[Review] Round 2: Swiftech Apogee XT Tested

[NaMcO]

What is this silicon trick exactly?

What dreamaxx said. I too am curious about this, the image doesn't really show

[Vapor]

See the white stuff in the center? That's 100% silicone caulk that I added. RTV silicone would also work. Between caulk and RTV varieties, a lot of colors should be available (clear included), I chose white since it's easiest to photograph. The 100% silicone part is also important...doesn't have to be anything specific, just as long as it's silicone and not acrylic or anything junky like that.

To do the mod, just gob some on, spread with a q-tip over the entire center section, and then use the straight-edge of a q-tip to remove the excess in one sweeping motion. Then go around the edges with another q-tip and clean up the edges at the inlet and at the 'moat'

[Humminn55]

I would not do that.
As far as I see the block is not worth replacing.
The block is more restrictive and is barely better for a 0,9°C.
It is a great block, I dont say that, but think about before doing this

Well, while I do have an HK 3.0 on my cpu, this block is really making me think of switching and not simply because of the temps. It's more a combination of the temps, which are not drastically different as you noted, and the vastly superior mounting system on the Swiftech block. Doing remounts with the HK is so tedious and horrible, I'm really thinking of dumping it and replacing it with this XT.

That's why one should consider the XT.....and keep rejecting the stupidly designed mounts of other blocks on the market until the other manufacturers produce mounting systems as simple and elegant as Swiftech's mount.

[shazza]

+1 to Humminn55's comments - I've kept my GTZ going just because it is so darned easy to work with the mounting system (and temps are fine, if not the best). Great to know we now have an even better alternative with the easy-to-use mounting mechanism.

[Manicdan]

can you explain a little more about the silicon mod, if you can get a full degree of performance, why is this mod not build into the blocks.

[Sadasius]

rge's tests?

Yup that was the one. Good response too btw and thankfully I don't lap my CPU's anyway! I just don't see the point if you have a good block, excellent TIM spread and enough pressure. I am sure I can squeeze even more out of this block with a little extra pressure.

Also since you used the silicon trick and it worked just like the HK that it would mean water was escaping the same way when bolted down and the plate flexed.

[Vapor]

can you explain a little more about the silicon mod, if you can get a full degree of performance, why is this mod not build into the blocks.

Yup that was the one. Good response too btw and thankfully I don't lap my CPU's anyway! I just don't see the point if you have a good block, excellent TIM spread and enough pressure. I am sure I can squeeze even more out of this block with a little extra pressure.

Also since you used the silicon trick and it worked just like the HK that it would mean water was escaping the same way when bolted down and the plate flexed.

I know how the silicone mod helped the HK: it forced all the flow through the microchannels (huge since I know a lot was going OVER the channels, which is useless) and it increased the bow (probably less huge of a gain in perf).

With the XT I'm a little less sure. There's already an o-ring there that should help with both aspects (bowing and forcing flow through the pins), maybe that's why the gains are smaller with the XT. When I disassembled, there was a clear pin pattern in the silicone (with the pins basically puncturing all the way through to the metal), meaning that there's not much flow going over the pins in stock guise. Maybe it's a combination of the more aggressive bow (it's a visible change) and increasing the velocity of the fluid at the very bottom of the "forest of pins" (10% drop in overall flow with a ~25% drop in cross-section area = 15-20% higher velocities)

[voigts]

Very nice work as usual. Roughly a 3c improvement over the GTZ is quite an accomplishment.

[Humminn55]

Very nice work as usual. Roughly a 3c improvement over the GTZ is quite an accomplishment.

And Vapor's testing almost exactly replicates the test data presented by Swiftech through their own testing. Guess this possibly will shut up all the "manufaturer data and testing can never be trusted" comments like were made or implied in the original thread about the XT by Gabe.

[Ketzer7]

Thank you for the great review Vapor.

Dumb question I was thinking about.

Given your measurements of the minimal flow reduction between the regular XT and the XT 'Alt' (the inlet plate rotated), would you expect to see the same/less/more amount of flow reduction with the inlet plate in its default installation but using a 45 elbow at the inlet instead?

Seems it would be a minuscule difference, but a 45 at the inlet may play better with my planned loop setup versus going straight in with the inlet plate rotated to clear the comp fittings..

[Fafeifa]

And Vapor's testing almost exactly replicates the test data presented by Swiftech through their own testing. Guess this possibly will shut up all the "manufaturer data and testing can never be trusted" comments like were made or implied in the original thread about the XT by Gabe.

So far there are only 2 tests been done, i am not saying vapor is cooking the numbers as he wouldn't have any benefit from doing so, right?

I would just wait for more reports from people with regular set ups, i mean 2 tests and 1 the HK wins the other the XT wins so it wouldn't be a bad idea to wait for more reports. Specially because vapor's test or rge's test config may be favoring either block.

[Sadasius]

So far there are only 2 tests been done, i am not saying vapor is cooking the numbers as he wouldn't have any benefit from doing so, right?

I would just wait for more reports from people with regular set ups, i mean 2 tests and 1 the HK wins the other the XT wins so it wouldn't be a bad idea to wait for more reports. Specially because vapor's test or rge's test config may be favoring either block.

I don't know if you mean to come off as insulting or not but to me it is. Between the two tests if you have read them you will see the similarities and also the problems with the hardware that gave the results they did. Both reviews are pretty spot on. On had a lapped cpu and the other did not. On used the HK LT and the other the HK CU.

[Fafeifa]

I don't know if you mean to come off as insulting or not but to me it is. Between the two tests if you have read them you will see the similarities and also the problems with the hardware that gave the results they did. Both reviews are pretty spot on. On had a lapped cpu and the other did not. On used the HK LT and the other the HK CU.

That is why i said you can't conclusion with just 2 reports, when the hk came out a lot of people was skeptic about the results. I don't mean no harm to no one specially to someone who is putting all the hard work. I truly believe vapor is not cooking the numbers but if you do is your own problem. The difference between the CU and LT are minimal if any.

[KURTZ]

thx for the review Vapor, nice job

[millertime359]

Well, considering the Swiftech mount is not really optimized for a lapped CPU, that would probably be the reason for the difference. One test, CPU was lapped, other test it wasn't.

[Titan7171]

Is this just for Intel? I didnt see any mention for AMD?

[millertime359]

Is this just for Intel? I didnt see any mention for AMD?

The AM2/AM3 bracket is in production.It will also include mounting holes for Xeon (771) server class platforms.

ETA is 4 weeks.

Here you go.

[JoeBar]

Thanks again for a great work!


However one thing has got me thinking. In your results XT is more restrictive than GTZ despite Swiftech's claims that it's less restrictive with original top config, on a D5...

[Titan7171]

Here you go.

sweet

[Salamndar]

Thanks again for a great work!


However one thing has got me thinking. In your results XT is more restrictive than GTZ despite Swiftech's claims that it's less restrictive with original top config, on a D5...

I for a moment thought the same thing. But again Vapor explaining eased my mind a little.

[Vapor]

I for a moment thought the same thing. But again Vapor explaining eased my mind a little.

I really don't have a good explanation as to why I'm getting numbers that the XT is more restrictive than the GTZ and they're seeing the opposite. That said, as I explained, flowrate itself isn't the important metric--what flowrate does in terms of temperatures is what's important and with that metric, the XT excels


Change of topic...
It seems a few people have missed my educated guess as to why rge is seeing different performance than me--it has nothing to do with Copper vs. Delrin top (though I can now confirm I'll be testing the Copper top effects with HK and the Supreme!), but rather much more dire for some prospective XT owners as it seems to be an incompatibility with the XT + LGA1366 socket when using a lapped CPU. Basically the structure of the socket assembly is preventing good contact on a lower profile CPU (i.e., lapped) because the base of the XT is so big.

Again, here's my longer explanation copy and pasted from an OCF post:

I saw rge's results (he mentioned a weird increase on two cores on XS--something that shouldn't happen with a block change) and I think I figured out why those two cores are oddly hot.

Unfortunately it hints that the XT is not compatible with lapped CPUs after all--it stems from the same basic issue that caused the Gigabyte issues, the block is big. 60x60mm is big and *I think* it breaks Intel's "no-fly zones."

Anyway, here's what I've found: first, Cores 2 and 4 are actually adjacent and are either the top two cores or the bottom two cores, I forget exactly. I don't know why it's done like that :screwy:

Second, when you lap a CPU, you make it shorter by a fraction of a millimeter, which may not seem like much, but it comes into play.

Third, the Torx screws on the socket assembly are not flush with the socket assembly--the top ones definitely stick out quite a bit and the bottom ones (once the socket is loaded with a CPU forcing the internal springs to compress) do a little bit as well.

Fourth, unlike the HK3.0, the Apogee XT overlaps both sets of socket screws and by a few mm.

Add that all up and I think the Apogee XT is not sitting right on his CPU (or really any lapped CPU) because it's sitting on one set of the socket assembly screws (not sure which, my guess is the top set, they protrude the most) and that's forcing the XT to be lifted off the (either the top or the bottom of) the IHS and resulting in increased temperatures on the corresponding pair of cores.

When he gets back from vacation, he can test this by either removing the socket assembly totally, or grinding down the four Torx screws on the socket assembly (or by grinding out grooves on the base of his XT, but that sounds like the hardest option of all). Or if he has a non-lapped CPU, he can do one or two really quick mounts of both the XT and the HK to see that the core temperature profiles are indeed similar (then go on to recompare if he wants to, he probably won't want to after a vacation, lol).

Here's what my core profiles were like with the two blocks.

XT = the Gigabyte-compatible base (I use a Gigabyte board for testing, so it was important)
HK3.0 = HK3.0
XTOld = original base SW sent--contact was horrible on the left side of the IHS because the 60x60 base was too big and resting on the adjacent bank of caps (even when pushed to the side as much as possible, they forced the base to lift off the IHS).

I've seen this same basic profile on every CPU block I've tested so far. Note that the numbers in the table above are the average of all 5 mounts, where-as the numbers in my charts are the average of the median 3 mounts (I drop the best and worst from my calculations), so they might differ by a few hundredths of a degree from what's on my charts.

[Vapor]

Thank you for the great review Vapor.

Dumb question I was thinking about.

Given your measurements of the minimal flow reduction between the regular XT and the XT 'Alt' (the inlet plate rotated), would you expect to see the same/less/more amount of flow reduction with the inlet plate in its default installation but using a 45 elbow at the inlet instead?

Seems it would be a minuscule difference, but a 45 at the inlet may play better with my planned loop setup versus going straight in with the inlet plate rotated to clear the comp fittings..

Not a dumb question at all actually!

I think the reduced thermal performance of the alternate position is totally due to the fact there's a nominal loss in flowrate. So because of that, I think the 45 degree bend will actually have better performance than rotating the inlet plate. The inlet plate force the water to take two kinda-rough 90-degree bends when in the alternate position (but it's a straight shot with the regular position). That said, the 45 bend fitting is an extra cost and can look a little clunky.

[gabe]

Thanks again for a great work!


However one thing has got me thinking. In your results XT is more restrictive than GTZ despite Swiftech's claims that it's less restrictive with original top config, on a D5...

yes, Vapor validated all our published thermal data, but for an unknown reason we are still at odds with respect to flow. We each re-validated our respective tests, and each re-confirmed our respective data, and we still don't find the same numbers. Could be instrumentation, could be methodology. For now, we just don't understand why; but I'm sure we'll figure it out eventually.

[zalbard]

Very nice review once again!
And good job Swiftech also! The king is back!

[mojododo]

To me, it seems like Heat Killer and Swiftech have similar cooling performance, like the blocks are approaching diminishing performance returns......

So it should now be a preference of looks and brand name?

for me the choice is mounting system. ive got a gtz because the hk isnt as easy to replicate good mounts. ill be getting an xt very shortly, once they hit aussieland.