2xD5 vs 2xDDC long loop comparison (serial scaling)

[luke997]

I though it will be worthwhile to share in a separate thread my observations of flow with 2xD5 vs 2xDDC (both with EK tops, serial), perhaps someone will find it useful (or not )

On system - let's call it "A" (see my worklog for details), with 2xD5 with EK top I was getting 200l/h.

With the following changes:
Change from MCR200 to RX360 (less restriction)
Change from serial GPU to parallel GPU (less restriction)
Change from 2xD5 EK top to 2xDDC EK top (3.25 - branded as Koolance PMP-400)

, flow is 205l/h

With the 1 pump on, 1 pump off the D5 did 130l/h while DDC does 160l/h.

So my 2 obvious observations are:

1) D5s scale much better in series - as 2xDDCs are not better for long loop than 2xD5s (and in fact they might be a bit worse if you account more than 5l/h improvement to the rad swap & parallel GPU)

2) 1 DDC is indeed better for long loop than D5 but it does make sense to add 2nd pump for redundancy/flow - so see (1)


Please note there are no scientific methods to the above - except for the flow meter observations - nor it is any testing - it is just observations of 2 specific scenarios on my system which may benefit other in similar situation.

[NaeKuh]

1) D5s scale much better in series - as 2xDDCs are not better for long loop than 2xD5s (and in fact they might be a bit worse if you account more than 5l/h improvement to the rad swap & parallel GPU)

Thats because u lacked the restriction component.

In a open flow, a D5 is better.
Once u start connecting things down the loop, the DDC's head pressure works out the restriction and pushes water.

The longer the loop and more complex it is, you LOOK at headpressure more then flow.

[PiLsY]

Yeah youve made your loop better suited to the D5's then replaced them with DDC's. Go back to the D5's with that setup.

[luke997]

Thats because u lacked the restriction component.

In a open flow, a D5 is better.
Once u start connecting things down the loop, the DDC's head pressure works out the restriction and pushes water.

What do you mean by lacking restriction component?

This is a closed loop with 7 waterblocks, 3 radiators and ~ 1m of tubing, there should be enough restriction for DDCs to shine.

[Johnny87au]

I thought having more of a restrictive loop the ddc 355 pumps actually work better as they have more head?? Also luke its been proven that running ur gpu's in serial or paralell basically give and take the same Temps, Not to sure though if it helps with restriction?

[luke997]

I thought having more of a restrictive loop the ddc 355 pumps actually work better as they have more head??

Yeah I thought the same - and it is certainly the case in my complex loop with 1 DDC (160l/h vs 130l/h) - but with the 2 pumps D5 scale much better than DDC - and with D5s being quieter it is better option (unless you struggle for space).

Also luke its been proven that running ur gpu's in serial or paralell basically give and take the same Temps, Not to sure though if it helps with restriction?

GPUs in parallel should help the flow.

[Johnny87au]

Biffa has just updated his rig from running 3x 480s from serial to paralell and he mentioned his temps were the same hmm.. Well yeah i guess .. Only reason im using 2x 355's is because all my loops have been quite restrictive and using the dual XSPC top which makes it into 1 pump but really your using 2 is just super easy..

[luke997]

Biffa has just updated his rig from running 3x 480s from serial to paralell and he mentioned his temps were the same hmm.. Well yeah i guess

It could be the case, from Gabe's serial vs parallel testing there was only a tiny difference in GPU's and a little in CPU's (like ~1C in his example) temps

Only reason im using 2x 355's is because all my loops have been quite restrictive and using the dual XSPC top which makes it into 1 pump but really your using 2 is just super easy..

My numbers comes from similar setup - restrictive loop and both D5s and both DDCs were in the dual top (although EK not XSPC).

Would be interesting to see more results of D5 vs DDC in dual tops in restrictive systems.

[Johnny87au]

It could be the case, from Gabe's serial vs parallel testing there was only a tiny difference in GPU's and a little in CPU's (like ~1C in his example) temps



My numbers comes from similar setup - restrictive loop and both D5s and both DDCs were in the dual top (although EK not XSPC).

Would be interesting to see more results of D5 vs DDC in dual tops in restrictive systems.

Interesting dude interesting! More restrictive loop the ddc wins by a few %, cant remember which thread but martin wrote it.. *edit found it*
D5 with top = 3600 Mhz
DDC3.2 with top = 3603 Mhz

[luke997]

Interesting dude interesting! More restrictive loop the ddc wins by a few %, cant remember which thread but martin wrote it.. *edit found it*
D5 with top = 3600 Mhz
DDC3.2 with top = 3603 Mhz

Was it dual or single? Any chances you could post a link? Thanks

[Johnny87au]

Yeh dude here ya go http://www.xtremesystems.org/forums/...=168059&page=4 , I couldnt find the exact thread as i have like 500 bookmarks haha, but yeah their ya go mate

[Waterlogged]

What do you mean by lacking restriction component?

This is a closed loop with 7 waterblocks, 3 radiators and ~ 1m of tubing, there should be enough restriction for DDCs to shine.

Yes, but 3 of those 7 are acting as 1 block so your loop actually thinks it has 4. Had you stayed serial and with your fancy elbow setup, the DDC's would be much better. You basically streamlined your loop to the point were it doesn't matter which pump you use now.

Funny thought (and I'm not making you or asking you to do it, just putting it out there), I wonder what the DDC's would do if you put them in parallel.

[luke997]

Yeh dude here ya go http://www.xtremesystems.org/forums/...=168059&page=4 , I couldnt find the exact thread as i have like 500 bookmarks haha, but yeah their ya go mate

Thanks, looks like this is a single though and this is rather for the dual comparison.

[Johnny87au]

Well yeah but seeing as its a single im sure even as dual it would run better, Use the DUAL top which makes it into 1 top, it rules..

[Dreddy71]

that is an interesting idea...
/Pump 1\
Loop< > loop
\Pump 2/
if i had 2 pumps and some stuf to test it i would but i dont. me = fail

[Johnny87au]

that is an interesting idea...
/Pump 1\
Loop< > loop
\Pump 2/
if i had 2 pumps and some stuf to test it i would but i dont. me = fail

Dont be so hard on yourself mate

[Church]

Waterlogged: IIRC dual pumps in serial = double head pressure, in parallel = double flow. DDCs have loads of head but less flow, who knows, maybe might show more flow that way .. but somehow it seems that you'll loose redundancy that way (as in if one DDC stops, big part of flow might go through failed pump subloop because it seems to be way of least resistance, instead of going all the way through loop with many resistive waterblocks. Should be easily workarroundable though by putting ball valve there to switch off that failed pump subloop though.

[Waterlogged]

Waterlogged: IIRC dual pumps in serial = double head pressure, in parallel = double flow. DDCs have loads of head but less flow, who knows, maybe might show more flow that way .. but somehow it seems that you'll loose redundancy that way (as in if one DDC stops, big part of flow might go through failed pump subloop because it seems to be way of least resistance, instead of going all the way through loop with many resistive waterblocks. Should be easily workarroundable though by putting ball valve there to switch off that failed pump subloop though.

Exactly, it would double the flow of the DDC's. They are already strong enough to keep the water flowing through the loop on their own so putting them in parallel should kick them into overdrive on the flow rate. As for redundancy, in either case (serial or parallel), there would still be 1 pump running in the loop which should provide more than enough to keep it running until a safe shut down can be initiated.

A good parallel setup for Luke would be something like this (not sure what his current order is, I went off older pics in his log to determine this order).

[luke997]

Yes, but 3 of those 7 are acting as 1 block so your loop actually thinks it has 4. Had you stayed serial and with your fancy elbow setup, the DDC's would be much better. You basically streamlined your loop to the point were it doesn't matter which pump you use now.

I think I'm lost here - would appreciate more input from you if I may ask for it

What would you exactly have in mind by saying streamlined?

The 200l/h is like ~0.8gpm so less than I would have expected - and that can be caused only by the restrictions in the blocks (can it?).

Funny thought (and I'm not making you or asking you to do it, just putting it out there), I wonder what the DDC's would do if you put them in parallel.

Would be interesting indeed
I'm sorry though - not going to do it in the short term - the temps are good, I'm happy my rig is working now and I've reached the point where after all the building and changes I want to bench it till it sweat. And then finally play some games :-)

Well yeah but seeing as its a single im sure even as dual it would run better, Use the DUAL top which makes it into 1 top, it rules..

That was the whole point of my post - to share information that in my loop 2x D5 scales much better than 2xDDC - even though as single DDC was better in it.
(much like the SLI, you may have let's say 5870 faster than 470 but 2x470 will be faster than 2x5870 cause they scale better in most title - just an example though, I'm not trying to discuss this here )

[luke997]

Exactly, it would double the flow of the DDC's. They are already strong enough to keep the water flowing through the loop on their own so putting them in parallel should kick them into overdrive on the flow rate. As for redundancy, in either case (serial or parallel), there would still be 1 pump running in the loop which should provide more than enough to keep it running until a safe shut down can be initiated.

A good parallel setup for Luke would be something like this (not sure what his current order is, I went off older pics in his log to determine this order).

Thanks for suggestions and diagram.
Changes look actually much easier than I though there would be (no need for another res, etc.), so I actually might pick up some fittings and single tops in the next couple weeks out of curiosity.

That's pretty much the order I have - so basically what you're saying is that there is not enough restriction in the loop for the serial setup to benefit it and the parallel setup would do the trick and improve the flow significantly?

Also, considering what the current values and scaling in series are any ideas what numbers for the parallel might be?

For example currently:
1 D5 = 130,
1 DDC = 160

2 D5 series = 200
2 DDC series = 200 (I've adjusted from 205 for GPU serial to parallel change)

2 D5 parallel = X?
2 DDC parallel = Y?

I realise it's hard to tell with that many variables but perhaps interpolating from some existing tests I could have some idea?

P.S. Sorry for double post it - didn't notice this post before replying earlier.

[Church]

Waterlogged: but without closing one of subloops in parallel, won't most of flow go skipping resistive-lot-of-waterblocks-loop part, when one of pumps fails? In serial pump connection it has no way to skip, not so with parallel without closing some valve or pressing tubing in it to close it. So one three way ball valve, or 2x two way ball valves or something like this @bad business practices shop.

[TJ TRICHEESE]

yea but he said earlier that the flow left would be enough to keep the system from killing itsself at idle untill the pump can be replaced

[Church]

Was such parallel pump setup with one pump switched off ever tested in real life? Or it "seems" that it will be enough? If second case, then i can counter with that it "seems" to me, that most flow will go skipping loop with waterblocks. I'm thinking this way because recalling few previous threads where paralelising parts of loops issue had been brought up, and recall vets saying to not paralelise loop parts with different waterblocks of different resistance, as almost all water will go to subloop with least resistance, leaving other one with almost no flow, and how hard it is to evenly balance hidrodynamic resistance with different waterblocks in subloops.

[Martinm210]

Fyi, 200lph is only .88gpm? That's not a really low restriction loop so I'm not sure.

As far as single pump vs series, the only way to really visualize that is by looking at the curves. D5s are more of a high flow lower pressure oriented pump. It just happens that most watercooling loops today are generally higher in restriction which is better matched to the ddc curve, but it is possible to have a low enough restriction loop to slightly favor the D5, and its also possible to have a low enough restriction to favor pumps in parallel vs series. Its just very unlikely and not typical at all.

[luke997]

Fyi, 200lph is only .88gpm? That's not a really low restriction loop so I'm not sure.

Couldn't agree more, that was behind my drive toward better through through various upgrades

As far as single pump vs series, the only way to really visualize that is by looking at the curves. D5s are more of a high flow lower pressure oriented pump. It just happens that most watercooling loops today are generally higher in restriction which is better matched to the ddc curve, but it is possible to have a low enough restriction loop to slightly favor the D5, and its also possible to have a low enough restriction to favor pumps in parallel vs series. Its just very unlikely and not typical at all.

So to summarise, my loop is so untypical that no changes may guarantee any decent flow increase and it's pretty much hard to estimate which change would give what improvement.

This perfectly describes my previous revisions, but hey, it's all fun isn't it (well, up to a point )