Originally Posted by MrGrievous
hello yet again I have a new question in trying to plan out this build I want to do. Well what I want to know this time which setup wold be better, gpu's in parallel or series? I'm thinking series is bette r:headscrat
Originally Posted by sakerfalcon
Parallel is better, but the difference is neglible.
Originally Posted by lowfat
Actually series is better but negligible. Easier to bleed plus a higher flow rate through the blocks. Twice the pressure drop though, so you'll have lower flow through the rest of the loop. Plus it generally doesn't look as clean.
Originally Posted by wermad
i prefer series because it maintains a higher pressure compared to parallel. This comes in handy when you're running more then two gpu blocks.
Other then that, it makes little differences w/ two gpu blocks. There's no temp difference either (I tested this a few times).
Originally Posted by joejoe69
This is how I look at it:
Parallel - if you got lots of blocks and rads but with a strong pump source.
Series - if you got less blocks and rads and with an adequate pump source.
Originally Posted by wermad
Id have to say its the reverse.
After running quad 480s and quad 580s, series works much better since it maintains a higher pressure through multiple blocks (including mb, ram, and the cpu).
As I mentioned before, if you're only running two gpu blocks (and a cpu), either works fine. If you're running three or more gpu blocks, series does a better job, especially at purging out air.
My advise would be to use which ever is appropriate to plumb and/or for looks when running two gpu blocks.
While it would seem that we're never going to reach a consensus on the serial versus parallel routing of GPU's, let me throw in a bit of thought here, that can make it clear why different setups get different results, and why you have to take more factors into account than just what pump and how many GPU's.
Everyone knows that putting blocks in parallel reduces the total amount of restriction by the number of blocks, 4 blocks equals 1/4 the restriction of 1 block.
Everyone also knows that each block then gets 1/4 of the total flow rate. If you have a flow rate of 1.2 GPM, then each block has a flow rate of 0.3 GPM.
We all pretty much also understand that there is a minimum flow rate that has to be maintained thru a block for it to function efficiently.
Rule of thumb for GPU blocks is ~ 0.5GPM
Now here's where the waters start to muddy up a bit;
It's often believed that reducing the overall restriction of the GPU component of the loop will allow the pump to have a big increase in flow.
While that may be reasonably true in a dedicated GPU loop, it's definitely Not True in a single loop setup.
The CPU block is normally the single most restrictive component in a loop, as a result of that fact, it by and large determines the highest possible flow rate you're going to be able to achieve.
Sure the rads add a bit of restriction, but not nearly the same magnitude of an impingement block.
In a multi-loop setup, where GPU's are on their own loop without the relatively huge restriction of the CPU block, the flow rates can be much higher, so that being able to maintain the minimal flow rate thru each GPU is no longer the problem it was in the single loop.
Where you have a dedicated GPU loop, it really doesn't make any difference whether it's series or parallel, both will maintain plenty of flow rate and possibly an edge, a degree maybe 2, to the parallel setup since all the GPUs get coolant at the same temp coming in.
But back to the most common setup . . . . . the single loop . . .
For up to 2 GPU's, the series versus parallel setup doesn't seem to make any difference.
Even a rig pushing the minimal rule of thumb rate of 1GPM, will make the minimal rate thru each of 2 GPU's.
It's when you get to 3 or 4 GPU's that results become so varied.
Some rigs that have enough pump to be running close to 1.5 GPM or more, could run parallel, and be fine.
Other rigs barely at 1GPM will suffer significantly with more than 2 parallel GPU's.
Quads just up the ante on required flow rate for parallel operation, and for pressure potential for series operation.
I would suggest that the best way to set up triples or quads in a single loop is to combine the best of both in a series of paralleled blocks.
Put the first 2 GPU's in parallel, then series that pair to the next single or pair of GPU's.
Flow rate wise, you'll never have less than 1/2 thru any block, and restriction wise, the system will not see it as any more than 1.5 blocks for the 3 gpu setup, and 1 block for a quad.
Anyway, that's my two rational cents,
Edit to add:
If you look at what EK has available for the FC Terminal style blocks, you'll see exactly what I referred to in what they call "semi-parallel":
They don't even show a quad parallel.Edited by IT Diva - 5/4/13 at 7:54am