Apogee GTZ

[skinnee]

Hardly. TAT (Intel's Thermal Analysis Tool) will get a dual core hotter than anything else can... but the bummer is it only works on dual cores. Gets The cores some 15C hotter than FAH can.

Maybe but it doesnt on my x3110 or Q66.
Plus they pass all test but F@H after 2 days.
With givin OC.

Don't skim a post from nikhsub...read the entire thing, read it twice. This is one guy among a few here in the LC section that when they speak, you listen.

[hotdun]

"topographically mapped CPU contact area"
What does this thing come bowed?

Any current blocks with comparable pin size, i.e. 0.009" ?

[sirheck]

Don't skim a post from nikhsub...read the entire thing, read it twice. This is one guy among a few here in the LC section that when they speak, you listen.

Im not knockin him.

But it still doesnt change the fact that the SMP client is the
hardest stress test on MY CPU.

You do know a x3110 is a dualcore and a Q66 is a quad core dont you?

[skinnee]

You do know a x3110 is a dualcore

Doh! My bad, I thought only the Xeon eXXXX's were duals.

[silverphoenix]

"topographically mapped CPU contact area"
What does this thing come bowed?

Any current blocks with comparable pin size, i.e. 0.009" ?

I've got my fingers crossed that it's that special IBM micro cross matrixing that increases TIM spread and increases thermal efficiency. Of course I can wish and hope lol.

[sirheck]

Doh! My bad, I thought only the Xeon eXXXX's were duals.

Well the x3110 is basically an E8400.
There is also an x3350 (something er other) which is the
equivelent to the Q66

[Martinm210]

"topographically mapped CPU contact area"
What does this thing come bowed?

Any current blocks with comparable pin size, i.e. 0.009" ?

The thermalright XWB-01 probably has the finest pins of the blocks I have on hand, it's probably in the neighborhood of that dimension or under .010" anyhow. But the TR is made from what appears to me, some sort of copper sheet method, where they stamp out the ultra fine pins and stack alternating sheets of pins, no pins, pins, etc, and then it's soldered to the base.

Trying to machine or cast something that fine would be very difficult.

Where did you find that introduction anyhow? I can't find it on their site..

[hotdun]

I've got my fingers crossed that it's that special IBM micro cross matrixing that increases TIM spread and increases thermal efficiency. Of course I can wish and hope lol.

Whoa, something like this?

============
"The clever bit is the layout of the channels. According to Big Blue boffins, the system uses a hierarchical network of pathways not unlike tree branches or human blood vessels. Bigger canals feed into smaller tributaries, down into a narrower rivers and streams.




IBM claimed the channels allow the thermal paste to spread more completely and more thinly (see above), improving its ability to conduct heat away from the chip and through to a heatsink, fan, cooling pipe or liquid-filled radiator. The paste is essential to allow the chip to maintain contact with its cooler even though the gap between the two devices changes as they both undergo thermal expansion and contraction."
============

Looking around, I found this too: LINK
I think I just had an accident...

[hotdun]

The thermalright XWB-01 probably has the finest pins of the blocks I have on hand...

Ya, I was reading your review on your website and saw you mention that....interesting.

I found the GTZ description on this random website here: LINK

[twwen2]

Woah i remember reading about that IBM thing but had forgotten about it. I wonder...

[rosco]

I've got my fingers crossed that it's that special IBM micro cross matrixing that increases TIM spread and increases thermal efficiency. Of course I can wish and hope lol.

Not at all, it's just a bowed base, a "convexo-concave" one, in order to match as good as possible the deformed shape induced by the hold-down socket plate and an original IHS.

[jonny_ftm]

Ahh, ok. I tried the 8K in place vs the OCCT 2.0.0a back and forth several times, and they are close. I do seem to have just a touch more heat coming from the OCCT runs and it's less button clicking, so I think I'm going to go that route.

OCCT 2.x in CPU mode uses same algorithm as prime95, but it is much more optimized and better multithreading. The FFT are iterations from 8-64k in OCCT while Prime95 stops at 32k (could be a plus or less as 64k will be out of the CPU cache)
OCCT v2.x in CPU mode in my comparisions testing will give the same temp as Prime95 small FFT but it will be reached much quicker.

Using OCCT in custom mode is a bad idea. It will make iterations between the CPU mode and the RAM mode at regular intervals. You can see it with an oscilloscopic temp curve on the cores (OCCT 2.x I talk about)

The hardest stress test for a quad or any cpu i know of is F@H.

Though, not sure if you would have the time or desire to run it.?

F@H will make a stress on the whole system, including the RAM and NB. Hangs ups won't be CPU only related. And F@H can't load a CPU hotter than Prime95. Also, far better than F@H at loading the whole CPU-NB-RAM system is OCCT v2.x in RAM mode.

Linpack could be a great tool, but only in its 64bit edition. The 32bits edition, in my testings, doesn't stand up to OCCT or Prime95, even with the excessive heat generated.

Finally, I'm not sure TAT really gives so much heat. I thought it was some "software simulation" of a 100% TDP state of the CPU, where all other programs, including Prime95 won't exceed maybe the 80-85% TDP of the CPU

All this was largely discussed in the "Intel" subsection of the forum. Maybe it's better to continue debating there if you like. Just make a search before. there are great topics discussing all this and whereI also posted comparision curves

[C'DaleRider]

Well the x3110 is basically an E8400.
There is also an x3350 (something er other) which is the
equivelent to the Q66

Well, you're almost right.......

The Xeon socket 775 cpus are arranged as this:

The X30xx's are Conroe cpus

The X31xx's are Wolfdale cpus

The X32xx's are Kentsfield cpus (like with the Q6600....the Xeon equivalent is the X3220)

The X33xx's are Yorkfield cpus

[menace2society]

Is this going to be a Intel only block Not one mention of anything AMD or how anything preforms on anything AMD is there absolutely no interest in AMD at all?

How the might have fallen.

[jonny_ftm]

All true, but you want a fixed length FFT/WU for any type of CPU block testing. In Prime95, you set that via the custom menu. For Prime it's between 8-64k. From what I've seen, "CPU" mode in OCCT uses a fixed length also I would expect. I say this because the heat load and power draw doesn't fluctuate in OCCT's CPU mode.

So bottom line, as long as the loading software allows one to control the heat dump in a "fixed manor" it's good to use.

From what Martin said, he observed a steady load with OCCT also.

andyc

No, it will iterate between 8-64k, I confirmed this with the author, no way to fix it. I participated in the beta debugging on HFR forums...
But as you can see, using small iterations won't affect temps of the CPU at all since the load on processing units will be the same with OCCT. It is optimized for that. I can confirm the CPU temps curves will be linear at a maximum heat

[sirheck]

OCCT 2.x in CPU mode uses same algorithm as prime95, but it is much more optimized and better multithreading. The FFT are iterations from 8-64k in OCCT while Prime95 stops at 32k (could be a plus or less as 64k will be out of the CPU cache)
OCCT v2.x in CPU mode in my comparisions testing will give the same temp as Prime95 small FFT but it will be reached much quicker.

Using OCCT in custom mode is a bad idea. It will make iterations between the CPU mode and the RAM mode at regular intervals. You can see it with an oscilloscopic temp curve on the cores (OCCT 2.x I talk about)




F@H will make a stress on the whole system, including the RAM and NB. Hangs ups won't be CPU only related. And F@H can't load a CPU hotter than Prime95. Also, far better than F@H at loading the whole CPU-NB-RAM system is OCCT v2.x in RAM mode.

Linpack could be a great tool, but only in its 64bit edition. The 32bits edition, in my testings, doesn't stand up to OCCT or Prime95, even with the excessive heat generated.

Finally, I'm not sure TAT really gives so much heat. I thought it was some "software simulation" of a 100% TDP state of the CPU, where all other programs, including Prime95 won't exceed maybe the 80-85% TDP of the CPU

All this was largely discussed in the "Intel" subsection of the forum. Maybe it's better to continue debating there if you like. Just make a search before. there are great topics discussing all this and whereI also posted comparision curves

Yeah im not gonna debate anymore, but SMP stresses my cpu only.
Almost 0 ram usage according to windows taskmanager.

[Kougar]

I will say only this: the Apogee GTZ only have two things in common with the GTX, it's name and it's function. Expect totally new stuff under the hood.

I cannot disclose anything else at this time, other than the fact that the product is really coming very soon. This is no vaporware. We have never done "coming soon" before, but I see that all my competition does, so what the heck..

Howdy Gabe! That looks distinctly LGA775 only... what about Nehalem/Bloomfield? Launching a 775-only block just a month or so before LGA1366 is officially lanched seems a bit odd?

You will, you will, it's a question of days now, if not hours.

make sure you bow the GT.. we only ship it bowed now. makes a big diff in perfs. can provide o-ring if needed.

Wow, this really carries some weight if the manufacturer recommends & now does this themselves. I guess I'll be inspecting my Apogee GT & CPU heatspreader closely when I do some loop work in a couple days to WC my first GPU. I have some Apogee homework to catch up on...

[jonny_ftm]

OCCT?

andyc

Yes, I can confirm you this for sure, from the author. OCCT is a stability test, not meant to be a heat generator only. If it runs only one iteration, it won't catch system errors effieciently

Back to GTZ:

- the day is getting to its end, still no GTZ
- At Gabe: Will it be even possible to upgrade it to Nehalem? The socket holes spacing are much distant than LGA 775. So it will need not only a backplate, but also a new top

[71 (Bryan)]

Will it be even possible to upgrade it to Nehalem? The socket holes spacing are much distant than LGA 775. So it will need not only a backplate, but also a new top

im pretty sure most block makers will sell a separate hold down plate for nehalem when its launched

[AllAgainstPaul]

Must See Internal Pics

[jonny_ftm]

Ok, it won't be 30/07

Good night

[71 (Bryan)]

take a chill pill bro, maybe he'll post it at 23:59 =P

[Martinm210]

Any chance you can test the Enzotech SWC-Rev. A?
http://www.enzotechnology.com/scw_reva.htm

FYI, I was fortunate enough to get one of these on the way also..

[SiGfever]

I always thought that the GTX with the copper top was one hell of a great looking block.

[Vapor]

So a stepped base and some really intricate/fine mesh-like copper stuff on the inside (or something along those lines).

Looking forward to it

Maybe it'll be good enough to get me back into WC, not that I ever use my desktops any more :-/