cathars g5 block probs wid it!!

[PHILIP1193]

well im affriad im not too impressed by what i got tonight with teh block, the base of it looks good but is far from actually flat!!!!!

how do i know? well i took it off my proc and only the edges of my proc were making paw print with the artic silver on eth base of the block, anyways after much sanding all teh way upto 1200 grit and lots of autosol i ahve a flat base and i have a nice full paw print of my cpu on my block.

also did a bit of re-tubing of the system and re-tightened down teh block till the springs on it were practically fully compressed!!!

Results are idle temps down 5°c and load temps down 9°c which is very good now im getting teh blocks full potential

Phil

 

[Phil Stanbridge]

Are you absolutely positive its not just a case of a bad installation mate? You have to be careful not to overtighten the screws; you may have actually put more pressure on one side of the block than the other, hence the pattern you were describing? Either way the new install has sorted out it - the temps are much better now.

 

[PHILIP1193]

well iv tightened it down so much now and temps are a lot better, possibly cuz of the weight of teh block it wasn't tightened enough or as you say not enough on one side although the pattern was around the edges of the ihs, who know just thought id mention it thats all!

But now im really pleased wid temps!!

phil

 

[nick25]

What are your temps/settings out of interest?

 

[Cathar]

Hi Philip,

Phil asked me to register and respond.

First I see that you've managed to fix your issue through better mounting. A 9C difference is only going to be caused by a poor mounting of the block, not through base non-flatness.

The bases are all machine-lapped to around a 10 micron flatness across the entire base. They are all checked for flatness before they are shipped. While there are unavoidably some small variations due to the machining and heat, there should not be anything more than what 1 minute with some 1200-grit paper can't fix, and at this level, we really are talking down to a 5 micron flatness.

Thermal pastes are designed to handle 25 micron non-flatness across surfaces at a minimum, and up to 100 microns for most thermal pastes. In short, at the shipped 5-10 micron flatness level, you simply should not be able to notice much of any difference through further lapping. Even though you may be able to improve the flatness somewhat with careful lapping it should not make any noticable difference.

I take base-plate flatness very seriously and I cannot believe that you were ever shipped a block so bad as to cause even a 0.5C variation due to non-flatness, let alone a 9C variation.

A 9C variation is going to be due to a poor mount of the block against the CPU, and nothing less. Please take the time to read the on-line manual and very, very carefully and slowly apply the block using the crossover method. Hold the block very firmly against the CPU and wind the wingnuts down until they just touch the tops of the springs. From that point on, wind the wingnuts down 1 x 360 degree turn each in a X-pattern until you have wound them down to the required amount. For a bare-die CPU, about 6 full-turns per post. For an IHS CPU, about 15 full turns per post.

 

[Minstadave]

Hows that for good support? Good effort Cathar. The IHS on A64 CPUs isn't always particularly flat, and needs some decent pressure to help good contact.

 

[nick25]

Sorry this is off topic but as IHS has come up - so you know what thermal paste AMD use between the core itself and the IHS. It used to be shin-etsu but apparently not anymore, anyone know?

 

[PHILIP1193]

OMG!!!!!!!!! the man him self i can't believe it!!!!!!

Well i did lap it wid 1200 grit and a bit of auto sol and it seams to be fine now, weather it be my poor mounting which i agree is most likely to be problem or a slight un flatness!!

Now teh block is cooling impeckably and am gettin 34°c idle and 44°c load temps even in this really hot weather and teh 44°c is measured using s&m the only program to heat the proc up properly (more than prime folding etc) does about 40°c during games which im more than pleased about.

Just outa interest how many g5 blocks ahve been made and shipping through out teh world as i knwo they are kinda rare and im more than happy to own one!

Nice to see you taking and active response in your work

Phil

Hi Philip,

Phil asked me to register and respond.

First I see that you've managed to fix your issue through better mounting. A 9C difference is only going to be caused by a poor mounting of the block, not through base non-flatness.

The bases are all machine-lapped to around a 10 micron flatness across the entire base. They are all checked for flatness before they are shipped. While there are unavoidably some small variations due to the machining and heat, there should not be anything more than what 1 minute with some 1200-grit paper can't fix, and at this level, we really are talking down to a 5 micron flatness.

Thermal pastes are designed to handle 25 micron non-flatness across surfaces at a minimum, and up to 100 microns for most thermal pastes. In short, at the shipped 5-10 micron flatness level, you simply should not be able to notice much of any difference through further lapping. Even though you may be able to improve the flatness somewhat with careful lapping it should not make any noticable difference.

I take base-plate flatness very seriously and I cannot believe that you were ever shipped a block so bad as to cause even a 0.5C variation due to non-flatness, let alone a 9C variation.

A 9C variation is going to be due to a poor mount of the block against the CPU, and nothing less. Please take the time to read the on-line manual and very, very carefully and slowly apply the block using the crossover method. Hold the block very firmly against the CPU and wind the wingnuts down until they just touch the tops of the springs. From that point on, wind the wingnuts down 1 x 360 degree turn each in a X-pattern until you have wound them down to the required amount. For a bare-die CPU, about 6 full-turns per post. For an IHS CPU, about 15 full turns per post.

 

[imported_mrbios]

welcome to the forums cathar

 

[Master_G]

As the man Cathar said, those blocks are the cream of the cream, there is no way the finish wasnt anything short of awesome, the very limited runs of the blocks ensure that. The amount of work that he puts into the blocks to optimise them to within an inch of their lives, he is not just going to throw away even .1C with a non flat finish.

The IHS on the A64 does suck, my one is so non flat that if i mount the stock aircooler with a thin layer of TIM, it only gets contact around the edges, not the centre.

Got any pics of the gorgeous block in action Philip?

G

 

[Cathar]

Am not disclosing the number of G5 blocks made, but it certainly was far more than I ever though I'd be making.

I feel that I must owe some measure of apology. Have been discussing with someone at OC forums about their G5 block that they received, and they were able to prove to me that when the block was assembled that it was bowed in a concave fashion slightly along the length of the block. When the block was disassembled, the bowing went away.

At my end I tested the blocks and found the ones that were bowed upon assembly and touched them up by hand when assembled. There weren't that many like this and I thought that I caught them all. The other thing is that this may be caused by people opening their blocks up to take a look, and overtightening the screws that hold the block together when reassembling. I do warn about this in the manual, and upon further testing I do note that if I apply excessive pressure on the bolts that clamp the block together that this can induce a small amount of concavity in the base along the length of the block.

This is not excessive, but it can happen. It should not however create a large difference as the degree of bow is still well within what thermal pastes are designed to handle, but can create non-flatness to about 3x larger than in an unassembled state in the worst case.

If this is an issue for you, then please accept my apologies, and a quick lap with some 1200-grit paper with the assembled block will fix the issue, although it should have a minimal, if noticable at all, impact on temperatures.

 

[PHILIP1193]

block before she went into comp

before tubed up

all tubed up althopugh i have changed tubing since that pic , will take oen later

well cathars i think your block is very good dispite the probs i ahd, and who ever's fault it was its working fine now mate.

I dont expect an exact number of blocks but are we talking 50+ or less as i have been boastin i own one and pimping my self saying theya re rare lol

phil

 

[unreal]

can you repost pictures please?

edit: bugger. it reloaded when I posted....

Looks good

 

[Master_G]

Thanks for the explaination there Cathar, i hate to think how many crazy people like phil will shell out for G5.

Their girlfriends must be asking them, why couldnt you buy silver in the form of jewellery for me?

G

 

[PHILIP1193]

hence why i dont ahve a g/f lol

phil