which version of the GPU you’re getting unless the PC maker displays the exact configuration including the TGP and base an turbo clock speeds on the box.
For a sense of the range of possibilities, you can check out the ComputerBase article for a complete list, but here’s an overview of the differences between the most energy-efficient and highest-performance versions of each RTX 3000 mobile GPU:
Model | TGP | Base clock | Turbo clock | CUDA Cores | Memory interface | Storage throughput |
RTX 3060 | 80W | 817 MHz | 1.282 GHz | 3,840 | 192 bits | 12 Gbps |
RTX 3060 | 115W | 1.387 GHz | 1.78 GHz | 3,840 | 192 bits | 12 Gbps |
RTX 3070 | 80W | 780 MHz | 1.29 GHz | 5,120 | 256 bits | 14 Gbps |
RTX 3070 | 125W | 1.215 GHz | 1.62 GHz | 5,120 | 256 bits | 14 Gbps |
RTX 3080 | 80W | 780 MHz | 1.245 MHz | 6,144 | 256 bits | 12 Gbps |
RTX 3080 | 150W | 1.35 GHz | 1.71 GHz | 61,44 | 256 bits | 14 Gbps |
via Tom’s Hardware
OK, so we release many many flavors and hide their names so the customers never know what they’re eating. Now every item on the menu just says “Food”, and all the prices are different. We’re gonna get sooo rich!
baseless speculation with no evidence whatsoever:
Maybe they’ve been having bad yields, and so all the variants are partially working chips with different parts of the chip disabled depending on what broke during fabrication.
Perhaps not as far as “partially working chips”, but it definitely could be a matter of their quality “binning”. Almost all chip manufacturers will put different quality levels of chips into different grade bins, and use them on different products based on what level of quality is required to achieve the voltage, temperature, clockspeed needed for a given product.
If you consider that Nvidia previous had a product line that spanned from the very low end GT 1010, all the way to the Titan, and now today their GPU lineup doesn’t extend nearly as low-end as it previously did (possibly due to improvements in Intel and AMD’s integrated GPUs). So perhaps this is Nvidia’s strategy to find uses for their lower-bin chips that have been displaced by their shrinking lineup.
Personally, I just wish they came up with a better marketing strategy. It seems unintuitive that the name “RTX 3080” could possibly represent two enormously different levels of performance. I’m certain that some people are going to call this misleading. Hopefully these specs are something that laptop makers will be upfront about.