Overclock SLI 970s?
Overclock SLI 970s?
Initially, I run at 1440p and my system handles this well. The only game where I had to lower settings is Fallout 4. I experienced drops into the 40s in some areas, but adjusting a few graphics options brought me back up to 60 FPS.
Performance has stayed solid for my PC. Fallout 4 was the sole exception and it resolved easily. I still enjoy tweaking my setup, so I’m thinking about overclocking my older 970s. Here are my current details.
i5-4690k at 4.4GHz
8GB 1600MHz RAM
Gigabyte GA-Z97X-UD3H-BK
2X EVGA ACX 2.0 SC GTX 970
EVGA G2 Supernova 850W PSU
I had some worries about overclocking. To be honest, I’ve never overclocked a GPU before, though I’m comfortable with Intel and AMD CPUs. I think SLI builds are trickier because you need both GPUs to run at the same speed, making it harder to identify the source of instability. That said, I’m not certain yet.
Second, others on this forum mentioned that a typical GPU overclock only adds 2-5 FPS. Does that match what I’ve heard? If so, it might not be worth the extra heat, noise, and power consumption.
Finally, is my power supply strong enough for an overclocked CPU plus two overclocked 970s? So far everything’s working fine, but I’m curious about the increased power needs when pushing GPUs to their limits.
I have a P2 1200 in my rig, 4790k which uses significantly more power than the 4690k, and two 780tis units—each with an 8pin power connector. I’m confident my system would work on your PSU, provided I don’t go overclocking. Will it make a difference? Only one sure answer: try it out. You could see a 10-20% increase in clock speed and improved RAM performance too. I have no doubt you’ll perform better than expected.
Nvidia experience is the ideal starting point. Configure the system to use just one card and adjust it. Continue refining until instability occurs, saving your configurations along the way. When instability arises, revert to the last stable settings. Restore experience to SLI and test both cards at the most stable configuration of card 1. If stability fails again, reduce the settings further.
if it's meant for synthetic testing with a clear advantage, but for gaming it's completely pointless, I'm still disappointed. I don't think I'll ever return to it honestly—it's frustrating that just one Radeon 4870 could ruin my 8800GT setup. I don't know how this is supposed to work anymore, but it really hurts.
The procedure appears straightforward, but I’m uncertain if my power supply can handle an overclocked CPU along with two overclocked GPUs. I also doubt whether this will deliver substantial performance gains. Some threads suggest only a modest improvement of 2-5 FPS in demanding games, which isn’t enough to justify the extra heat and noise.
I have a P2 1200 in my rig, 4790k which uses significantly more power than the 4690k, and two 780tis units—each with an 8pin power connector. I’m confident my system would work with your PSU, provided I don’t go overclocking. Will it make a difference? Only one sure answer: try it out. You might see a 10-20% increase in clock speed and improved RAM performance too. I have no doubt you’ll perform better than expected.
if it's meant for synthetic testing with a clear benefit, but if it's intended for gaming it's completely pointless. I'm still disappointed in SLI and I doubt I'll ever return to it honestly—it's frustrating that just one Radeon 4870 could ruin my 8800GT setup. I don't know what the current situation is, but it really hurts.
SLI isn't entirely useless. I've been using SLI 970s for about six months now, and it has significantly improved my setup compared to having a single 970. Benchmarks indicate that SLI 970s match the performance of a 980Ti. From my experience, a single 970 does have some issues with certain games on a 1440p display, but running two in SLI gives me around 60 FPS across all titles.
The drawbacks of SLI are often overstated on this forum. People constantly complain about its performance, but my own experience is quite different. It might have been a nightmare on older 8800GT cards, but that seems to be changing now.
Of course, if you're purchasing new today, a 980Ti would be a better option since it costs the same as two 970s and offers a single-card solution. If you already own a single 970 and want a performance boost, adding a second one for SLI is a much more economical choice than spending $650 on a 980Ti.
Most AAA titles support SLI from the start, while games like Fallout 4 usually get it within a week or two. Nvidia has been improving its Game Ready drivers for SLI support recently. Arkham Knight remains the only major game from this year that I played without SLI support—it's a mess on PC.
I have a P2 1200 in my rig, 4790k (it uses a lot more power than your 4690k) and two 780tis units (each has several power connectors). I'm sure my setup would work with your PSU, as long as I don't overclock. Will it make a difference? Only one way to find out—just try it.
it's nice when the Sli support improves, because I had it be pretty sparse and most games didn't support it. I think Sli was more for synthetic testing, but that's changed its value. Still, I'm a bit unsure about setting up Sli, especially since heat is an issue for me—my home doesn't have an air conditioner, and my GTX 970 has a reference Nvidia heat that gets quite hot (around 80°C).
I experimented with this overclocking project for a few hours last night and struggled to achieve a consistent GPU boost. I used Unigine Heaven at 2560x1440 with ultra settings, extreme tessellation, and 4X AA for stress testing and benchmarking.
I adhered to the overclocking instructions and noted that I didn’t raise the voltage as suggested in the guide; instead, I capped it at its maximum of 110%.
I observed no artifacting even when VRAM clocks reached high levels. I halted at +400MHz due to rising temperatures.
When adjusting the core clock, I never encountered artifacts, but anything above +100MHz led to display crashes during stress tests. The screen would freeze after a few minutes of benchmarking, though the system remained operational and didn’t shut down.
At normal speeds, my FPS was 75.8 with a score of 1909.
With +100MHz core and +400MHz VRAM, FPS improved to 83.0 and score to 2090, but the display still crashed after the benchmark.
At +89MHz core and +300MHz VRAM, FPS dropped to 81.9 and score to 1455.5, with the screen going black shortly after the test.
I understood that lowering the numbers further wouldn’t significantly boost performance without adding excessive heat or noise.
I also realized I might have been able to stabilize the build or push overclocking higher if I’d started increasing voltage, though my top GPU was already nearing its 78°C limit during testing without a voltage boost. I suspect raising the voltage would cause the card to throttle below the 80°C threshold.