Did lowering my GPU power cause my tests to get worse because of an over-boosted CPU?
Did lowering my GPU power cause my tests to get worse because of an over-boosted CPU?
Hey everyone, I just built a computer for my nephew and have been testing it out. To make sure everything works right, I've been tweaking the clocks around them. Here are the details: AMD Ryzen 5 1600X paired with Corsair H100i Aio 16GB RAM, running at 3200MHz. For graphics, I used an EVGA SC GTX 970 on an Asus TUF B450M Plus Gaming board, which is part of the build. I've been using Passmarks Performance Test 10 as my benchmark. I know this isn't the absolute best bench, but it's one I'm comfortable with because I only use it to check how much the system improved or stayed the same. So here comes the problem. When everything was set up by default at stock settings, the CPU ran at around 3.6GHz and 1.375V while the GPU ran at 1317MHz and 7010MHz. The scores were: CPU - roughly 13,000 with temps between 50 and 55 degrees. GPU - roughly 8800 with temps between 65 and 70 degrees. After I tried to speed up the GPU while keeping the CPU at stock settings, the GPU scored about 10,000 but got hotter, reaching 70 to 80 degrees. Is that good news? So then I turned my attention to the CPU and started messing with its clock speeds until it ran stable at 4.0GHz and 1.4V. When I re-tested everything after this change, the scores were: CPU - roughly 14,000 with temps between 60 and 70 degrees. GPU - still around 8000 (the same as before) but stayed hot at 65 to 70 degrees. Now I realize how surprised I got when I saw such a big drop in the GPU score compared to what it used to be, even though the CPU speed went up and temps were actually fine. You probably know that sometimes high power or heat on the CPU can hurt the graphics card, but since my CPU was still cool and the scores only improved slightly by 200 points, I didn't think this would happen to me. But I do have to admit it's been my first time overclocking a Ryzen system so I'm not sure if I did everything right. Also just to be clear, I went back down to the default CPU speeds and the GPU score immediately went up all over again, which makes me pretty sure something was wrong with how much I pushed the CPU. Thanks so much for any help or advice on what's actually going on here!
Most people get the power supply question wrong. The quality of inside parts matters more than how much wattage you have. Either way, everything is working fine here: http://www.jonnyguru.com/blog/2015/11/09...-supply/6/
Ha ha, yeah that makes sense. I never skimp on power supplies before, even though I got some serious headaches with cheap, unbranded parts in the past. I always stick to brands I've used for years because I know them. =) Any thoughts on what could be making my bench test numbers drop after I overclocked the CPU? It's the first time I've seen this exact problem and I'm honestly stuck as to why it's happening.
Honestly, I'm still not sure about this. But if we look at how the CPU and GPU work together, maybe the overclock isn't really working well right now. It might just be that you haven't tested it enough times yet. Like, have you tried running them side by side?
Oh sorry, maybe I should have explained more clearly in my first post. The three samples I sent were just for context. I did many tests across different clocks and voltages, and the results are pretty consistent. As soon as I breathe on the CPU clock speeds, the GPU performance drops a lot (at least according to the benchmarks). Honestly, I can't imagine that this would be causing any bottlenecks anywhere in the system. But as I said before, this is my first time messing around with Ryzen so maybe I'm just totally uneducated about how this specific chip works. To be brutally honest, I don't really care much about overclocking this rig anyway because it's a Christmas present for my nephew who lives over a country away from me. So I'll just reset everything back to stock before I ship it off. I'm only experimenting right now out of morbid curiosity. Ha ha.
Try climbing up to 4.0ghz and checking scores at 3.7, 3.8, 3.9. If your CPU overclock isn't working smoothly, those little glitches will ruin the GPU performance.
Yeah, I was thinking about doing that tomorrow. Maybe a good night of sleep will help clear my head too. Ha ha! I'll have another play around tomorrow and try increasing the clocks slowly and see how it goes. I'll post back here with an update once I've got some more solid data to go off.
I just ran a few quick checks after work and nothing looks conclusive yet. I really want to keep going too. I also changed the graphics card from the 970 to a new 1060 just for fun to see if things behave differently with another chip. Everything else in my setup stayed the same. My computer is getting quite warm; about 36 degrees on the CPU and 50 degrees on the GPU. It's really hot right now, especially since I work in a small office so it gets sticky inside. Before testing, I waited for the room temperature to cool down some. Here are the results so far: Bench score > average temp reached > average wattage reached Reboot Clocks look like this: Memory at 2133mhz; CPU running at 3.6ghz using 1.375V voltage; GPU Core clocking at 1885mhz and Mem at 4006mhz with 1.062V power. Power set to 100% and Temp target at 83C. The first run: Memory was 2173, CPU hit 54C for 70W, GPU ran at 10133MHz/62C, 2nd run saw memory rise slightly to 2180 with the same CPU and 52C/GPU drop to 9172/61C, then again 2182/CPU hit 53C/GPU stayed at 10089/61C. After another reboot: Memory back down slightly while clocking higher; CPU surged up to 3.8ghz with 1.375V; GPU Core went to 1885mhz, Mem 4006mhz, Power hit 116%, and Temp target bumped to 92C. First run: Memory at 2168, CPU jumped to 12601MHz hitting 53C for 70W, GPU at 10055/61C; second run memory fell slightly while clocking faster, CPU surged again to 12638/62C/GPU hit its target of 9164/62C, then third run memory rose a bit more while clocking steady, CPU stayed at 12615 but GPU dropped back down to 9174/62C. The mystery keeps popping up with some weird results that I don't know what to do about yet. It's still just a small group of tests though, so maybe a bigger testing plan would help things out more. What really surprises me is how much the CPU temp and power draw went up after only changing the multiplier by a couple of steps. I haven't even adjusted the voltages yet. I didn't think a simple .2ghz bump would cause such big changes in temps and watts. Anyway, I'll run some more tests later, but for now that's everything I can share here. Any thoughts?