Xigmatek Battle-Axe: First Direct-Touch Heatpipe VGA Cooler

Want to talk about one of the articles in SPCR? Here's the forum for you.
Post Reply
MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Xigmatek Battle-Axe: First Direct-Touch Heatpipe VGA Cooler

Post by MikeC » Fri May 02, 2008 11:50 am


widowmaker
Posts: 239
Joined: Sat Mar 29, 2008 7:05 pm
Location: Toronto Ontario

Post by widowmaker » Fri May 02, 2008 12:06 pm

On the third page first image, I assume it's supposed to be "Fin shape" as opposed to "F in shape" which can be taken to mean something else. :wink:

FartingBob
Patron of SPCR
Posts: 744
Joined: Tue Mar 04, 2008 4:05 am
Location: London
Contact:

Post by FartingBob » Fri May 02, 2008 12:10 pm

Image
As soon as i saw that photo i knew the heatpipes were way to far apart to be really effective.
When most of the high end cards (read: 95% of cards this will likely be installed on) arent able to use half of the heatpipes then its clear they didnt terst this enough in real world conditions.
Still, performed quite well and its nice to have soft mounting fan plugs as standard, which is my one complaint about the S1.

jolynsbass
Posts: 46
Joined: Tue Apr 03, 2007 5:33 am

Post by jolynsbass » Fri May 02, 2008 1:19 pm

Maybe that crazy closed fin design is actually somewhat helpful to the performance of the heatsink, since it redirects all of the air to the hottest part of the fins - the part immediately around the heatpipes. Impressive numbers, considering that only half of the heatpipes were really doing anything. Too bad about the frail hardware, though. If they beef those screws up a bit, they'll have a real contender.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Fri May 02, 2008 1:37 pm

Aside from better screws (easily done even by the DIY), how about a copper piece soldered to physically connect all the heatpipes -- just outside the GPU/base contact area? A super thin copper heatspreader might actually help in this HS.

walle
Posts: 605
Joined: Tue Dec 26, 2006 10:52 am

Post by walle » Fri May 02, 2008 1:46 pm

This cooler was mounted on a X1950XTX card which has a *smaller* GPU footprint, mounting it on a GT/GTS or GTX card with its bigger footprint suggests that not only those two heatpipes in the centre would make contact, but all four. Perhaps the outer heatpipes would not make as good contact? be that as it may; they would surely still make contact. I think that the *smaller* GPU footprint on the ATI card should be taken into account here prior to the Xigmatek design being criticised as such.

Reservation:
Unless I'm mistaken about the smaller GPU footprint of the ATI card, in which case you are (should) free to fully dismiss the input.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Fri May 02, 2008 1:57 pm

walle wrote:Reservation:
Unless I'm mistaken about the smaller GPU footprint of the ATI card, in which case you are (should) free to fully dismiss the input.
Look at the die of Asus EN9600GT on this page -- it's not appreciably bigger than the ATI 1950's core.

walle
Posts: 605
Joined: Tue Dec 26, 2006 10:52 am

Post by walle » Fri May 02, 2008 2:10 pm

I stand corrected on the GT and GTS cards Mike (unnecessary mistake that was) which leaves the GTX card, one out of three is not that bad eh. 8)

Cheers

Note to self:
Should have known better when owning a GTS card, absent-minded...erhm.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Fri May 02, 2008 4:03 pm

MikeC wrote:
walle wrote:Reservation:
Unless I'm mistaken about the smaller GPU footprint of the ATI card, in which case you are (should) free to fully dismiss the input.
Look at the die of Asus EN9600GT on this page -- it's not appreciably bigger than the ATI 1950's core.
It's hard to google up exact numbers but here is what I'm seeing:

G94 9600GT ~200 mm2
G80 8800 480 mm2
G92 8800GT 290 mm2

X1800 XT R520 264 mm2
X1950 Pro RV570 ~230 mm2
X1900 XT R580 315 mm2
X1950XTX R580+ 315 mm2

It's very very confusing looking at model names and trying to line them up vs the chip. There are way more than one name per chip, I just picked one and listed a size.

My point in all this is that if your x1950 whatever is in the 230mm2 to 315mm2 range it is no where near the largest possible chip size. 480mm2 dwarfs that easily. Add to that the heat spreader seen in pics like this and I have to wonder if all 4 heat pipes wouldn't contact such a wider package?

It's not just how does it do on any one card but how well would it do for really large GPUs past, present, or future?

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Fri May 02, 2008 4:20 pm

dhanson865 wrote:It's hard to google up exact numbers but here is what I'm seeing:

G94 9600GT ~200 mm2
G80 8800 480 mm2
G92 8800GT 290 mm2

X1800 XT R520 264 mm2
X1950 Pro RV570 ~230 mm2
X1900 XT R580 315 mm2
X1950XTX R580+ 315 mm2

It's very very confusing looking at model names and trying to line them up vs the chip. There are way more than one name per chip, I just picked one and listed a size.

My point in all this is that if your x1950 whatever is in the 230mm2 to 315mm2 range it is no where near the largest possible chip size. 480mm2 dwarfs that easily. Add to that the heat spreader seen in pics like this and I have to wonder if all 4 heat pipes wouldn't contact such a wider package?

It's not just how does it do on any one card but how well would it do for really large GPUs past, present, or future?
Our primary VGA test card is an X1950XTX -- and I just measured its die -- it's around 190mm sq. So... examining the base of the Xigmatek, if the die is about 215mm sq or smaller (and square, not at a diagonal angle), it will only touch the 2 central heatpipes in the base. The core would have to measure around 250mm sq for it to straddle about 1/3 of each of the outside heatpipes. It would have to b 330mm sq to straddle all 4 heatpipes.

This info will be added to the review.

I'd still say that the xigmatek base can be improved:

1) space the heapipes closer together
2) make the aligning "seat" piece not out of aluminum but copper

walle
Posts: 605
Joined: Tue Dec 26, 2006 10:52 am

Post by walle » Fri May 02, 2008 5:00 pm

Nice of you to shime in dhanson giving us the numbers bringing it to our attention. I’ve never been giving this any thought to be quiet honest but merely noticing that the footprint of the GTX card was that much bigger compared to that of the ATI card, I mistakenly including the GT and GTS cards (which easily could have been avoided). Nevertheless, I’ve learned something new here well worth keeping in mind till next time, your thoughts around larger GPUs in the future was a well worth pointer too as I see it, after all, its nice if the cooler you spend money on could last a second generation of cards, at least different models without having to replace it (thinking of Thermalright here although that one was more related to a different mounting bracket, which in its own right was somewhat of a disgrace on their part as far as I’m concerned)

MikeC wrote: I'd still say that the xigmatek base can be improved:
No disagreement on that one Mike and (as an aside) I mentioned in another post regarding one of Xigmateks HSF coolers that narrowing down the spacing between the heatpipes would be beneficial. I apologize if I came across as attacking you in my first input in this thread, it was not an attack nor meant to be perceived as such, felt the need to clarify that.

Your second point about making the aligning seat piece out of copper instead of aluminium would most likely be somewhat more effective. Clearly the Battle-Axe can be improved efficiency wise as you say, but what I found really flabbergasting was actually the poor quality of the mounting gear, I assume Xigmatek are (mainly) targeting Thermalright, and if so, they can ill afford such blunders as I see it.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Fri May 02, 2008 5:26 pm

walle wrote:I apologize if I came across as attacking you in my first input in this thread, it was not an attack nor meant to be perceived as such, felt the need to clarify that.
No problems, not perceived as such. I, too, had my eyes opened to the die size issue by both you and dhanson865.
...what I found really flabbergasting was actually the poor quality of the mounting gear, I assume Xigmatek are (mainly) targeting Thermalright, and if so, they can ill afford such blunders as I see it.
Totally agreed. The mounting hardware is just miserable. I think I could improve it in 15 minutes with better screws and nuts.

BTW, the article has been changed -- see page 4 and the final page.

walle
Posts: 605
Joined: Tue Dec 26, 2006 10:52 am

Post by walle » Fri May 02, 2008 6:30 pm

I’m glad to hear you didn’t take it the wrong way Mike, the update you’ve done looks great and surely covers all the bases, talk about in depth. This is one of the reasons why I hold this page is such high regard, not because you and I and dhanson saw eye to eye on this one (identifying the issue, not at all), but because you strive to achieve the highest possible quality in your reviews, this is something I feel that you’re not getting to hear often enough and as such I wanted to put it out there, for the record.

Time for some sleep over here, getting late.

Cheers

Note:

I could perhaps have chosen a better time for bringing above to your (and others) attention, given it could be viewed as smoothing over what smoothing over didn’t need, yet I swear I’m senile at times so better to go with it before it slips ones mind. :wink:

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Fri May 02, 2008 7:16 pm

Thanks for the note. I just want to say for the record I don't know how accurate the numbers I posted are.

Your average google hit for "die size G80" without the quotes (or any other search term I used) will be an enthusiast site or a news site that specializes in prerelease rumors. I also checked Wikipedia but I can't vouch for the accuracy of that source either.

With any luck someone can provide better data for us after seeing our discussion. If not, at least people reading the review can keep an open mind about the wide variety of GPUs that this product will be facing.

MoJo
Posts: 773
Joined: Mon Jan 13, 2003 9:20 am
Location: UK

Post by MoJo » Sat May 03, 2008 3:25 am

Mike, I think you hit the nail on the head when you said the performance was probably limited by the fact that only the centre heat pipes were making contact.

I imagine nVidia cards would be even worse, since most of their GPUs seem to be very concave. Certainly both the 7800GT and 8800 Ultra I checked are visibly concave.

All GPU coolers have to deal with this problem. It seems like the stock coolers just try to use pressure and thermal paste to close the gap, while some after market heatsinks (like the HR-03) are a bit convex to compensate. Since it's virtually impossible to lap a GPU as it is soldered to the card, it's a bit of a problem.

Considering how large that heatsink is, I don't know why they just don't go for full coverage instead of relying on RAM heatsinks. GDDR3 produces massive amounts of heat - underclocking the RAM on the 8800 Ultra as low as it would go took over 10C of idle temperatures. Shame the driver can't do it dynamically.

cavarr
Posts: 5
Joined: Thu Nov 29, 2007 10:46 am
Location: Poland

Post by cavarr » Sat May 03, 2008 6:03 am

Great review :D, I wait on article with Thermaltake DuoOrb, Thermalright HR-03 GT and Auras Fridge :twisted:

And all will be bright :D

Gillian Seed
Patron of SPCR
Posts: 25
Joined: Sun Apr 13, 2008 9:05 am
Location: England, UK.

Post by Gillian Seed » Sat May 03, 2008 7:31 am

Another excellent article, especially with the added information on die sizes.

I do have one suggestion that I would like to see incorporated into future GPU cooler reviews... I would like to see a photo of the base of the cooler, and outlines of various core sizes overdrawn onto this photo.

Much like this image.

http://www.silentpcreview.com/files/ima ... eaxe19.jpg

Without the heat transfer compound, and with outlines of various cores overlaid. This way anyone can instantly see how much of the core will make contact with the base/heatpipes.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Sat May 03, 2008 8:25 am

Gillian Seed wrote:Another excellent article, especially with the added information on die sizes.

I do have one suggestion that I would like to see incorporated into future GPU cooler reviews... I would like to see a photo of the base of the cooler, and outlines of various core sizes overdrawn onto this photo.

Much like this image.

http://www.silentpcreview.com/files/ima ... eaxe19.jpg

Without the heat transfer compound, and with outlines of various cores overlaid. This way anyone can instantly see how much of the core will make contact with the base/heatpipes.
That's an interesting suggestion, but I wonder how useful it really would be. With all other heatsinks, because they have a one-piece base that makes contact with the die, as long as it covers the die, the heat transfer effectiveness would not change. Things like mounting pressure & evenness, flatness/smoothness of the base, flatness of the die -- these would matter much more. It's only with the bare wide spread heatpipes in the base of the Xigmatek that GPU die size became significant. I can't think of a single video card cooler where base size was too small... or too large.

Gillian Seed
Patron of SPCR
Posts: 25
Joined: Sun Apr 13, 2008 9:05 am
Location: England, UK.

Post by Gillian Seed » Sat May 03, 2008 8:36 am

MikeC wrote: That's an interesting suggestion, but I wonder how useful it really would be. With all other heatsinks, because they have a one-piece base that makes contact with the die, as long as it covers the die, the heat transfer effectiveness would not change. Things like mounting pressure & evenness, flatness/smoothness of the base, flatness of the die -- these would matter much more. It's only with the bare wide spread heatpipes in the base of the Xigmatek that GPU die size became significant. I can't think of a single video card cooler where base size was too small... or too large.
Good point! If another manufacturer produces a heatsink that has heatpipes making contact with the die, I'd be most pleased if there was a photo like I suggested. :-)

colm
Posts: 409
Joined: Tue Jan 31, 2006 8:22 am
Location: maine

Post by colm » Sat May 03, 2008 8:23 pm

that is alot of effort for all those mistakes by xigmazu, i mean xigmatek :)

The heatsink is not functioning given those numbers.
I trialed and errored with the accelero on a hd2600 (rv600 hotty)
and got into the 60s C and new darn well it wasn't correct. So, I shined a light at the base, and it was cockeyed less than a millimeter. I straightened it out, still held by 2 screws (also breaking a stud with the accelero and made a stud setup much stronger) and it functioned incredibly. (49c max no matter what I do, and it is still passive!)

anyway, from the description of the xigmatek base, if to get to a point.. it is way wrong. So very wrong, I would call it broken just to look at it.
Furthermore, the springs are an insane attempt at sustaining pressure. Hot metal is weak metal, and that is when the heatsink is needed the most. An oxymoron if I do say so myself. to see it on such a largely useless heatsink...and rather good lookin one ...
If to go aftermarket oem vga cooler, the accelero has its dainty quirks, but well well worth the reward if to figure them out on your own. Seems to be a trend with these vga coolers...don't get frustrated I guess is an answer, as no one has an easy one.

Alex
Posts: 185
Joined: Sun Dec 17, 2006 12:49 pm
Location: Stockholm

Post by Alex » Sun May 04, 2008 2:25 pm

Edit: Typing errors are now fixed. :D

Thank you very much for this excellent article.

roadie
Posts: 166
Joined: Fri May 12, 2006 2:07 am
Location: Liverpool, UK

Post by roadie » Wed May 07, 2008 6:44 am

As soon as I saw this product, I knew it wouldn't do any better than the Accelero S1 rev2 and Turbo Module I have on my 3870. The product is really wasted without an IHS on the GPU.

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Wed May 07, 2008 6:08 pm

My reaction to this HS is different from most.

My first thought was "why are the fins so small?". My second thought was "why more than two heat pipes?".

Well, I think the first was off-base, but the second was entirely justified.

Think of it this way: even with only two of the heat pipes touching the GPU, and even with the tiny fin cross section, this HS cools the GPU almost as well as an S1, regardless of fan speed.

There are two conclusions to draw from this: the other two heat pipes add no value, and two direct-touch heat pipes (even with tiny fins) can cool pretty much any GPU with minimal air flow.

The deciding factor IMO is the fragile retention hardware. The S1 performs marginally better, but is a more robust product.

Post Reply