ChrisLB Posted July 22, 2020 Posted July 22, 2020 I built a new soft tubing water cooled system a few months ago. This time I decided to use the XD5 pump. It seemed that it was a good value and my motherboard has built in Corsair lighting controls. I noticed that with the custom loop, my CPU temperatures weren't much better than with a $140 AIO and my EVGA RTX 2080 Super GPUs weren't much cooler than with their factory 120mm AIO cooling. I did some testing and removed the GPUs and just had the CPU cooled by a 360 60mm thick radiator and a 280mm 45mm thick radiator. Both radiators have push pull fans. My CPU temperatures only dropped 1C after removing the GPUs from the loop. Other people using the same CPU block were claiming to have much lower temperatures with similar systems. Most of them said their CPU temperatures were 6C or 7C cooler than what I was getting with the same tests. I tested the system at both a 3700 RPM fixed pump speed and full pump speed with all fans at 100% to see what the maximum cooling abilities were. I decided to build another water cooled PC using my old PC. I was going to use the corsair XD5 in that system and bought an EKWB D5 PWM pump for my current system. The EK pump and mount matched the all black theme better than the Corsair pump and mount. When I swapped in the EK D5 pump into my current system and ran the same tests at both 3700 RPM and full pump speed, my CPU temperatures dropped by 6C in both AIDA 64 and the Intel Burn Test. My CPU temperature dropped 8C in Prime 95. It was obvious that the XD5 pump has much less head pressure or flow at the same RPM and at full speed. Was my XD5 pump defective from the day I purchased it, or are EK's D5 PWM pumps that much stronger(better flow and head pressure) than the Corsair D5 pumps? Basic system specs if it helps: R9 3950X Dual EVGA 2080 Super GPUs Optimus Foundation AM4 CPU block EKWB EK-FC RTX 2080 +Ti Classic GPU blocks EK-CoolStream XE 360 radiator EKWB EK-CoolStream CE 280 Radiator Corsair XD5 pump/res or EKWB D5 pump/res
BlaiseP Posted July 22, 2020 Posted July 22, 2020 (edited) Cory from Designs by IFR did a comparison between a few pumps (including the XD5) in a video: [ame=https://www.youtube.com/watch?v=sS8QlSf3snA]The ULTIMATE PC Water Cooling Race.. Who Will Win?[/ame] Edited July 22, 2020 by BlaiseP
LeDoyen Posted July 22, 2020 Posted July 22, 2020 (edited) Technically they should be both identical since they use the same pump inside. But EK specs give 1500l/min at 3.9m head pressure Corsair for the same pump states 800l/min at 2.1m head pressure. Either it's two typos, or they need to redesign the pump top.. In most builds it doesn't matter, but, with 3 waterblocks and 2 rads, if it's a series loop, and these numbers are true, i can see why corsair's underperforms. Edit: saw the video.. nevermind, that pump sucks... keep the EK. Corsair's specs are correct, they did manage to kill the one advantage of D5 pumps with their pump top... Edited July 22, 2020 by LeDoyen
c-attack Posted July 22, 2020 Posted July 22, 2020 (edited) Well something wasn't quite right, but the specs on the pumps aren't the story. You can't cut 8C off anything by increasing flow rate unless you are in an effective standstill state. I expect you don't run 8C warmer if you cut the EK pump speed in 1/2. I ran my XD5 in a three block, 2x360 set-up with a ridiculous 14x90 degree fittings. It was not done with flow performance in mind. I would later replace the pump with something else and there was no change in performance at any level or pump speed. It's hard to troubleshoot it after the fact. Usually when something is wrong with the pump it will let you know in a clear and loud fashion. Strange vibrations, odd noises, something. You could hook it up out out the back in a quick loop around where flow rate is easy to judge. Edited July 22, 2020 by c-attack
LeDoyen Posted July 22, 2020 Posted July 22, 2020 past a certain flow rate, it starts to matter, but it's usually pretty low. I see negligible temperature changes going from full speed to like 2500 rpm, then at a point temps increase "dramatically".. between 2500 and 1500 rpm, temps go up by 6 or 7°. I guess that's roughly the kind of flow the XD5 may give at 3700 rpm. Kinda comparing apples and oranges, different loops, but, the point is the XD5 will make temperature rise sooner than any other pump that takes advantage of the full flow of a D5. 800L/min is still useable, but with a more complex loop, given the lower head pressure, i would probably favor a full parallel loop, to avoid losing more flow.
ChrisLB Posted July 22, 2020 Author Posted July 22, 2020 Well something wasn't quite right, but the specs on the pumps aren't the story. You can't cut 8C off anything by increasing flow rate unless you are in an effective standstill state. The Optimus CPU blocks have been measured to have much higher than average pressure drop due to their design. According to one test I found, its about twice the pressure drop of the EK Supremacy CPU block and about twice as much pressure drop as the corsair XC7 CPU block. At a standstill, my CPU hits 95C rather quickly then I get the high temperature alarm. So, no the loop was nowhere near a standstill. I found that out one day when I booted the PC after doing some more cable management and I forgot to hook the power back up to the pump. As you might guess, 16 core CPUs heat up quickly. I expect you don't run 8C warmer if you cut the EK pump speed in 1/2. I ran my XD5 in a three block, 2x360 set-up with a ridiculous 14x90 degree fittings. It was not done with flow performance in mind. I would later replace the pump with something else and there was no change in performance at any level or pump speed. Luckily, its easy to test that. I set the EK D5 pump to 1850 RPM and ran Prime95, AIDA64, and Intel Burn Test and recorded the temperature results. The results with the EK D5 pump at 1850 RPM were within 1C of the Corsair XD5 pump at 3700 RPM. So it looks like the Corsair D5 pump does have a much lower flow rate than the EK D5 pump. It's hard to troubleshoot it after the fact. Usually when something is wrong with the pump it will let you know in a clear and loud fashion. Strange vibrations, odd noises, something. You could hook it up out out the back in a quick loop around where flow rate is easy to judge. Its pretty obvious now that the Corsair XD5 pump has a much lower flow rate and lower head pressure than the EK D5 pump. It could just be a poor pump top design.
c-attack Posted July 23, 2020 Posted July 23, 2020 Its pretty obvious now that the Corsair XD5 pump has a much lower flow rate and lower head pressure than the EK D5 pump. It could just be a poor pump top design. Except of course I have two EK D5s and one of them is what replaced the XD5. No difference in performance at 1800 or 3500 rpm in a very entangled loop. If you are running any pump at 3500 rpm and there is a flow issue, something else is wrong. The basic functionality of a water cooling system cannot hinge upon whether the CPU block is low, moderate, or high resistance. It can't be or every loop would fail when some novice took a weird turn. Was this 8C better coolant temperature? Or 8C at the CPU cores? Are your GPUs in parallel? That might cause a large enough pressure drop especially if the CPU block is right after them in the flow path. Either way, I am sure you are not going to be putting the XD5 back in, so the original question was "is it defective?". I think the only way to know that is to run it outside the case in base loop or with any radiators or blocks lying around. Even in a loop back to itself, usually motor problems make themselves known.
ChrisLB Posted July 23, 2020 Author Posted July 23, 2020 Except of course I have two EK D5s and one of them is what replaced the XD5. No difference in performance at 1800 or 3500 rpm in a very entangled loop. If you are running any pump at 3500 rpm and there is a flow issue, something else is wrong. The basic functionality of a water cooling system cannot hinge upon whether the CPU block is low, moderate, or high resistance. It can't be or every loop would fail when some novice took a weird turn. Loop order goes: pump/res -> GPUs -> radiator -> CPU block -> radiator -> pump/res No different than what I've seen with many other loops. I had the same temperature results when I tried this order also: pump/res -> GPUs -> CPU block -> radiator -> radiator -> pump/res Was this 8C better coolant temperature? Or 8C at the CPU cores? Are your GPUs in parallel? That might cause a large enough pressure drop especially if the CPU block is right after them in the flow path. Its a 6C to 8C difference at the CPU cores depending on the test run. GPUs are in parallel. Either way, I am sure you are not going to be putting the XD5 back in, so the original question was "is it defective?". I think the only way to know that is to run it outside the case in base loop or with any radiators or blocks lying around. Even in a loop back to itself, usually motor problems make themselves known. The motor isn't audible outside the case unless its above 3700 RPM. Even then its a normal humm, just louder as the RPM go up. No other noises. Without taking the pump apart or without having some sort of flow meter available, I don't know an accurate way of testing it other than having it pump water form one bucket to another to see how much it pumps in 60 seconds. Even then that's just sort of accurate and probably not the best way to determine the actual flow rate.
LeDoyen Posted July 23, 2020 Posted July 23, 2020 Well, Corsair specs themselves state it's half the flow of a normal D5.. A last ditch resort would have been to also put the CPU in parallel with the GPUs. In series, i would have put CPU before GPU, simply because GPUs do warm the water up substantially if the flow is this weak. Corsair fixed the XG7 manifold, now they have a pump top to widen...
ChrisLB Posted July 23, 2020 Author Posted July 23, 2020 Well, Corsair specs themselves state it's half the flow of a normal D5.. A last ditch resort would have been to also put the CPU in parallel with the GPUs. In series, i would have put CPU before GPU, simply because GPUs do warm the water up substantially if the flow is this weak. Corsair fixed the XG7 manifold, now they have a pump top to widen... I ran all the tests with the GPUs idle and I also reran the tests with the GPUs removed from the loop and it made no difference in CPU temperature when I was just testing CPU temperatures with or without the idle GPUs in the loop. If the pump is delivering adequate flow, loop order doesn't matter at all. Obviously the XD5 doesn't deliver adequate flow. However with the GPUs idle or out of the loop, it shouldn't make a difference.
LeDoyen Posted July 23, 2020 Posted July 23, 2020 GPUs idle yea it won't make a difference. I meant when they are loaded. At idle mine sits 1° above water temp. Not much to heat up with. with very low flow, you start to see big temperature differences between water supply and return on GPUs. That can raise the CPU temp by the same amount, but at this point, the CPU block also doesn't dissipate as much so it's a mess.. bottomline.. need flow
c-attack Posted July 23, 2020 Posted July 23, 2020 I ran all the tests with the GPUs idle and I also reran the tests with the GPUs removed from the loop and it made no difference in CPU temperature when I was just testing CPU temperatures with or without the idle GPUs in the loop. Having the GPUs in or out of the loop certainly should matter and removing them should have removed half the resistance in the loop. Things should have improved if pressure and flow rate were the restriction. I am afraid that does little to resolve the original question, but it seems weird there would be such a difference in a CPU block + 2 rad loop. At this point, all I can say is that must be one very restrictive CPU block. I haven't found an AMD version test yet, but the Intel one is about the double the norm. Presumably that is how they create the advertised improvements.
ChrisLB Posted July 23, 2020 Author Posted July 23, 2020 (edited) Having the GPUs in or out of the loop certainly should matter and removing them should have removed half the resistance in the loop. Things should have improved if pressure and flow rate were the restriction. I am afraid that does little to resolve the original question, but it seems weird there would be such a difference in a CPU block + 2 rad loop. At this point, all I can say is that must be one very restrictive CPU block. I haven't found an AMD version test yet, but the Intel one is about the double the norm. Presumably that is how they create the advertised improvements. From what I've seen in other testing two EKWB GPU blocks in parallel is about 0.75 to 0.85 PSI pressure drop depending on the model. If the specs stated above are correct: EK specs give 1500l/h at 3.9m head pressure Corsair for the same pump states 800l/h at 2.1m head pressure. Then its possible that two radiators, a few 90 degree fittings and the Optimus Foundation AM4 block are just too much for the corsair pump since it has about HALF the flow and head pressure of the EK D5 pump. From what I've seen of the Otimus AM4 block, its not only just a larger block, its a much larger surface area that is covered with those small fins compared to an older EK block or the Corsair XC7 block. That could mean much higher pressure drop than the average CPU block, and even more pressure drop than the intel version. Edited July 24, 2020 by ChrisLB fixed typos
ChrisLB Posted July 23, 2020 Author Posted July 23, 2020 A last ditch resort would have been to also put the CPU in parallel with the GPUs. From what I've seen, if the CPU block has much higher restriction than the GPU blocks, as the Optimus blocks do for example, then the CPU block gets a lot less flow and that gives you higher CPU temperatures.
LeDoyen Posted July 23, 2020 Posted July 23, 2020 From what I've seen, if the CPU block has much higher restriction than the GPU blocks, as the Optimus blocks do for example, then the CPU block gets a lot less flow and that gives you higher CPU temperatures. I hadn't seen how this block was. you are totally right. It uses microfins, which makes more sense in AIOs with small flow pumps. On a parallel loop it would send all the water to the GPUS.. so, nevermind ^^ Looks like a good block to run on a dual loop setup. I'd still recommend thick fins for custom loop with high flow pumps. Easier to match restriction between blocks = more setup possibilities (and the block doesn't behave like a filter for the ocasional impurities you'll inevitably find :D: )
ChrisLB Posted July 23, 2020 Author Posted July 23, 2020 Looks like a good block to run on a dual loop setup. I've actually thought about running a dual loop setup in that case. I figured out a way to do it. The XD5 should be adequate to cool the GPUs. However, the dual loop would just be for ePeen points and no practical reasons.
c-attack Posted July 23, 2020 Posted July 23, 2020 Not that much extra space in there. Were you thinking about one those flat panel reservoir/pumps against the side glass?
ChrisLB Posted July 23, 2020 Author Posted July 23, 2020 Not that much extra space in there. Were you thinking about one those flat panel reservoir/pumps against the side glass? I could mount a DDC pump/res where the rear fan is located and use the top radiator for the CPU. I'd mount the D5 pump/res on the side radiator and use a side and bottom radiator for the GPUs. Its pointless really. There probably would be little to no performance benefit.
c-attack Posted July 23, 2020 Posted July 23, 2020 I had considered that when I first built in the XL with the air of making a smoother appearance around the side mount position. Ultimately didn't like the close corners between CPU block, top radiator, and GPU. There's space to be had in back if you move the PSU to the top shelf, but as you stated there doesn't seem to be much purpose unless the CPU is banging against the upper limit and you need to separate from the GPU heat in the coolant stream.
Recommended Posts