Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
Can I just say how excited I am to have our 'proper' graphics card power consumption equipment and software finally in place? Trusting what the AMD or Nvidia drivers tell you in regards to power (e.g., via GPU-Z) is just asking a bit too much, I think. With the proper equipment, we can also definitively state how much power the various graphics cards are actually using. Reporting 'GPU-only' power (AMD) is at best deceptive, especially when the competition (Nvidia) reports total board power use.
But let's hit the charts — I've reworked everything so that we can hopefully present up to 12 GPUs in a chart while still keeping things legible. AMD GPUs are dotted lines, Nvidia GPUs are solid lines, and the card we're reviewing — the EVGA GTX 1650 GDDR6 SC Ultra in this case — is indicated with an extra thick line. Let's start with a look at gaming, using Metro Exodus, and then we'll look at the worst-case results with FurMark.
EVGA GTX 1650 GDDR6: Metro Exodus Power, Temps, Clocks, and Fan Speed
Looking at the average power use chart, the EVGA GTX 1650 GDDR6 ended up tied with the 1650 GDDR5 model at 81W. The EVGA card sips power compared to mainstream GPUs, but it's not substantially different from the GTX 1050 Ti. And by that, we mean that power use increased right alongside performance. The previous generation GTX 1050 Ti averaged 58W in the same test, so Nvidia created a more powerful chip without massive efficiency changes. That's perhaps expected given the transition from 16nm to 12nm TSMC nodes didn't change as much as the numbers would suggest (i.e., it wasn't a 25% smaller process).
Overall, ignoring the outliers like the RX 590 and RX 570, the GTX 1650 delivered about 18% lower performance than AMD's RX 5500 XT 4GB while using 35% (44W) less power. And that is why we're using Powenetics again rather than GPU-Z's power figures. We're not showing the charts for the PCIe slot or PEG connector, but the EVGA card easily stayed in spec. It drew peak power of 45.8W from the x16 slot and 39.3W peak power from the PEG connector. None of the cards shown here exceeded spec while running the Metro Exodus benchmark, if you're wondering.
The EVGA GTX 1650 GDDR6 GPU core clocks came in way above the listed 1710 MHz boost clock — 200 MHz, higher in fact. Throughout the gaming portion of the benchmark (omitting the loading at the end of each run), the EVGA card pushed 1916 MHz clocks on average. It might run at lower clocks in some other games, and the Gigabyte GTX 1650 card and Zotac GTX 1660 card do post slightly higher average clocks, but this is about as good as it gets for Nvidia's Turing GPUs. AMD's RX 5500 XT 4GB averaged 1836 MHz, just a hair below the 'maximum' 1845 MHz rated boost clock.
GPU temperature has a direct inverse correlation with fan speed, which we'll look at next — the higher your fan speed, the lower your temperature. Heatsink size and design are also factors, as well as the power use of the chip. Since the EVGA GTX 1650 GDDR6 uses a relatively low power chip, even with the high clock speeds it runs at, temperatures peaked at 62.8C and averaged just 59.9C throughout the gaming test. Basically, anything below 70C is more than fine on modern chips, and the EVGA cooler is certainly doing its job.
Last up before we kick things into high gear with FurMark, fan speeds are also fine. You can see in the charts that some of the 'hotter' GPUs in the temperature chart correspond to cards that also have lower fan speeds. Basically, manufacturers typically design the fan speed curve around temperature rather than the reverse, so if EVGA, in this instance, tries to keep temperatures below 60C, there will be a ramp-up in fan speed to do that. (Note: the 1650 Super didn't log fan speeds properly — ignore the "0" result.)
Interestingly, though EVGA doesn't mention this in its product information, the GTX 1650 GDDR6 can shut its fan off. When the GPU hits 45C or higher, or perhaps under any significant GPU load, the fan starts to spin at 40% or about 1250 RPM. Maximum fan speed during our gaming test hit 1850 RPM. It's higher than many of the other GPUs in our chart, but the dual 85mm fans are not particularly loud.
EVGA GTX 1650 GDDR6: FurMark Power, Temps, Clocks, and Fan Speed
As usual, FurMark shakes things up quite a bit. Average power use during the stress test climbed to 92.6W for the EVGA GTX 1650 GDDR6 and peaked at 93.3W, a modest 11.5W jump. That might seem big, but looking at other GPUs, it's not too bad. AMD's RX 5500 XT cards, for example, used about 45W more power in FurMark. This is not a 'normal' GPU workload by any means, and while it might theoretically be possible for a game to push the GPUs to similar power levels, in practice it never happens — there's a lot more going on in a game than just rendering a fuzzy donut as fast as possible.
Along with higher power use, clock speeds were lower on all the GPUs with FurMark. The EVGA GTX 1650 GDDR5 averaged 1755 MHz, down about 160 MHz from the Metro test. And yeah, I realize the line is hard to find in the bunched up data, but it was pretty stable throughout the test period. It's interesting that the GTX 1650 GDDR5 model maintained higher clocks, which is likely from its larger cooler that also keeps temperatures down.
Temperatures were far more stable as well, partly because FurMark is a constant workload, unlike games. The EVGA GTX 1650 GDDR6 averaged 63.2C during the test run and peaked at 65C, which is only about 3C higher than in Metro Exodus. As with the power charts, AMD's GPUs didn't fare as well. The 5500 XT 4GB was 8C higher in FurMark, though all of the GPUs are still under 80C and shouldn't have any issues.
As before, fan speeds influence temperatures, and all the GPUs have to ratchet up the RPMs a bit for FurMark. Also, you should just ignore the fan speeds on the old GTX 980 and 1070 cards — both have been around the proverbial block and I had to manually adjust the fan speed curves to keep the GPUs stable during testing. It's about time to put the 980 out to pasture, I think.
MORE: Best Graphics Cards
MORE: Desktop GPU Performance Hierarchy Table
MORE: All Graphics Content
Intel reportedly demands all board partners implement Intel Default (Baseline) Profile by May 31 — company hopes to fix issues with some Core i9 chips
'Digital Twins' research gains $285 million in CHIPS ACT funding — virtual chip tech could revolutionize semiconductor manufacturing
Explosive HBM demand fueling an expected 20% increase in DDR5 memory pricing — demand for AI GPUs drives production cuts for standard PC memory
