Printable View
Picked up 3 blocks from the machinist today, the storm, apogee and fuzion. Now I just need to test to see if it works... I think I may have gone a mm or two too big on the step (diameter wise, not depth). I also had him lap them so they are FLAT, no BOW WHATSOEVER. This required the fuzion to be lapped ASSEMBLED. Will update when I have some data, but now a couple of pics.
http://www.anonforums.com/builds/tes...n/3stepped.jpg
http://www.anonforums.com/builds/tes...stormclose.jpg
I'll admit, I'm sort of lost in seeing the benefit here. I'll wait for your results.
Same, unless you also had the "core" area shaved down as well.Quote:
Originally Posted by EnJoY
Time will tell. :D
i can see benefits of doing it if diameter (outer) was less than size of ihs ...
Think of the dynamics of an IHS for a second. Have you ever seen the underside of it? Well it has a ridge around it for support round the outer edge. If you can apply pressure INSIDE of that ridge you can get more pressure on the die(s). An unmodded or unbowed block can be bolted down with 200 lb. of pressure yet the outer support of the IHS is taking the brunt of this pressure off of the center. This is why a bowed block works well. This is the same concept but IMO done correctly which is FLAT.Quote:
Originally Posted by EnJoY
We shall see.
Cant wait to see the results... But I'm also a little confused...
Perhaps, The step is there so the block will bow ITSELF inwards when mounted? I could be very wrong, But we shall see. If the middle was already lapped to a bow, I could see that idea working very well, Seeing as it would offset the inwards bend from the mount.
Crap IDK if I made any sense, I've been awake for 30 hours now... And counting.
EDIT: My god, He's on to something!
I have a feeling I may need to bring my Fuzion into Techshop soon...
nikhsub1, is ur cpu lapped?
The center step is SIGNIFICANTY smaller than the IHS, perhaps another picture will help.Quote:
Originally Posted by rafrag
http://www.anonforums.com/builds/tes...mountpaste.jpg
that answers all questions (mine anyways) wonder how storm will preform.
darn 1 more arent u concern with putting 2 much pressure on silicon?
u know
"the outer support of the IHS is taking the brunt of this pressure off of the center. This is why a bowed block works well. This is the same concept but IMO done correctly which is FLAT."
thank you nikhsub1.. this is the first time I actually seen an explanation about bowed blocks that actually makes perfect sense in my noob brain :up:Quote:
Originally Posted by nikhsub1
that's some damn nice looking work on those blocks...can't wait for results :D
As long as you put the pressure on EVENLY, It's pretty strong. Think of it as a piece of glass, If you get the pressure even on all corners, you can apply hundreds of pounds. But if the weight favors one side, Then you can chip a corner off. TBH, Idk how much you have to chip off to kill it. I've chipped some AXP's, NF4 chipsets, And a few GPUs... They still work. :woot:Quote:
Originally Posted by rafrag
wonder how easy this will be to get done in Tampa, FL if it works
easier then brazil? :stick: :p:Quote:
Originally Posted by nealh
Looking forward to the results, I like the idea of concentrating a flat not bowed surface on the IHS center. Without the outer edges of the IHS to support the weight it should give equal pressure across the cores that I don't feel that we get with bowed blocks.
<gets out popcorn and waits> :up:
Me tooQuote:
Originally Posted by SiGfever
maybe with some copper 90 degree elbows and proper mounting, they could be our new high performance gpu blocks =D
nikhsub1;
Am i seeing that correctly? do you not only want to cool the top of the ihs but the vertical wall of it too? If so arn't the block's milled the wrong way?
Hmmm, would be interested seeing the results...
Same trick I talked in an article to avoid bow ;)
http://www.cooling-masters.com/image...es/contact.png
I too would like to see the results. If they are good I know of a few cnc shops nearby. May have to see what it would cost to have done.
Nice one Nick, really good work done on them blocks. I look forward to the tests
Long time no see rosco! Yeah I guess so, lol. What size do you think would be optimum for the center 'patch'. I think I've gone a bit too big... mine is 24mm sq. and .75mm high. I'm thinking 22mm sq may be better? I wish I knew the dimension of a quad core without the IHS.Quote:
Originally Posted by rosco
You, my friend, are a genius. I was skeptical, but seeing this does prove that you did have a doozy of an idea.
Excellent work, and also applicable on any platform...
I'm pretty sure this will bring the Storm back...
Nice! Eagerly awaiting your testing results:up:
No. Not my test bench CPU.Quote:
Originally Posted by rafrag
when will you have some before/after mod data?
I'll probably make it a bit smaller than 24˛. Too big and WB touchs IHS edges so it's not good, but too small and you loose too much contact area even if pressure contact is increased (to get a thinner TIM2 essentially). Balance is not easy, but final difference won't be enormous too (heavy load needed). Don't put too much thermal paste and that's all, problem is the necessary pressure to minimize its thickness, what bow achieves easily because contact area is very reduced when analyzing contact paths. A flat-flat situation remains the easier solution for everyone in comparison of milling again the baseplate.
0.75 mm is not really necessary because max IHS flatness defect is ~0.10 mm based on my measurements on several CPU (Intel says 0.05 mm max, but never achieved :D). That's why I put 0.1-0.3 mm maxi to make the base step, that's enough to avoid the IHS ridge along the perimeter. If you have a very thin base like MP-05 for instance (1 mm), you can't make a 0.75 mm step because you'll fragilise the whole WB and it will be bend easily under pressure. The Storm stiffness isn't really impacted by the 0.75 mm because base is thick.
Central IHS area is generally quite flat (a bit convex on dual dies), IHS edges are (always) the major problem. Let's say that a ~4 mm square ring is bad on IHS, that gives us a ~22˛ area. You can see the "copper ring" when you lap the IHS, it's the bad part to eliminate in priority, all the rest is pretty good and flat if you continue to lap it because copper appears everywhere in a same time very quickly. The Swiftech bow only gives a ~0.1-0.15 mm convex height difference at the contact and it's sufficient to make a difference, even with a non lapped IHS.
Well we have the before... I do have a busy week this week, but monday I might be able to get 3 runs of the storm done... If it doesnt do as I hoped, I'm gonna make the center patch smaller and retest. There must be an optimal size... obviously the smaller we go the more PSI of pressure we create over the die(s) but go too small and you may end up with parts of a quad core that are literally uncooled.Quote:
Originally Posted by Bail_w
Does anyone know the measurements of a quad without the IHS? I'm talking the area that the 2 dies occupy.
Yeah, I'm thinking 22mm to 20mm sq would be where its at. IIRC Cathar thought 22mm (and he is usually right LOL). rosco are you doing some experimenting with this as well? This whole thing begs the question as to how or why the 'Mfgrs' have not tinkered with this idea...??Quote:
Originally Posted by rosco
Edit: nvm.
Here it is (from one topic on this forum). It's for Kentsfield IIRC but it's not important, it's almost the same each time when dealing with 2 dies. A ~23x14 mm area here just above dies (without taking account of the heat spreading).Quote:
Originally Posted by nikhsub1
No time to make experiments with that trick for the moment, but obviously it's better than nothing if you don't want to lap your IHS and keep your warranty :)
all i found is:
Die Size (combined) 2 x 143 mm2 or 286 mm2 total die area
it sems 24mm squared is slightly big(23 looks correct??)....22mm squared would be too at edge of too smallQuote:
Originally Posted by rosco
axiously waiting results. I wish I was able to help some how. This stuff is exciting to me.
--pak
For sure, you won't be able to measure any difference between a 22 and a 23 mm² contact :)
looks like the bow without the bow... :up:
No doubt. Damn. I need to go back to the machinist now and do 22mm sq. :(Quote:
Originally Posted by rosco
I can get you the quad's dimensions, Just need a picture of a naked quad and a spare IHS (Got the IHS and found a picture on google) and I can do some dimension scaling.Quote:
Originally Posted by nikhsub1
I'll get back to you later tonight. ;)
See post # 33 by rosco... 23mm x 14mm for a quad's AREA. 22mm sq for the patch should be good.Quote:
Originally Posted by kevinbo03
then why do it..sorry for my ignoranceQuote:
Originally Posted by nikhsub1
is it becasue the smaller area will fit on the IHS better and provide better transfer
Meh, now I'm wondering if a 14sq mm patch would just not OWN on a single die CPU... that would be massive PSI. Block Mfgrs would not like to make 2 bases though...
Neal, the SMALLER the patch, the more PSI we can apply to the core(s). The more pressure the better. Make sense?Quote:
Originally Posted by nealh
I can see the point of all this easily enough but am a tad concerned about
*introducing 4 corners as pressure points to the IHS surface.
*whether pressure is evenly distributed or becomes less towards the center.
*a pocket of TIM becomes trapped (if point *2 is true). This seemed to be the benefit of bowing to me.
Are we merely reversing the IHS square outer frame and putting it on our block instead? Although, admittedly much closer to the dies themselves.
Can't wait for some results.
:clap: niksub1 and rosco for waking us all up again.:up:
LOL, metric system is not one of ur strong points is it?Quote:
Originally Posted by nikhsub1
OK..but you want to make sure you cover the cores completely, right to cool the entire core evenly, no?Quote:
Originally Posted by nikhsub1
will 23 vs 22 make any difference in reality?
Good thinking and work!
I was always curious as to why the stock Intel air cooler for the QX series CPU has a circular base, yet the one size fits all waterblocks had square ones.
...and as to why some air coolers (CNPS7500) which has a circular base, works better than the stock square base coolers on earlier CPUs.
well done :)
Matched page 32 of this specsheet with some naked quad photos, Got 22.8 x 13.3; I hadn't seen rosco's post until now :p: :
http://download.intel.com/design/pro...s/31559205.pdf
I noticed in there, That the IHS is 2.03mm Thick, to a tolerance of .05.
So, My thinking is, Would your block design work better if the IHS was lapped down by 1mm? 1.5mm?
Wouldn't lapping the CPU kind of defeat the purpose?
--pak
yup
this is just what i thought you were going to do :D
but i don't think you will see a noticeable difference to bowed blocks :(
Not entirely, the purpose of His design (So I think! Correct me if I'm wrong) is to focus all of the pressure directly over the Die/Dies. Bowing achieved this, Lapped or not lapped, And his design eliminates the more solid and raised edge of the IHS from protruding in the path of the block.Quote:
Originally Posted by pak
What I meant, is by lapping a significant amount off of the IHS, There would be less metal between the block and the cores, And therefore, Less pressure would be dissipated laterally throughout the IHS.
A good example would be playing Jenga with fish sticks Versus wood blocks.
Quote:
Originally Posted by kevinbo03
Go ahead and call me a dumb ass, but I am missing your point.
As far as taking a mm off the IHS, wouldn't that already improve temps for two reasons?
1. if your going to take 1 mm off, it might as well be nice and shiny.
2. Less metal=better heat transfer?
3. Your point which I am missing, probably cuz you mentioned fish sticks and I have no tartar sauce.
--pak
Please correct me if I am wrong but would not the bow cause the heaviest pressure at the apex of the bow and the areas radiating out from the bow have less contact pressure?Quote:
Originally Posted by dinos22
Which is why this should yield better results, the question is how much?Quote:
Originally Posted by SiGfever
--pak
I would be very surprised if this affects temps in a measurable way....
The only thing I can see coming from this is better consistency between mounts compared to bowing...
Hate to be a damp dishrag on everyones hopes but, I have to agree.Quote:
Originally Posted by Linus@ncix
I'm also not very enthusiastic, However it is the first time a waterblock with a surface the size of the DIE is being mounted, So all bets are off.
Holding an IHS in my hand right now is making me think too much. :shakes:
Me too! Thank god, someone finally posted some pictures so I can understand this. I have sort of scratching my head all along trying to figure out this bowing business. :up:Quote:
Originally Posted by hoax
I always thought the bow was to compensate for the concave bending of a mounted block....just like how long flatbed semi's have a bowed stucture that flattens under the weight.
Doesn't the concave bending still take place with this design? Granted, it won't negatively effect the mount as much since the edges are still within IHS limits, but it'll still be there.
I'd imagine a very shallow bow + small contact patch would be the best combo....
Took the words right out of my mouth :)Quote:
Originally Posted by Vapor
Looking forward to seeing some results Nik (can't wait actually), are you going to do any testing with the 24mm^2 base or are you sending them straight back to the machinist?
Here's what I'm thinking in pictures....
Here's what we've had for years, not so great...
http://img101.imageshack.us/img101/9679/normaldf7.jpg
Then the bow came, and it was pretty darn good (almost too good in this pic, oh well, these were done quickly):
http://img216.imageshack.us/img216/1870/bowedvd4.jpg
Now nikhsub's great idea to increase pressure by reducing contact area, but I imagine it will still have a concave tendency (as depicted):
http://img216.imageshack.us/img216/3346/nikhno1.jpg
Now just a slight bow with Nikhsub's reduced contact patch would be 100% ideal, IMO (but damn hard to machine I'd bet :():
http://img101.imageshack.us/img101/4...wednikhvm6.jpg
This assuming the bow is created because mounting pressure is along the top edges, and will happen to the entire base uniformly (or nearly). Nikh's idea to put the pressure back in the middle is awesome, but until we can mount with pressure down the middle, there will always be a bow (theory).
Actually the blocks are all FLAT, not concave or convex at all. I would imagine that this sort of base would work best with a lapped CPU.
My pictures are what happens when mounting pressure (on the edges) is applied....there should still be a concave nature to those three blocks (since mounting pressure is along the edges, and not in the center). Granted, the increased local pressure will negate a lot of it in the end (since it's a huge theoretical advantage on its own right), but it'd be interesting to see if you could have a slight bow added and compare.
My thought is that a bow with my mod or a bow with an unmodded base will yield the exact same thing... I don't see how it would be different?
A bow with an unmodded base still has a large (full-size) contact patch, no?
A bow on your base would not have to be as drastic for 'perfect' mounted-flatness.
EDIT: even doing the larger O-ring style bow would be less drastic considering the thinned part of the base would take more of the torque.
I follow you Vapor--but I do not think the blocks go convex when tightening...the edge force is too low to affect the massive crossection. Theory. The mount bends--not the block.
@Nikhsub1 Gutfeeling: the slight oversize you have now will be just as effective as a slight undersize.
Bow base might be affecting for dual core, but not so good for quad.
Four passes with a file on your fuzion and you could find out.
If they'd sell the bottom piece alone I'd try it.
I tend to Vapour's line of thought, as my earlier thoughts indicated.
We're forgetting the corner points. They may need radiusing.
reduced block/die area, and bowing some...
Quote:
Originally Posted by Jedda
Good thinking there... plus chamfering the corners slightly as well, so the corners don't dig into (bite) the IHS.
Maybe so..but most heer are fighting for 1-2C improvementQuote:
Originally Posted by Linus@ncix
Making a more consistent mount can seem to "yield huge temps improvements"... improper/porr mounts are a pain...if this seemly fixed it is a huge benefit IMHO
I think there will some benefit and the idea of a small bow with it seems real good...but how to create the "exact bowing amount" seems the biggest issue
Interesting ideas. I'm kinda thinking of radiusing the edges of my lapped IHS now so that the block mounting forces are no longer transmitted into the vertical surfaces of the IHS resulting in more of this pressure going into the other loading area...the CPU cores. I'm using Coolaboratory LM which has a rather low C/W in comparison to typical TIM, so TIM layer thickness on my setup isnt too much of a concern, but very interesting possibilities nonetheless if better thermal transmission can be accomplished with more pressure over the CPU cores with the goal of an ideal 0 C/W for the interface.
That said, I'm wondering about the posts I've read a couple months ago about Intel switching from a solder based TIM below the IHS to a paste based TIM. Anyone been popping IHS' off lately to see what the latest CPU manufacturing spec is shipping? Might be another can of worms to work around if the paste interface is being used below the IHS.
Now that it's been explained to me (dee dee dee :p: ) I'm eager to see the improvements.
Is there an ETA on results Nick? I can see this possibly affecting evaps as well
As a representative of a retailer, and not just an anonymous member of a forum, I don't want to get too far into this debate, but IMO there are far better ways to improve mounting consistency than filing down the bases on all of your blocks.
2/3 times with a bowed APOGEE GT/GTX and 3/3 times with a FuZion I can get the exact same temps with my plastic backplate, bolts, and springs. That means at most I'd have to do a couple mounts to get the best temps. To me that's worth not voiding the warranty on my blocks, and not replacing them every time a different sized die comes out.
Quote:
Originally Posted by nealh
what about just putting a support in the center of the block(gt/gtx) to keep it flat under mounting pressure?
All very well said and done, but this is Xtreme Systems where pushing the edge of current tec is what we do.Quote:
Originally Posted by Linus@ncix
Taking the best and tweaking it to find a way to make it better - nowt wrong with the mate. Thats what devlopment is and R&D and thats what Nick does best
Quote:
Originally Posted by Pete
And lets not forget that this kind of work isnt something that the average person will go out and do. I can comfortably say that even if this gave a 10c improvement in temps that I would not go out and have my blocks milled.
Not saying its a bad idea, far from it, its through innovations and experiments like this that watercooling is pushed to its limits. Cathar did the same with the Whitewater as well as the Storm, and its very possible nik is doing the same with this mod. Its these innovations that get companys that have the means to incorperate these ideas into their existing or or future products so that the average person who is lacking the resources or funding justification to do something such as have their blocks milled.
Im looking forward to seeing the results on this one.
Nick 1st post about it all he was saying he'd been talking with Cather no end about it and stuff.Quote:
Originally Posted by Jimmer411
Take the man that crated the WC sceen with the right kit and very talented guy and you get this.
I'd defo give it ago to try it for my self, old kit, if it cane be made to more advanced then nowt wrong with it
http://www.stealth-and-mutilation.co.../core_spec.jpgQuote:
Originally Posted by rosco
I just found this thread today ... good thing I read through all the posts first ... I was about to link nikhsub1 to that image on my server, but I see that you beat me to it.
It's about as exact as I could get it. My measurements were taken from a QX6700 diagram. As you said though, until the next die-size emerges these should be fairly accurate for the Quad-Core procs. :up:
Quote:
Originally Posted by Jimmer411
Bingo
Although given a fuzion is $60-65..I would get it milled if I could find someone to do it and right without charging me a arm and a leg
Wait, what is the possibility that the sharp corners could actually depress the material of the IHS after awhile? If that was so then you couldn't use anything else to cool whatever cpu is used under the block.
lol... i was going to comment about that.Quote:
Originally Posted by ColonelCain
Your basically SOL on the next line of blocks that came out. But i figure if the block technology doesnt change any. Then there at its pinicle.
However, if its goign to be a one bang thing on the CPU, and not just the block, might as well go all out and silver it.
Should not be a problem...Quote:
Originally Posted by ColonelCain
Anyone tried a round 26.9mm diameter step?
It'd get rid of the corners.
What's the width of the IHS outer rim?
nothing like looking outside the square, eh?
Does anyone have an idea what the base of the Fuzion weighs? :confused:
I don't know much about removing the IHS, but it seems like that would be the easier/cheaper solution. Although, I guess the size of the IHS doesn't change too much throughout the generations. You are talking about tailoring water cooling components to specific systems. If you really need those extra degrees, then maybe TEC, refrigeration would be easier.
At the same time, I do think its cool to try and get the most out of plain ol' water.
Your theory does make sense though, and I am curious too see the results.
How's this for a wild idea.
With the IHS still attached to the cpu.
Have your machinist bottom tap the corners of the IHS.
Then have him machine a groove near the perimiter for an o-ring and
have him carve out a pin configuration on the IHS.
Then all you have to do is get a matching top.
That way the IHS is the water block. Cut out the middle man.
Sorry, I know it's a stupid idea. But its something that always comes up when I'm thinking about how to get rid of the thermal losses assosiated with the thermal interface and irregular mounting pressure.
IHS's are no longer 'easily' removable. They are now welded to the cores and require not-so-safe quick heating of the chip to melt the solder and attempt removal. It also requires a big set of balls, and someone who isn't scared by the possibility of killing their chip. That being said, it's been done before by many people, but I wouldn't even come close to calling it a mainstream option.
2 mins it took to my X6800. No problems but when it pops it pops and you need a way to stop it going flyingQuote:
Originally Posted by EnJoY
nice :up:
OK! The idea totally works! Now, does it work better than a bow? That we will see soon! The Stepped storm performed about 2C better than the best normal storm and almost 4C better than the worst mount... I've only done one mount thus far, now I want to get a apogee GT on there and see how she goes vs. a bowed apogee gt.
^^^how are you measuring temps^^^
and how are you controlling testing conditions >> temps/humidity etc
nice! at the very least there still improvement...
Sigh, I guess no one sees any threads I make.Quote:
Originally Posted by dinos22
http://www.xtremesystems.org/forums/...d.php?t=151781
Go speed racer, Go!
:woot: