My new Build Project, with

[Helios451]

Today, another job possible 'sans processeur' is sorting out the passive thermal efficiency of the X570 chipset heatsink and fan.

I've read just too much dissatisfaction on different forums about chipset fans running noisily. The position of the heatsink arrangement on my 'carte mère' (yes, that's French for motherboard!) is also close to being overhung by the end of the GPU, with a potential associated reduction in airflow.



So, I decided to see if I could upgrade in order to improve the passive thermal efficiency and reduce the likelihood of the fan needing to start up.

First of all, the heatsink is screwed to the PCB with four screws. None of these can be easily accessed as they are all underneath the back plate of the mobo. So, first job, remove all back plate screws.



Exposing this:



These are the four screws that hold the chipset fan and heatsink assembly onto the mobo:



Having removed the screws I turning the board over and gently removed the heatsink.





The existing heat transfer material was a little pad of plastic, which had been so tight onto the chip that the lettering was embossed onto it:



The old pad was peeled off:



A quick visit to the local chemist, adhering to 2 metre distancing etc, yielded some good cleaning medium:



Which I used to clean the two relevant surfaces:




I appraised myself of the right way to use the thermal grease:



and proceeded to cover the chip:




I then just put it back together - the heatsink assembly has four feet that slot into the screw holes in the mobo so it's easy to locate, hold in position and screw down. The only wrinkle was the little fan power cable insisted on looping across and blocking one of the screw holes, it just needed to be persuaded to move over a bit.

Job done

[ik9000]

Those connectors are supposed to be maximum 7A per terminal. 8 conductors is 4 pairs (what comes out one yellow wire loops back in a black ground wire) so that's 4x7 = 28A total, which at 12V would be 336W. That's the point you risk melting it, but I expect for most of us our cpu cooler will have given up before we hit the limit of a single 8 pin connector.

In case the internet lied to me about the current rating of that molex connector (the Internet is always right, isn't it?) a rough estimate is that the rather similar PCIe 8 pin PEG connector is conservatively rated at 150W, but one of the connectors is a sense line not for power delivery so you only get three pairs from the possible 4. So scaling the 50W per pair for 6 active cables an 8 pin connector should allow 200W.

So that single 8 pin connector should deliver 200W all day no problem, possibly up to 300W. Driving even the bigger Ryzen chips from an old 4 pin EPS connector should be possible, but peaks will be on the limit.

If you intend using PBO overclocking where the motherboard will tell the CPU that it can deliver an unholy deluge of power, then it might be wise to connect it all up so it isn't lying. But then most of us don't overclock.

OTOH, if you have two 8 pin power connectors then I think it would look nice to have them both in use. But then I drive a car that really doesn't need twin exhausts at the back

I agree - as I said it can run off one, but the mobo manual says use both - why I can't explain, but you can see the warranty rejection coming "did you use both ports?"

[MrJim]

I agree - as I said it can run off one, but the mobo manual says use both - why I can't explain, but you can see the warranty rejection coming "did you use both ports?"

Having watched a lot of Buildzoid videos, I understand that a standard 8 pin motherboard connector can carry up to 384W, and for almost all purposes (perhaps with the exception of extreme overclocking using liquid nitrogen), that's enough without the need to use and additional 4 or 8 pin connector.

[ik9000]

yup but I'd hate to be the poor person who gets a claim rejected because he only used one.

[Ciber]

yup but I'd hate to be the poor person who gets a claim rejected because he only used one.

If you were taking the heatsink off the mobo you might not care about invalidating the warranty though!

[Helios451]

If you were taking the heatsink off the mobo you might not care about invalidating the warranty though!

Yes, but I've been careful to stow the old heat transfer material (that square of plastic) on the underside, so if there are any issues, it'll be simple to return it to it's prior position.

Also, I'm happy to admit that I bear full responsibility for my actions, so if I do something that invalidates a warranty, I do so happy that I can risk the consequences.

[ik9000]

how was your thermal grizzly kryonaut? Any abrasive grains in there?

[Helios451]

how was your thermal grizzly kryonaut? Any abrasive grains in there?

Thanks for asking, I tested it beforehand as you suggested and no grains at all, nice and smooth, if a little sticky.

If you compare these two pictures you can see how it spilled over the edge of the chip at the beginning - just a bit too much pressure on the tube gave just a little too much paste in one place. I then did quite a bit of tidying up to make sure there wasn't excess around the chip, which is where I found in to be sticky, and the result is the second pic.

[Helios451]

A quick update on the CPU delivery, which I was last told by phlegmatic Stefan would come around June 16th.

I had been monitoring the situation and picked up that the company supplying the CPU (Xstra Digital Storage of the Netherlands) was advertising the 3900X on Amazon for delivery next week and also showing it as in stock on its own website. I therefore called the head office again and once more spoke to the phlegmatic, almost to the point of being taciturn, Stefan.

I pointed out my observations and asked whether there was any update on my order. Stefan went away to talk to a colleague and returned to say, in his slow and quiet voice, that my order was able to be dispatched today and would be handed over to DHL between 7 and 8 this evening for delivery in the next two days.

Well, will wonders never cease or will pigs fly. Until I have an update from DHL, I have no idea which!

[ik9000]

patience young grasshopper! Time will reveal all.

[Helios451]

patience young grasshopper! Time will reveal all.

My head is bowed in patience, Master

[Helios451]

Patience is a virtue to be rewarded in the fullness of time:



So, wonders never cease and pigs have flown...

[Helios451]

Well, at around midday DHL delivered this:



Containing this:



At last!



Separately, my new PSU arrived:







Tonight, the bench test!

[Xlucine]

Not really with modern PSUs which are designed to work like that. Old ATX PSUs, sure.

It'll work, but will lose more efficiency going from 50% to 10% load than going from 50% to 100%. Hexus reviews show only a couple of percent difference between 50% and full load, but ~10% less efficiency at 40-60 W

These are figures for the 650W PSU I've ordered:

80 PLUS certification - Platinum

Efficiency at 230V, 20% load - 92.4%

Efficiency at 230V, 50% load - 93.7%

Efficiency at 230V, 100% load - 92.6%

The 80+ spec only used to cover down to 20% (so 130 W for your unit, whereas a modern machine under low load will be more like a third of that). The newest titanium rating helps, specifying the efficiency down to 10% load, but even then with looser requirements than at higher loads. A modern PSU won't be bad at idle by any measure, but if we're talking theoretical ideals...

[Helios451]

I'v done the bench test and all running ok - a different post later on will cover that with plenty of pics.

First, thought, is the question of the RAM - if anyone recalls, I had two sticks of 8Gb Corsair Vengeance LPX 3200MHz supposedly rated for Intel and another identical pair rated for Ryzen. The plan was to see which would run better and whether they could run in a four set.

Having run then both through Thaiphoon, the interesting thing is that both sets are Samsung B-die but they have remarkably different settings:




Am I right in thinking that I won't be able to put the two pairs together, despite them having identical XMP settings as the supply voltages and speed grades are different?

Cheers

[ik9000]

you might get away with it. they are both stating an XMP row of 1600 with the same timings so that would give you 3200 operating speed. I'd give it a shot OR do an extensive ebay hunt for a matching batch no to one of the sets (though they probably won't be on the 2nd hand circuit for a while generally). I'd also try dropping down to CAS 14 at 1600 and see if you can get that to work. There is little net difference between 3200 at CL14 and 3600 at CL16.

btw stated supply voltages appear to be identical in those lists. Manually set RAM voltage to 1.35 (or try 1.30V first and go up from there - mine is happy at 1.31V or thereabouts IIRC)