Analyzing Temperatures & Frequencies Over Time
Analyzing Temperatures & Frequencies Over Time
Modern graphics card designs make use of several advanced hardware and software facing algorithms in an effort to hit an optimal balance between performance, acoustics, voltage, power and heat output. Traditionally this leads to maximized clock speeds within a given set of parameters. Conversely, if one of those last two metrics (those being heat and power consumption) steps into the equation in a negative manner it is quite likely that voltages and resulting core clocks will be reduced to insure the GPU remains within design specifications. We’ve seen this happen quite aggressively on some AMD cards while NVIDIA’s reference cards also tend to fluctuate their frequencies. To be clear, this is a feature by design rather than a problem in most situations.
In many cases clock speeds won’t be touched until the card in question reaches a preset temperature, whereupon the software and onboard hardware will work in tandem to carefully regulate other areas such as fan speeds and voltages to insure maximum frequency output without an overly loud fan. Since this algorithm typically doesn’t kick into full force in the first few minutes of gaming, the “true” performance of many graphics cards won’t be realized through a typical 1-3 minute benchmarking run. Hence why we use a 5-minute warm up period before all of our benchmarks.
So let’s start things off with the overall temperatures observed by both GPU-Z and EVGA’s Precision over the course of a 5-minute static benchmark from Hellblade where the character is literally standing still in one place. . As you can see, the RTX 2080 doesn’t go above 72 degrees Celsius which seems to be what NVIDIA’s default peak is for this card. The Boost algorithm won’t allow it to go beyond that relatively low temperature. This means you could probably get some additional clock speed headroom by simply increasing the Temperature Limit in overclocking software.
The RTX 2080 Ti on the other hand shows us something interesting too since its temperatures are allowed to peak at a point that’s 5 degrees higher and that’s where it remains. Basically both of these are pretty cool running cards but it really does look like NVIDIA gave their higher end GPU a bit more leeway. That’s likely because it it needs that headroom to achieve expected performance numbers.
Something did strike me as interesting though. NVIDIA could have easily squeezed some more performance out of the RTX 2080 by simply allowing it to hit a higher temperature like the RTX 2080 Ti. Maybe this was done to insure a bit more separation between the two cards or it could have been for another reason altogether different. Only NVIDIA knows and they’re not telling.
And here you can see what kind of effect NVIDIA’s GPU Boost 3.0 technology has on clock speeds as temperatures increase. Rather than pushing fan speeds up so the RTX series gets overly loud, they fluctuate core frequencies to balance temperature and power consumption. This also goes to show why it is VERY important to either perform longer benchmark runs or insure the cards have a warm-up period before testing each game.
While the RTX 2080 remains around 1900MHz for the first thirty seconds or so, by the end of the 5 minute test that gets reduced to 1845MHz. This is literally right in line with NVIDIA’s Boost specification. The RTX 2080 Ti gets a bit more shaved off by going from 1785MHz to 1680MHz after 5 minutes, it too remains right near the 1635MHz Founders Edition spec. Not bad at all but a bit different from previous generations where we saw the cards exceeding the stated Boost speeds by (sometimes) pretty significant margins.
For those of you wondering about those odd spikes on the RTX 2080 Ti, those are areas where the fans increase their rotational speed by a completely unnoticeable ~200RPMs. That effectively lowers the temperatures by a bit and the core is able to subsequently boost to a higher frequency.
All things considered, I was expecting the TU102 within NVIDIA’s RTX 2080 Ti to exhibit a bit more fluctuation than it did. We’ve already seen other hot-running cards like AMD’s Vega64 and Fury series literally cut 300 to 400MHz off their initial frequencies. So count me moderately impressed.