GPU Block Differential Pressure vs. Flow Test

[Radical_53]

Die size maybe? GPUs are much larger than CPUs normally, at least all those I know of are. That's why you'd need more "tech" to keep a CPU down than a GPU...

[Baenwort]

Any chance we'll get some of the older blocks added to this list so those of us keeping a eye out on the used market will have some data to compare with? Data for a 6002 would help satisfy my geek lust.

[Petra]

Any chance we'll get some of the older blocks added to this list so those of us keeping a eye out on the used market will have some data to compare with? Data for a 6002 would help satisfy my geek lust.

I'll take care of that once I start doing CPU block testing...

[epion2985]

Why is everyone concerned with pressure drop, isn't that going a bit backwards? I would be more concerned with which block under load X and setup Y nets lower temperatures. Weather it kills the flow or not is not really relevant at that point as long as it performs better. Am I missing something?

[Radical_53]

Not if it only cools slightly better but gives you a much higher restriction. Die sizes are still big enough to make it fairly easy cooling a GPU. Even putting a good CPU block on one wouldn't give you much of an advantage over a "normal" GPU block.

[epion2985]

Not if it only cools slightly better but gives you a much higher restriction. Die sizes are still big enough to make it fairly easy cooling a GPU. Even putting a good CPU block on one wouldn't give you much of an advantage over a "normal" GPU block.

That still doesn't make any sense. If block X get you the best temperatures then I could care less what the pressure drop was. Low pressure drop wont get you a better oc, but lower temperatures will.

[Radical_53]

Well, it still does make sense, maybe not for you If your loop consisted only of that one block, ok. But if you have several blocks and a rad, all with improved performance at higher flow levels, then you might see the problem.
You normally use a restrictive CPU block as only those can give peak performance, the performance difference to very low restriction blocks is large. So if you put a block like that in, the overall gain is still there.
But for a GPU block, where you might use 2 of them, the performance gain is so small (if there is any, haven't seen a more restrictive block that would perform better than good low restriction ones) that you'd get from the block itself that it will be less overall performance as the other blocks and the rad suffer due to the decreased flow.

[epion2985]

oh multiple block loop, lol its late my brain is numb. So when looking at cpu block you look at best temperatures, when looking at gpu block you look for least pressure drop because temperatures will be good anyway, and the precious flow you save for the cpu block.

[Radical_53]

Well, it's always a compromise somehow I woulnd't use a CPU block that shaves off half a degree and was 2 times as restrictive as another.
That's because the rad performs better also with higher flow, and the performance of the loop is the sum of all things added up... and as the performance would suffer heavily with warmer water of course, I woulnd't want to restrict my rad more than I had to.
For the GPU blocks it's so far a simple choice, as even low restriction blocks can show very good performance, performance that even higher restriction blocks can't beat really. So why use a higher restriction one, just because it might be better in theory? No way...

[Petra]

That still doesn't make any sense. If block X get you the best temperatures then I could care less what the pressure drop was. Low pressure drop wont get you a better oc, but lower temperatures will.

Look at it this way, for a moment: Put a restrictive GPU block (e.g. MP-1 GPU) on your video card that performs, say, 2*C better (under certain conditions) than a non-restrictive GPU block (e.g. Maze4/FusionHL/FusionSL)... Your video card isn't going to care about the difference--if it does provide more overclocking headroom (doubtful), it'll be minuscule. The trade-off for this tiny drop in GPU temp is significantly decreased flowrate, which results in poorer performance from your CPU block (CPU being a more temperature sensitive component that has a lower operating temp ceiling) and a decrease in radiator performance.

Does that sound, at all, like a smart trade? If you're using multiple loops, then, sure, go to town... but, as far as smart single loop construction goes, it just doesn't make sense.

*Ninja Edit* Radical_53 beat me to it.... see what happens when I talk on the phone and try to type at the same time... sheesh

[epion2985]

Well, it's always a compromise somehow I woulnd't use a CPU block that shaves off half a degree and was 2 times as restrictive as another.
That's because the rad performs better also with higher flow, and the performance of the loop is the sum of all things added up... and as the performance would suffer heavily with warmer water of course, I woulnd't want to restrict my rad more than I had to.
For the GPU blocks it's so far a simple choice, as even low restriction blocks can show very good performance, performance that even higher restriction blocks can't beat really. So why use a higher restriction one, just because it might be better in theory? No way...

ARRRRRRRRRR, but thats my point. In a given setup if block X will give you the lowest temperature with the rest of your great on everything (radiator, cpu etc), then thats the block you want, pressure or no pressure.

[Jochenp]

yeah, but they say the rest will always suffer
I get what you mean, if temps of cpu are unchanged adding a highpressure gpu block and gpu's temperatures went down, who cares that flow is lower
they are saying the lower flow will ALWAYS negatively affect performance

[Radical_53]

@epion: Well, still your point doesn't make sense.

Just an example. If you used a normal, low restriction block, you'd get maybe a delta of 15k for the gpu, 15k for the cpu and 5k for the radiator. So the GPU would be 20k above ambient in total.
Now you tried the super-restrictive super-performing block for the gpu, dropping temps by 2° to 13k delta. But now your cpu suffers going up to 18k and your rad goes up to 7k.
In total, your gpu isn't any cooler than before, but your cpu runs 3° warmer. See my point? It's just an example, but that's how using a high restriction gpu block shifts around those values. It normally doesn't get you anything.

[epion2985]

yes I see your point but you misunderstood mine, but don't worry about it I just realized we are going in a loop lol its late ><

[Radical_53]

Seems like... yes now I see what you meant on your last post. That's correct, you're searching for the block that might give better peformance than another without giving you worse CPU temperatures...