Question New Phanteks G6000 waterblock not performing as expected.

[Justinus]

I'll start with my expectations for how the block should perform are based on my experience having a 6900XT reference card with an EK Quantum block for a while, where at normal overclock loads (340-360W) it would see mid 40's for edge temp and around 10C delta to hotspot, around mid 50's.

I've been waiting for a block to come out for the 6900XT Strix that I got, and since Phanteks was first to market and seem to be a reputable brand, I ordered it as soon as I learned about it.

I installed it last night, and noticed two things - there is a deep (for what is supposed to be flat) machining line that runs across the coldplate and also the heights of the screw standoffs don't seem to be consistent, so the card is bowed out at the die.

Nonetheless, the mounting looked good and all the thermal pads make contact so I installed it into my loop.

Sadly, upon testing it, I see edge temps that are somewhat in line with the old EK block at similar thermal loads, but a little higher. I'd probably be OK with that if the delta to hotspot was still around 10C....

But it's not. Hotspot regularly runs 20-30C hotter than edge. The highest delta I observed was 36C! It really makes me think that machining line is preventing the die from making good contact along it, or the uneven standoff height causing it to bow out is lightening the mounting pressure in places.

I plan to reach out to Phanteks, and also take the block out this weekend and dismount it to get a look at what might be the culprit.

Am I making a mountain out of a molehill? Is it too much to ask for a Phanteks block, which costs the same as an EK block, to perform similarly and have similarly good die mount/coldplate flatness?

Sample 360W load comparison
EK: 45C edge/55C Hotspot
Phanteks: 48C edge/72C hotspot

If I run the card to the ragged edge with a 470W load, edge is 57C and hotspot is 93C! The fact the hotspot delta grows with increasing load really makes me wonder.

In the same loop, similar ambient, similar coolant temps.

 

[Justinus]

Running port royal stress test, where the old EK/6900XT would still maintain a 10C delta, it's stabilized at 52C edge and 81C hotspot, a darn near 30C delta. Sigh.

 

[aigomorla]

The differences should not be that great...

I would try a remount... or inspect mounting pressure or warpage.
Yes without a doubt eK is a better block.
But load values should not deviate that great unless something else is bottlenecking your system, and that block somehow increased the bottleneck even more.

 

[Justinus]

The differences should not be that great...

I would try a remount... or inspect mounting pressure or warpage.
Yes without a doubt eK is a better block.
But load values should not deviate that great unless something else is bottlenecking your system, and that block somehow increased the bottleneck even more.

I removed the waterblock and the mounting looked perfect. I remounted the OEM AIO cooler and the delta is generally 15-20C.

I contacted Phanteks and they immediately agreed it seemed like a defect and they're sending a new block from a new batch they will receive next week. I guess I'll hope it does not share the same defect, being from a different batch.

Another guy I know bought the same block the same time I did and has the same defect, and is getting a replacement from a new batch as well.

I'm pretty sure any competent cooler should see 10-20C edge to hotspot delta, not 30C.

 

[Justinus]

With regards to bottlenecking, there aren't any. I massively overbuilt this box intentionally. 3x360mm HWLabs GTX nemesis rads, EK dual D5 serial pump for higher pressure, 12 NFA12x25 fans on the rads, water temp sensor with fans calibrated to keep water temps below 30C.

I use GC extreme paste, which is top 3 pastes IMO with Kryonaut and Mastergel Maker.

Also I was not exaggerating about the 6900XT reference with EK block being in this exact same system - I quite literally just used a connector to stick the GPU hoses together when I took it out, and unconnected then and hooked them up to the Phanteks block.

 

[Justinus]

Yes without a doubt eK is a better block.

Is this true? I had both blocks in my hand, side by side and I can't say I agree. The EK and the Phanteks both shared similar superficial machining lines on the coldplate (excepting the one large line we're calling a defect). The EK and Phanteks both have 0.5mm fins at a 0.5mm fin pitch. The acrylic machining looks better on the Phanteks. The coldplate and acrylic are thicker on the Phanteks, possibly contributing to less coil whine. The fin core is absolutely massive on the Phanteks at 53x50mm vs. 32x24mm, but I'd have to see the flow simulations to be able to tell if the flow across a smaller fin core is more effective or if the flow across a larger fin core is more effective.

I'll update once I receive, inspect, and mount the new Phanteks block. So long as it has an acceptable hotspot delta, I don't see why it wouldn't be a better block than EK. Excepting quality control issues, that is.

 

[aigomorla]

Is this true?

most definitely.
I can go on for probably half a page on why the eK block is superior except the nickle plating job.
But as for performance, you can automatically see that just from looking at the fact the eK uses an injector at the core base plate for increased turbulence, while phantek does not.

 

[Justinus]

A brief follow up:

Phanteks allowed me to return the second block they sent, which performed identically to the first, which is to say poorly and very high hotspot delta. They gave a full refund and didnt give me any hassle whatsoever, so props to them.

I ended up ordering the EK block even though I desperately wanted to explore other brands. I wish alphacool made a STRIX block but they only released one that fits the 6800XT STRIX and cannot fit the 6900xt due to the third 8 pin power connector and some components being moved due to it.

The EK block performs excellently and I am ashamed to say I should have known you were right, aigomorla. It's like 10-15C lower average temps and 15-20C lower hotspot temps under the same conditions as the Phanteks block. At least I can sleep now knowing I have something performing as good as I'd hoped.

 

[solidsnake1298]

In a hypothetical situation where Phanteks did not give you a refund, perhaps you could have lapped the cold plate. You shouldn't HAVE to, but I wonder if a good sanding would have improved temps.

 

[aigomorla]

In a hypothetical situation where Phanteks did not give you a refund, perhaps you could have lapped the cold plate. You shouldn't HAVE to, but I wonder if a good sanding would have improved temps.

i think his problem lied in the actual alloy.
Lapping would not have helped, only if the block was uneven or not flat.

And lapping would also destroy that Nickle plate job.

 

[solidsnake1298]

i think his problem lied in the actual alloy.
Lapping would not have helped, only if the block was uneven or not flat.

And lapping would also destroy that Nickle plate job.

I was mostly referring to the edge vs hot spot temps. The fact that the hot spot temps were way higher than the edge temps indicates to me that there was a flatness problem. If it was just a matter of the alloy being less thermally conductive, the delta between edge and hot stop temps would still be much closer together.

Obviously lapping would remove the nickel plating. But if given the option between improving temperatures with the block I am stuck with or maintaining the superficial appearances, I will always choose the former. Other than appearances, what functional purpose does nickel provide? Prevents copper oxidation? Sure. But the part of the block that is in contact with the GPU is sealed off from the air by the thermal paste. You could also argue that it is more resistant to staining when used with liquid metal (not applicable here). Again, a purely superficial benefit.

Also, what do you mean by "the problem lied in the alloy". Aren't both the Phanteks and EK blocks made from nickel plated copper? Nothing on either of their product pages indicates that they use anything but pure copper. Wouldn't the overall temperature deltas come down to the microfin design, density, surface area, and water flow restriction?

 

[solidsnake1298]

Also, isn't it possible that the uneveness also affected edge and overall temperatures? Poor heat transfer in a portion of silicon would heat up adjacent portions is my thinking.

 

[Justinus]

I think it was a combination of the coldplate still not being flat (it didn't have a hard line but it did have a few obvious circles still), the fin core being too big (leading to slower water flow, essentially wasting the water flowing across parts that aren't in contact with the die), the design not utilizing an injection plate or some other method to force the water into the fin core evenly, and the fin core being the wrong direction/orthogonal to the short dimension of the die.

It was clear the Phanteks fin core design was much more affected by flowrate, but even with pumps maxed out it didn't perform as well as the EK block at 40% pump speed.

At any pump speed below ~80% it performed about as well as the AIO the card came with, which is awful and totally unacceptable. Maybe some people build custom loops without performance in mind but I certainly don't. I expect high end/expensive custom waterblocks to significantly outperform an off the shelf, cheap AIO.

My backup plan was to lap the coldplate if I was stuck with it, it probably would have improved the edge to hotspot delta but I doubt it would have made a significant difference to edge temps. I think that's largely due to the factors I mentioned above.

Under a torture test at 470W, the Phanteks block would see 65C edge and 95C hotspot where the EK block sees 45C edge and 70C hotspot.

 

[aigomorla]

Also, what do you mean by "the problem lied in the alloy".

Wouldn't the overall temperature deltas come down to the microfin design, density, surface area, and water flow restriction?

A chinese oem vendor which supplied the metal could of cheezed on the actual composition.
Could of had a higher zinc content, or it could of been a straight up covered piece of brass.
I have seen issues with alloy back when i was active.
Silver Coils were notorious for being tainted and fake, which completely destroyed the market for them as you could not trust which vendor used pure 99.999 silver and not Sterling or a cheaper quality silver. Tainted silver kill coil meant to keep your distilled system in check ended up corroding and gave vendors an excuse to deny RMA on cheap nickel flaking job.

A block is made by balancing flow with turbulence.
Increasing flow increases the efficiency of water.
Increasing turbulence increases the efficiency of the heat transfer. (injector plate)
Microfins, extend surface area to allow more water in contact with heat plate. as well as waterways.

So you need to balance these 3 to make a block.
If you have too much flow, it usually means to little turbulence, unless you have a LOT of pumps in serial and your pumping out a high pressure system.
If you have too much turbulence, you have too little flow making the water overall not as efficient.
If you lack surface area, well, you get the point.

Block contact with IC's also play a big role.
Normally there are ways to test the block, like getting water coolant temp, and also looking at TIM imprints on the block.