Tucked inside NVIDIA’s announcement of their spring refresh of their cell GPU lineup, the corporate included a brand new low-end cell half, the GeForce GTX 1650 GDDR6. Exactly because it says on the tin, this was a model of the corporate’s GTX 1650 accelerator, besides with newer GDDR6 as a substitute of the GDDR5 it launched with. Now, in considered one of NVIDIA’s extra poorly stored secrets and techniques, their desktop product stack is getting a model of the cardboard as nicely.
While not a launch (as NVIDIA likes to border it), the desktop GTX 1650 GDDR6 has none the much less lastly turn out to be an official product this previous Friday, with companions unveiling their playing cards and NVIDIA including the specs to their web site. Sitting alongside the present GDDR5 model, the GDDR6 model is meant to be a parallel, usually equal SKU. As NVIDIA makes the transition from GDDR5 to GDDR6 on the backside fringe of their product lineup, the up to date card will get entry to sooner reminiscence, however apparently the GPU clockspeeds are additionally tapered again a bit.
| NVIDIA GeForce Specification Comparison | ||||||
| GTX 1660 | GTX 1650 Super | GTX 1650 (G6) | GTX 1650 (G5) | |||
| CUDA Cores | 1408 | 1280 | 896 | 896 | ||
| ROPs | 48 | 32 | 32 | 32 | ||
| Core Clock | 1530MHz | 1530MHz | 1410MHz | 1485MHz | ||
| Boost Clock | 1785MHz | 1725MHz | 1590MHz | 1665MHz | ||
| Memory Clock | 8Gbps GDDR5 | 12Gbps GDDR6 | 12Gbps GDDR6 | 8Gbps GDDR5 | ||
| Memory Bus Width | 192-bit | 128-bit | 128-bit | 128-bit | ||
| VRAM | 6GB | 4GB | 4GB | 4GB | ||
| Single Precision Perf. | 5 TFLOPS | 4.Four TFLOPS | 2.85 TFLOPS | three TFLOPS | ||
| TGP | 120W | 100W | 75W | 75W | ||
| GPU | TU116 (284 mm2) |
TU116 (284 mm2) |
TU117 (200 mm2) |
TU117 (200 mm2) |
||
| Transistor Count | 6.6B | 6.6B | 4.7B | 4.7B | ||
| Architecture | Turing | Turing | Turing | Turing | ||
| Manufacturing Process | TSMC 12nm “FFN” | TSMC 12nm “FFN” | TSMC 12nm “FFN” | TSMC 12nm “FFN” | ||
| Launch Date | 03/14/2019 | 11/22/2019 | 04/03/2020 | 04/23/2019 | ||
| Launch Price | $219 | $159 | ~$149 | $149 | ||
By the numbers, the brand new GDDR6 model is essentially the identical because the GDDR5 model. Both are 75W playing cards primarily based on NVIDIA’s entry-level Turing TU117 GPU. However the GDDR6 model of the cardboard each features some and loses some within the course of. NVIDIA swaps out the GDDR5 for newer GDDR6 – and thereby lastly confirming that TU117 is GDDR6-capable – nonetheless the playing cards additionally take a slight clockspeed nerf. As a consequence the GDDR6 model of the cardboard has a whopping 50% extra reminiscence bandwidth – bringing it to 192GB/sec – however 5% decrease GPUs clocks and throughput.
In discussing the matter with NVIDIA, we had been instructed that the GPU clockspeed change was to equalize efficiency and energy consumption between the 2 components. Which is sensible to a level – the GTX 1650 is a very particular half in NVIDIA’s lineup because it’s the quickest card they provide that may be powered completely by a PCIe slot, which is to say it will possibly’t have a TDP over 75 Watts. So with the GDDR5 model already near that restrict, if the change to GDDR6 reminiscence drives up energy consumption in any respect (be it the reminiscence or the GPU’s reminiscence controllers), then one thing else needs to be dialed again to compensate.
Meanwhile, equalizing efficiency is one thing of a secondary purpose on this state of affairs, particularly due to the efficiency of GDDR6 reminiscence. NVIDIA would not intend for the GDDR6 model of the GTX 1650 to be its personal product; the subsequent card up after the GTX 1650 stays the GTX 1650 Super. But given what we’ve seen on different Turing components such because the GTX 1660 sequence, the place an analogous change netted an additional 10% in efficiency, I might count on the GTX 1650 to see the identical type of modest advantages from the…
