Mid-Life Crisis

[ArchStanton]

After chasing my tail for a bit, I've reached the conclusion that its RMA time.  I got a fair amount of guidance from OFFICIAL 5900X and 5950X two chiplet Zen 3 CPUs Overclocking thread | Page 64 | Overclock.net (from post # 1197 onward) on what to test and how.

 

Will have more to add here once I'm back up and running again .

 

[J7SC_Orion]

13 hours ago, ArchStanton said:

After chasing my tail for a bit, I've reached the conclusion that its RMA time.  I got a fair amount of guidance from OFFICIAL 5900X and 5950X two chiplet Zen 3 CPUs Overclocking thread | Page 64 | Overclock.net (from post # 1197 onward) on what to test and how.

 

Will have more to add here once I'm back up and running again .

 

 

Wow, what happened ? All the test you showed in this thread like 7 & 8 hrs Prime95 and CPU tests showed that it was doing well. 

[Avacado]

Sorry man. Thats been the energy of 2021 and this year. I have had more RMA's, failing components in the past year and a half than since 2000. 

[ArchStanton]

Just now, J7SC_Orion said:

 

Wow, what happened ? All the test you showed in this thread like 7 & 8 hrs Prime95 and CPU tests showed that it was doing well. 

I expanded the size range of FFT's I was testing with.  I believe the "out of the box" defaults for the versions of prime95 and CoreCycler I was using don't trigger the peak boost frequencies the CPU can try to hit under the right circumstances.  My current understanding is that CoreCycler in SSE mode with runtime set to "auto" and FFT size set to "all" will find instabilities that the defaults won't.  Obviously, AVX and AVX2 would also need to be tested, but my CPU core 1 will fail CoreCyler within 5 minutes at an FFT of between 20k-1000K without any DOCP or PBO.

[ArchStanton]

4 minutes ago, Avacado said:

Sorry man.

Not to worry, it has been an extremely informative branch of my "OC Adventure" .

[ArchStanton]

Besides, the down time gives me an excuse to add a second (better) pump to the loop...

[ArchStanton]

So, while my "adventure" has been temporarily sidelined with a layover in "RMA Purgatory", I decided to try and learn a little more about the much maligned "XD5" pump/res combo from Corsair.  I've seen the included pump described as "gimped", "knockoff", etc. here and elsewhere, and I certainly don't make any claims to the contrary, but I'm confused as to how the pump actually differs from others on the market.

 

I do have some firsthand measurements of the unit's performance.  My loop consists of 480x40 radiator, 360x64 radiator, 360x20 radiator, CPU block, GPU block, flow meter, the XD5 itself, and a stupid number of 90-degree/45-degree fittings used mainly because of my aesthetic preferences.  Under these constraints, the pump produces a measured (Koolance INS-FM19) flow of roughly 3100mL/min (0.82gal/min or 186L/hr).  While we are a long way from the 1.5gal/min league, I would think my loop is of the "high restriction" variety, and the flow seems to reflect that.  The tach signal for the pump indicates it is turning 4800 RPM +/- at 100% PWM as expected.

 

I purchased the unit prior to joining this forum, and I would likely have gone a different route had I been aware of the unit's less than stellar reputation in the community.  However, this review led me to believe the unit was viable:  

CORSAIR Hydro X Series XD5 Pump/Reservoir Combo Review

WWW.TECHPOWERUP.COM

---

We add to our detailed coverage of CORSAIR's Hydro X series of custom watercooling products with their XD5 pump/reservoir combo unit. It uses the ever-reliable Xylem D5 PWM pump paired to a...

 

 

My question is this, specifically, how is the included pump "gimped" in comparison to its competition.  The linked review makes note of the fact that Corsair's published specs reflect an expected flow rate for a "typical" system pressure rather than the theoretical maximums.  Is this the source of the unit's bad reputation?  The published electrical specifications indicate it will peak at about 30W output.  This seems in-line with the rest of the D5 herd.  Did Corsair change the model of included pump subsequent to the publication of the linked review?  Inquiring minds want to know , and like Bartles & Jaymes, we thank you for your support.

 

 

 

P.S.  While typing this up, I recalled reading Avacado's and ciarlatano's comments in the Post your Last Purchase - Page 38 - Chit Chat General - ExtremeHW thread.  After rereading them, it seems my question(s) have largely already been answered there.  ciarlatano indicates that the pump isn't truly "gimped" but rather that it is a slightly less powerful pump (30W vs 35-37ish maybe?) that Corsair is blatantly overcharging for.   This certainly coincides with my experience so far.  The pump is "adequate" in a general sense without being a cost-effective option compared to its competition.  Hearing the XD5 described as "half flow" so many times prompted me to investigate a little deeper to see how accurate the statement is.  I now regard it as just a general "knock" against Corsair's current business practices rather than a statement intended to be taken as measurable fact.  The pump isn't "half flow", it's just overpriced for what it is.

 

P.P.S.  I debated deleting this entire post after coming to the conclusion indicated in the paragraph above, but I think I will post it anyway on the off chance that another noob like me might benefit from it.  Especially the sort of noob that replies to every answer from an expert with the follow up question, "but, why?". 

 

[Sir Beregond]

I think the main stat that was being looked at was the Corsair flow rating marked as 800L/h while a regular D5 like from Aquacomputer for example is 1500L/h.

 

Source: 

 

Edited February 8, 2022 by Sir Beregond

[ArchStanton]

1 minute ago, Sir Beregond said:

I think the main stat that was being looked at was the Corsair flow rating marked as 800L/h while a regular D5 like from Aquacomputer for example is 1500L/h.

Aye, but the 800L/h is at a pump outlet pressure of 2.1m (of H2O @ 4 Celcius I presume), or roughly 3psi.  The 1500L/h rating has no associated pressure value, leading me to believe it is the theoretical maximum flow in a mythical "zero restriction" loop.  The Aquacomputer D5 variant may produce a flow of around 900L/h in a loop where the output pressure of the pump is 3psi, but I do not believe it will produce anywhere near 1500L/h in a loop with sufficient restriction to induce a pressure of 3psi at the pump outlet.  In a nutshell, I believe there are better D5's on the market.  I do not believe there are D5's on the market that will produce twice the flow of the unit Corsair includes with their XD5 combo if used in the same loop.  The charts in the linked review support my contention.  However, if we factor in cost, and calculate the flowrate per dollar (i.e. Liters per hour per dollar) now we may be getting into territory where the Aquacomputer unit is "twice as good" .

 

Apologies in advance for the ramble.  I just happen to find this particular topic to be of "more than rational" interest at the moment 

[Avacado]

1 hour ago, ArchStanton said:

Aye, but the 800L/h is at a pump outlet pressure of 2.1m (of H2O @ 4 Celcius I presume), or roughly 3psi.  The 1500L/h rating has no associated pressure value, leading me to believe it is the theoretical maximum flow in a mythical "zero restriction" loop.  The Aquacomputer D5 variant may produce a flow of around 900L/h in a loop where the output pressure of the pump is 3psi, but I do not believe it will produce anywhere near 1500L/h in a loop with sufficient restriction to induce a pressure of 3psi at the pump outlet.  In a nutshell, I believe there are better D5's on the market.  I do not believe there are D5's on the market that will produce twice the flow of the unit Corsair includes with their XD5 combo if used in the same loop.  The charts in the linked review support my contention.  However, if we factor in cost, and calculate the flowrate per dollar (i.e. Liters per hour per dollar) now we may be getting into territory where the Aquacomputer unit is "twice as good" .

 

Apologies in advance for the ramble.  I just happen to find this particular topic to be of "more than rational" interest at the moment 

It's all good, your well thought out ideas and postings are why I willed you here . My arguments specifically come from each manufacturers specification sheet on each product. The 1500l/h DOES come with a discharge head of 3.7m. That is the point I was trying to make. The best answer I received in return was that "Corsair made a mistake in listing the specifications". If that is true it means they are either really stupid or the specifications are indeed true. If you were to look at a spec sheet for a pump that read as follows:

 

800l/h at 2.1m head using 30w

 

or

 

1500l/h at 3.7m head using 23w

 

Which would you choose? Again, these are directly from each manufacturers technical specification sheets. It could be a "Theoretical maximum", but why would you (Corsair) NOT list the highest capability of your product? It may not be true, but given the information at hand, I would not be willing to take that chance. No pump achieves it's highest flow rate with addition of restriction, that is not disputed. The only way to put this to bed once and for all is to either do a test with these pumps or have Corsair re-word it's specifications so it doesn't appear misleading.  

 

 

Edited February 8, 2022 by Avacado

[ArchStanton]

18 minutes ago, Avacado said:

It's all good. My arguments specifically come from each manufacturers specification sheet on each product. The 1500l/h DOES come with a discharge head of 3.7m. That is the point I was trying to make. The best answer I received in return was that "Corsair made a mistake in listing the specifications". If that is true it means they are either really stupid or the specifications are indeed true. If you were to look at a spec sheet for a pump that read as follows:

Now you've really got my "testing" juices flowing! I'm going to get one of the Aquacomputer units and install it in place of the Corsair "Xylem D5" for science .  Thank you for the clarification on the published specs of the Aquacomputer variant.  I did not interpret the spreadsheet you included in your post to indicate that the 1500L/h AND 3.7m pressure would be observable concurrently.  This is exciting .

[Avacado]

27 minutes ago, ArchStanton said:

Now you've really got my "testing" juices flowing! I'm going to get one of the Aquacomputer units and install it in place of the Corsair "Xylem D5" for science .  Thank you for the clarification on the published specs of the Aquacomputer variant.  I did not interpret the spreadsheet you included in your post to indicate that the 1500L/h AND 3.7m pressure would be observable concurrently.  This is exciting .

They may not be. I am glad you are going for it! I do not have the hardware to test actual head pressure. Sure I can test flow, but that's about it. 

[ArchStanton]

8 minutes ago, Avacado said:

Sure I can test flow, but that's about it. 

Same here, and your flow meter is likely better calibrated than mine, BUT it will be interesting none the less even if all I can provide are relative flow measurements .

 

Aquacomputer D5 Pump

WWW.TITANRIG.COM

---

D5 pump body (does not include pump top)Max. Discharge head: 3.7mMax. Flow: 1500 l / hrPower supply via 4-Pin Molex power connector, speed signal via separate 3-Pin fan connectorDimensions:...

 

Should be here later this week (out of stock at Mod My Mods, I did try them 1st). 

 

Edit:  I saw no reason to get the PWM variant.  That * is gone be at max or off no matter what 

 

2nd edit:  The forum auto filters profanity...did not realize that.

Edited February 8, 2022 by ArchStanton

[Avacado]

6 minutes ago, ArchStanton said:

Same here, and your flow meter is likely better calibrated than mine, BUT it will be interesting none the less even if all I can provide are relative flow measurements .

 

Aquacomputer D5 Pump

WWW.TITANRIG.COM

---

D5 pump body (does not include pump top)Max. Discharge head: 3.7mMax. Flow: 1500 l / hrPower supply via 4-Pin Molex power connector, speed signal via separate 3-Pin fan connectorDimensions:...

 

Should be here later this week (out of stock at Mod My Mods, I did try them 1st). 

Given you are doing flow calculations. I will give you reference.

 

Here on the office rig I am running an AC D5 same as you 100% at 4822rpm.

 

1 ACool UT60 360mm rad

1 EK 3070 WB

1 Ncore V1 CPU block

 

Current output: 250lph/1.1GPM. I hope it ends up being better. If it doesn't then you have yourself a hell of a 2nd pump to integrate into your loop!

 

 

 

 

Edited February 8, 2022 by Avacado

[ArchStanton]

2 minutes ago, Avacado said:

you have yourself a hell of a 2nd pump to integrate into your loop!

Indeed.  Having taken the "water cooling leap" finally, I do not intend to go back to air in the future.

[J7SC_Orion]

1 hour ago, ArchStanton said:

Indeed.  Having taken the "water cooling leap" finally, I do not intend to go back to air in the future.

 

...water-cooling makes eminent sense if you run a CPU and GPU with aggressive boost algorithms that have temps as an input 

 

I typically use a minimum of 2x D5s in a circuit (mostly because I also build water-cooled servers and want the fail-over) but sometimes I use 3x D5s, like the white-ringed-ones at the back, if I have an otherwise unused D5. I prefer the ones with both PWM and manual adjustment, though I rarely connect the PWM. In the pic below, I 'slowed' the 3x D5s to about 55% or so via the manual adjustment knob on the back ....they're also rubber-mounted and perfectly quiet at that speed.

 

[ArchStanton]

@J7SC_OrionOut of curiosity, your multi-pump loops have the pumps in series for fail-over?  Or you run then in parallel for increased flow with check valves for fail-over?  I tried to see how you had them plumbed in your build log, but I couldn't quite make it all out.

[tictoc]

It's just silly for Corsair to list anything other than the max flow rate.  There is almost no way to estimate what the flow rate will be in an "average" loop, at least without doing some pretty involved fluid dynamic calculations.  It looks like Corsair did change the spec page and now it says: "800L/h at 2.1m pressure head (1500L/H theoretical max)".  The AquaComputer spec sheet is nice, since it includes some numbers for what can be expected in the real world, alongside the actual rating of the pump.

 

A well "calibrated" bucket and clock, with the return line running to the bucket, will give you an exact measurement.

 

At the end of the day, as long as water and component temps are acceptable, and the pump doesn't have an unreasonably short life, then that's all that really matters.

[ArchStanton]

54 minutes ago, tictoc said:

At the end of the day, as long as water and component temps are acceptable, and the pump doesn't have an unreasonably short life, then that's all that really matters.

No argument there.  This whole build is just for entertainment purposes.  Strange as it may seem, I anticipate testing the AC pump when it arrives to be entertaining, for my weird self if no other .

[Avacado]

3 minutes ago, ArchStanton said:

No argument there.  This whole build is just for entertainment purposes.  Strange as it may seem, I anticipate testing the AC pump when it arrives to be entertaining, for my weird self if no other .

If it does in fact raise your flow over 1GPM, you realistically can only expect a 1-3c reduction in temps. You simply must make a video. 

[Avacado]

1 hour ago, tictoc said:

It's just silly for Corsair to list anything other than the max flow rate.  There is almost no way to estimate what the flow rate will be in an "average" loop, at least without doing some pretty involved fluid dynamic calculations.  It looks like Corsair did change the spec page and now it says: "800L/h at 2.1m pressure head (1500L/H theoretical max)".  The AquaComputer spec sheet is nice, since it includes some numbers for what can be expected in the real world, alongside the actual rating of the pump.

 

A well "calibrated" bucket and clock, with the return line running to the bucket, will give you an exact measurement.

 

At the end of the day, as long as water and component temps are acceptable, and the pump doesn't have an unreasonably short life, then that's all that really matters.

Well that is weird.

 

It would seem there are 2 variants? This product (CX-9040002) only lists 800 lph

 

https://www.corsair.com/us/en/Categories/Products/Custom-Cooling/Pump-Reservoir/Hydro-X-Series-XD5-RGB-Pump-Reservoir-Combo/p/-WW#tab-tech-specs

 

While this product (CX-9040006) lists the "Theoretical max" like you said toc. Very strange. 

 

https://www.corsair.com/us/en/Categories/Products/Custom-Cooling/Pump-Reservoir/Hydro-X-Series-XD5-RGB-Pump-Reservoir-Combo/p/-WW#streaming-gear

 

I really don't understand it. Are they two different pumps for the same price?  Either way, it does not address the head pressure or wattage consumption difference. 

 

If by some chance there is a Corsair rep stalking my postings or an AI algorithm monitoring the net, I want demand a test sample.  Or you could just hire me to comb over your products and tell you about your typo's. I'm totally fine with double my nursing salary to wear a sailboat tee-shirt. 

Edited February 9, 2022 by Avacado

[ArchStanton]

I went back and checked my original order along with the actual box it came in. If it turns out to make a difference, I ordered and received the "06" version .

 

 

[J7SC_Orion]

5 hours ago, ArchStanton said:

@J7SC_OrionOut of curiosity, your multi-pump loops have the pumps in series for fail-over?  Or you run then in parallel for increased flow with check valves for fail-over?  I tried to see how you had them plumbed in your build log, but I couldn't quite make it all out.

 

Currently, it is more serial than parallel, but it can all be switched around via QDs.  Essentially, there's a common reservoir with 5 in/outlets, so if one pump fails (rare to never), the others continue to pump liquids through the shared loop - and will equalize pressure along the loop. In fact, during initial filling and bleeding operations, I started with only one D5, then powered the second one, and finally added the third.

 

This pic from EKWB is for illustration purposes only since the triple D5 system above uses different D5 pumps (each rated at up to 1500 lph / 4700 rpm) and a different common reservoir.

 

 

Via QDs, I can make my 5950X system a true dual parallel loop system (CPU separate from GPU) with a second identical 5 in/outlet reservoir, but that is for later when these loops will be doing some HEDT again. The loops and rads are designed to handle up to 2000 W of heat energy. For example, each of these maxes at 400W (before the CPU) 

 

 

As an aside, until I got into the 3950X and 5950X, I hadn't build a 'desktop' system since '14, only HEDT (basically, the pro-sumer / light server space).  But AMD and Intel have been slow this time around in bringing out corresponding HEDT parts, and of course AMD has the 3950X/5950X which sit somewhere between regular desktop and HEDT / pro-sumer.

 

With AMD, the Epyc servers now offer up to 8x 'Ryzen 5K type' cores while the corresponding Threadripper/Pro parts are still based on the 3800X cores'...question is whether AMD will offer updated Threadripper/Pro parts based on the 5800X (or even 5800Xv3cache?) or just keep on trucking as is until Zen4 chiplets are called upon for future Threadripper.

 

Intel is similar, nothing new in the HEDT space since the 10980XE, though there are 56 Core Sapphire Rapids Xeons ES around, hinting at a HEDT pro-sumer Saphire Rapids X w/ Raptor cores...

 

Sorry, I digressed - the point is simply to configure a cooling system that has fail-over, very high flow and can be updated / reconfigured / expanded via QDs down the line w/o tearing down the whole loop. 

 

 

Edited February 9, 2022 by J7SC_Orion

[ArchStanton]

So, as I was twiddling my thumbs waiting for the RMA process to complete, I realized my hydraulic test kit at work likely contained a G 1/4 test fitting.  Low and behold, it did.

 

I'll need to acquire a low-pressure gauge to have decent resolution (smallest of my existing is 0-50 psi).  I think I'll pick up something in the 0-10 psi range.  So, I should be able to provide both pressure and flow readings in the not-too-distant future .

 

 

 

 

 

 

[Sir Beregond]

Well isn't that handy, nice lol.