NVIDIA GeForce GTX TITAN X Overclocking & Best Playable Settings

by Rob Williams on April 2, 2015 in Graphics & Displays

NVIDIA’s GeForce GTX TITAN X is the world’s latest fastest single GPU solution. But does that mean we should just leave it alone, and ignore its true potential? We could, but that’d be boring – so join us as we take a look at the TITAN X from an overclocking perspective, and provide some ‘Best Playable’ results for good measure.

When I posted my look at NVIDIA’s GeForce GTX TITAN X at its launch last month, there were a couple of tests I had to skip over due to a lack of time, one of which was overclocking. As it is now, though, I have to admit that I’m glad I waited, as it’s allowed me to spend a lot more time to find that “perfect” overclock.

Alright – so a “perfect” overclock doesn’t actually exist. “Great” ones do, though. I’m not talking about those sorts of overclocks that break records, but 100% stable overclocks that prove genuinely worthwhile – those that can add at least 10% to the performance.

Can the TITAN X achieve that level of greatness? On account of the fact that this ~2,000 word article exists, you can probably already guess the answer!

TITAN X Overclock: Stable, Not Stirred

In advance of TITAN X’s launch, NVIDIA told us that a top-end clock speed of about 1,400MHz could be a common sight. At this point, I’m not sure if that’s true or not, as results are all over the map from fellow reviewers – some show as low as a 100MHz gain, while others show upwards of 250MHz. Given my experience with TITAN X thus far, I feel confident in saying that a +175MHz boost should be possible on most cards – an overclock that doesn’t require a voltage boost.

All told, I ended up reaching a top stable overclock of +210MHz on the GPU and +400MHz on the memory. When inching near the +200MHz mark, I didn’t require a voltage boost, but I did at +200MHz, and subsequently +210MHz. I am sure I could have pumped even more voltage to breach +210MHz, but +24mV seemed like a good place to stop given the temperatures I was seeing after a long stress.

+210MHz core, +400MHz memory, +24mV, 60% fan speed

To get a peak temperature and make sure this overclock was in fact stable, I looped 3DMark’s Fire Strike Ultra (4K) test for about 100 minutes. The AIDA64 overall results can be seen here, while the entire log can be seen here.

That log highlights a couple of notable points. For starters, like other recent GeForce cards, the clock speed you’ll get will be largely dependent on the temperatures the card’s dealing with. This even applies at stock speeds – the card might not hit its full GPU Boost potential if the card runs hot (but it should never dip below the normal 1,000MHz clock speed).

What that means is this: If you overclock TITAN X (or other high-end NVIDIA GPUs), you’ll probably want to increase the fan speed to compensate for the added heat. While you might find that an overclock is 100% stable, it doesn’t necessarily mean that it’s running as fast as it can because of the heat.

I have found that a good “sweet spot” for fan speed on TITAN X is 60 or 65%. That might not seem like much, but it is in fact much louder than the card ever gets (based on personal testing) during intense stress sessions at stock speeds. That said, it’s not at all what I’d call “loud”; it’s just noticeable. If you’re listening to game audio at all, the sound of the card with its fan pegged to 60% isn’t going to be a problem.

If you don’t care about noise, at all, you could increase the fan speed higher and potentially earn better stable overclocks. I stopped where I did because at above 65%, the fan became louder than I’d expect most people would to deal with, and if a fan is run at top speed all of the time, it could cut its lifespan short. If you’re the type of gamer that upgrades often, that might not matter too much. All I can say is that it took some good effort to get the overclock I did, but I am sure it can be pushed even further if you’re willing to add more stress to the card.

Putting The Overclock To Work

When it comes to overclocking, I don’t produce results unless I consider the overclock to be 100% stable. For that to be the case, it has to pass a long stress test (in this case, 100 minutes of 3DMark Fire Strike 4K), and also prove flawless through all regular benchmarking. That proved to be the case here.

To provide what I feel is the most accurate apples-to-apples representation of reference vs. overclocked performance, I forced the fan speed to 60% for both configurations. As I mentioned earlier, even if a card like TITAN X is running at stock speeds, its GPU Boost clocks can throttle a little bit if heat becomes an issue. That means that one TITAN X might appear to run a little bit faster (or slower) than someone else’s, because their PCs and ambient room temperatures are different.

That means a couple of things: We can assume that the TITAN X’s “stock” results below represent its best possible performance; it also means that if I did leave the fan to its normal auto speeds, the results below would work even more in the overclock’s favor, given that configuration’s fan speed boost.

Enough of that; let’s get on with the performance results, shall we?

NVIDIA GeForce GTX TITAN X - 3DMark Fire Strike Overclocked Results

This simple graph helps show just how much more beneficial a card like TITAN X is on the high-end versus the low-end. At 1080p, we see an 11% increase, while at 1440p and 4K, it bursts to 15% and 16%, respectively. Will that carry on over to the real-world tests?

Yes, it will:

	Stock		Overclock		Gain (Avg)
4K (3840×2160)	Min	Avg	Min	Avg	+ / –
Assassin’s Creed IV	43	52	48	57	+10%
Battlefield 4	31	44	38	50	+14%
Crysis 3	35	51	42	58	+14%
Dying Light	37	46	42	53	+15%
GRID 2	58	66	69	78	+18%
Metro Last Light	27	39	32	45	+15%
Shadow of Mordor	30	36	35	41	+15%
Sleeping Dogs	43	54	50	62	+15%
Splinter Cell: Blacklist	48	58	56	66	+14%
The Crew	26	34	30	39	+15%

I’ve mentioned in the past that I don’t overclock my own GPUs because I don’t consider the gains to be that worth it, but this is one of those times when an exception would be made. The overclock proved to add at least 10% to the framerate, with the average being 15%. Given the clock boost was 20%, this shouldn’t come as a surprise, but it’s still nice to see.

What’s not seen in the table above but should be emphasized once again is that with this overclock, the TITAN X ran hotter than the reference configuration. Even with that locked 60% fan speed, the card peaked at 88°C – a 5°C gain over the peak 83°C I’ve seen at stock.

Because I love repeating myself, that means you should either scale back the overclock a little bit (at least to the point where you don’t have to add voltage), or increase the fan speed even further. Regardless of the overclock, your target temperature should be not much more than 80°C.

As with the “Best Playable” sections in our regular graphics card reviews, I went through most of the games benchmarked above (the exception is Metro Last Light) and modified their settings to get as close to 60 FPS as possible on average. While the TITAN X is an undeniably fast card, 4K is likewise an undeniably mammoth resolution. You can’t expect to max everything out – overclock or not – but the settings you will be left with will still look fantastic.

Onward we go.