Arctic Cooling Accelero S1 VGA Cooler

Want to talk about one of the articles in SPCR? Here's the forum for you.
Keto
Posts: 6
Joined: Mon Dec 31, 2007 1:31 pm
Location: N'awlins

Post by Keto » Mon Dec 31, 2007 1:32 pm

Anyone know a store that is confirmed selling Rev. 2 yet?

mattthemuppet
*Lifetime Patron*
Posts: 618
Joined: Mon May 23, 2005 7:05 am
Location: State College, PA

Post by mattthemuppet » Wed Jan 02, 2008 8:03 pm

I wonder if it's power regulation circuitry or RAM cooling that's caused the artifacts at 5+7V in the passive test? The VF900 comparison made me think this might be the case as at 7V the VF900 cooled card had a higher GPU temp but lower card temp (without artifacts) than the S1 passive/ case fan at 7V (with artifacts). The stock heatsinks are designed to cool those components whereas the S1 will just leave them heating up underneath the card and HS.

Obviously any fan on the S1 will provide enough air to cool those components down enough, but even a bit of a (directed) breeze from a nearby fan might be enough. Having a fan hanging off the end of the S1 might do just as well without blocking another slot.

Just a thought :)

pcy
-- Vendor --
Posts: 137
Joined: Sat Aug 13, 2005 2:46 am
Location: Central London, England

Post by pcy » Tue Jan 08, 2008 7:18 am

Hi,


Do we have an actual power output available for the graphics card used in the test? so we could caluclate the actual C/Watt performance of this cooler and compare it with CPU coolers.


If I understand correctly the S1 was keeping the GPU 10C above local ambient, which puts most CPU coolers to shame.

Any chance of modifying this to use as a CPU cooler?

Do you damage the internal wick if you bend the heat pipes?




Peter
[url=http://www.paqt.co.uk]PaQ[/url]

Luminair
Posts: 223
Joined: Wed Mar 21, 2007 10:45 am

Post by Luminair » Tue Jan 08, 2008 8:40 am

pcy wrote:Hi,


Do we have an actual power output available for the graphics card used in the test? so we could caluclate the actual C/Watt performance of this cooler and compare it with CPU coolers.


If I understand correctly the S1 was keeping the GPU 10C above local ambient, which puts most CPU coolers to shame.

Any chance of modifying this to use as a CPU cooler?

Do you damage the internal wick if you bend the heat pipes?

Peter
Heatpipes have a certain amount of bend left in them after being bent.

There would be no modifying S1 to work on a CPU, and no reason to either. There are plenty of good CPU heatsinks.

You could compare S1 performance to CPU heatsink performance if you were bored, or you can just compare the number of heatpipes and rough fin surface area for the same effect.

The Ultra 120 Extreme is probably the best CPU heatsink ever, and it got to be that way by having just the right configuration of heatpipes and fins. There isn't much room for improvement over it because you just can't fit many/any more fins in the case and heatpipes over the CPU.

pcy
-- Vendor --
Posts: 137
Joined: Sat Aug 13, 2005 2:46 am
Location: Central London, England

Post by pcy » Tue Jan 08, 2008 9:47 am

Hi,


Yes - but I'm interested in actual engineering, not dogma, and anyway the Ultra Extreme won't fit in my case.


Unless I've mis-read the article, the S1 is showing a temperature rise of 10C over local ambient with a Nexus 120 fan attached running at 5V.

The article does not mention whether that's at idle of ATI burn, so maybe the power output is only 140W.


Even so, that gives a C/W of 10/140 = 0.07 compared to a typical CPU cooler C/W of 0.14 (excellent) to 0.5 (atrocious).

So the S1 is apparently twice as effective as the best CPU cooler. I agree this seems unlikely, but by no means impossible, which is why I asked.

Consider: GPUs do in fact produce upto 300W, compared to a typical Intel Quad core with a TDP just under 100W. The graphics cards seem to be able to run at higher temperatures; but even so cooling a GPU can be more of a challange than cooling a CPU.

The large plan view surface area and relatively thin cooling fins would produce a mcuh more efficient heat distribution compared to the typical CPU cooler. It isn't just about surface area - it's surface area x temperature delta (integrated over the entire cooling surface) that counts. A large thin design is inherantly more efficient.

In addition the S1 has it's cooling fins much more loosley spaced than most CPU coolers, so the fan speed (and hence noise) can be much lower and still provide adaquate airflow. You can see this very clearly in the article on p5 where increasing the fan voltage produces very little improvement ot the cooling.


Overall... if the S1 could be made to fit onto a CPU it might prove to be remarkably effective.




Peter
[url=http://www.paqt.co.uk]PaQ[/url]

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

Post by cmthomson » Tue Jan 08, 2008 12:40 pm

pcy wrote:Even so, that gives a C/W of 10/140 = 0.07 compared to a typical CPU cooler C/W of 0.14 (excellent) to 0.5 (atrocious).

So the S1 is apparently twice as effective as the best CPU cooler. I agree this seems unlikely, but by no means impossible, which is why I asked.
This is not because the S1 is a better heat sink. It is because the GPU is a bare die, whereas the CPU is in a case. This means that GPU/S1 cooling passes through two fewer interfaces (the path is junction-substrate-TIM-base-heatpipe-fin-air) than with a CPU, where the path is junction-substrate-TIM-IHS-TIM-base-heatpipe-fin-air.

In a low-airflow system, a Ninja (or Ninja Mini) would be the CPU heat sink of choice; it shares many of the good properties of the S1. The Ultra Extreme is intended for higher-airflow systems.
Consider: GPUs do in fact produce upto 300W, compared to a typical Intel Quad core with a TDP just under 100W.
The card used in the SPCR article is around 100W.
Overall... if the S1 could be made to fit onto a CPU it might prove to be remarkably effective.
Probably. But then so is a Ninja Mini or an SI-128, and neither needs to be modified.
i7 4790K CPU@4.6 GHz, ASUS Z97-PRO, 16GB G.Skill 2400C10, Intel 335 240GB SSD + WDC EFRX 1TB, Internal i7 graphics, Antec P180 case, Seasonic X-400 fanless PS, Megahalems CPU HS, Nexus 3-pin & AC PWM fans < 600 RPM, AcoustiPack foam, homemade ducts.

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

Post by MoJo » Tue Jan 08, 2008 4:08 pm

pcy wrote:Consider: GPUs do in fact produce upto 300W
What GPU is this? An 8800 Ultra is about 175W TDP. That's for the graphics chip, so the RAM will add some but real world conditions will take some away.

300W for an entire quad core system might be closer to the mark.

I agree that some hard numbers would be interesting, but it's not easy to obtain them without a lab. It's not really SPCR's area of expertise either; you would be better off trying overclocking web sites for that kind of info.
http://world3.net

pcy
-- Vendor --
Posts: 137
Joined: Sat Aug 13, 2005 2:46 am
Location: Central London, England

Post by pcy » Tue Jan 08, 2008 6:46 pm

MoJo wrote:
pcy wrote:Consider: GPUs do in fact produce upto 300W
What GPU is this? An 8800 Ultra is about 175W TDP. That's for the graphics chip, so the RAM will add some but real world conditions will take some away.

According to this review:
http://www.sudhian.com/index.php?/artic ... onsumption
the card used in the SPCR test produces 133W on idle and 293W max.

However this one:
http://www.techpowerup.com/reviews/ATI/X1950XTX/7.html
effectively claims the figures are 30W and 115W, the discrepancy is rather large - makes things hard.

This article:
http://guru3d.com/article/Videocards/428/6/
quotes the total power consumpyion under load of a machine with an 8800Ultra as 371W, which implies the graphics card is producing solidly over 200W, though not 300W. I agree.

I saw the power for overclocked 8800s quoted at over 300W, though, early last year. May have been moonshine of course.

Maybe all these articels report total system power consumption (because that's easy to measure) and then ascribe all of it to the GPU, which would explain a great deal. Even so, some of these GPUs are cleary dissipating well over 100W under load, and a few must be hitting 200w if total system power consumption of nearly 400W is correct.

And if not... what are all these PSUs in the 600W to 1000W for (even if actual output is no more than half their nominal capability).


That's why, initially, I asked if we actually knew wha the Power output of the test card was...




Peter
[url=http://www.paqt.co.uk]PaQ[/url]

seraphyn
Posts: 322
Joined: Wed Nov 28, 2007 1:26 pm
Location: Netherlands

Post by seraphyn » Wed Jan 09, 2008 1:36 am

pcy wrote:And if not... what are all these PSUs in the 600W to 1000W for (even if actual output is no more than half their nominal capability).
Triple SLI / Crossfire + 10 HDD's and Dual Quadcore setups, maybe. :P

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

Post by MoJo » Wed Jan 09, 2008 5:20 am

pcy wrote:According to this review:
http://www.sudhian.com/index.php?/artic ... onsumption
the card used in the SPCR test produces 133W on idle and 293W max.
It looks like they measured the power draw for the entire system, not just the graphics card. 133W for the whole system seems reasonable.

For comparison I have an 7800GT, X2 4200+, two HDDs and water cooling which does about 110W idle.
This article:
http://guru3d.com/article/Videocards/428/6/
quotes the total power consumpyion under load of a machine with an 8800Ultra as 371W, which implies the graphics card is producing solidly over 200W, though not 300W. I agree.
They were using a 680i chipset mobo, and those things run very hot. Say 50W for the mobo. Core 2 Duo, say 60W. They have water cooling, 1 WD Raptor so say 20W for other bits.

8800 Ultra 175W
Mobo 50W
CPU 60W
Others 20W
Total 305W

Factor in say 75% PSU efficiency, that gives 381W.
And if not... what are all these PSUs in the 600W to 1000W for (even if actual output is no more than half their nominal capability).
I bet if you really tried to draw 1KW from most of those they would explode.

Even a top end quad core system with multiple HDDs and an 8800 Ultra would be okay on a good quality 500W supply. You only really need more if you are going SLI.
http://world3.net

Tephras
Posts: 1140
Joined: Tue Sep 07, 2004 11:03 am
Location: Europe

Post by Tephras » Sun Jan 13, 2008 6:34 am

pcy wrote:According to this review:
http://www.sudhian.com/index.php?/artic ... onsumption
the card used in the SPCR test produces 133W on idle and 293W max.

However this one:
http://www.techpowerup.com/reviews/ATI/X1950XTX/7.html
effectively claims the figures are 30W and 115W, the discrepancy is rather large - makes things hard.
According to X-bit labs the power consumption of the Radeon X1950 XTX is 33W at idle and about 125W at load. Their method for measuring VGA power consumption is detailed here and here.

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

Post by MoJo » Sun Jan 13, 2008 8:58 am

http://www.hexus.net/content/item.php?item=1359&page=19

Scroll down a bit. That's the difference between a real 650W PSU and a QTec.
http://world3.net

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

Post by roadie » Wed Jan 30, 2008 3:41 pm

How many PCI slots does the S1, plus the 120x25mm fan take up?
I Customised Antec Solo I Antec NeoHE 550W I MSI P67A-GD65 I Intel i5 2500K @ 4.5Ghz I
I 8Gb Samsung @ 1866Mhz I Corsair Force GT 60Gb I Zotac GTX 670 I Dell SP2208WFP I Windows 8.1 x64 I

martdj
Posts: 5
Joined: Mon Nov 13, 2006 7:59 am

Looks like 3

Post by martdj » Fri Feb 01, 2008 1:21 am

Judging from the pics in the review it looks like 3 slots, which is a lot.
I was thinking of using a 92mm fan on the had drive bay of my Antec Solo blowing sideways on the Accelero to provide proper airflow without wasting another slot. What do you guys think. Would that work? I want to cool a Ati HD3850 card.

Edit:
I actually found a pic from someone who put their Accelero in an Antec Solo case and used a 12cm fan and it shows it's using 4(!) slots.
http://www.hardforum.com/showthread.php?t=1261100
Using the Arctic Turbo Module should keep it down to 3 though.

Putki
Posts: 20
Joined: Sat Feb 16, 2008 4:32 am
Location: Finland

Post by Putki » Thu Mar 06, 2008 1:36 am

I ordered this cooler (Rev. 2) and a 120mm Nexus Real Silent for 8800GT. Two 92mm fans would fit there well and 120mm fans also somehow. But I was just wondering does anyone use this with two fans. Is it better to have one blowing air in the middle or two in the edges? For comparision: Accelero S1 Rev. 2 costs 20€ which is $30.66, Nexus Real Silent 120mm 8.42€=$12.90 (I bought two of these: one for S1 and one for Ultra-120 eXtreme) and Nexus Real Silent 92mm 7.30€=$11.20 (I bought one to be put in the front of the case). I also ordered Antec Sonata Plus 550, which is the same case than P150 but with a 550W NeoPower PSU. So it is going to be a silent system. If someone knows is S1 a lot better with 2x92mm fans or 2x120mm fans than with one, it would be nice to know. I could put two on S1 and take TriCool away from the back of the case and put there a Nexus. Thanks.

And Noctua NF-S12 is everywhere around 20€=$30.70, (Inno3D) 8800GT 512MB (cheapest price in Finland) 169€=$259, Thermalright Ultra-120 eXtreme (cheapest price in Finland) 45.60€=$69.90 :P :?

PUTALE
Posts: 1
Joined: Thu Jul 13, 2006 10:21 pm

Post by PUTALE » Thu Mar 06, 2008 2:29 am

ac makes some of the best coolers.

Immortals
Friend of SPCR
Posts: 77
Joined: Mon Dec 12, 2005 5:02 pm

Post by Immortals » Sun Mar 16, 2008 3:04 am

Wow. I replaced my "silent pipe 2" gigabyte 2600xt with one of these. temps are about 12-15 degrees cooler at idle and 20 - 25 degrees cooler at load. Obviously I have rev. 2 and I had no problems sticking on the memory heatsinks. I don't know whether it was because they have upgraded their solution or because my Gigabyte didn't have any cooling on them thus making it unnecessary for me to clean off

LuckyNV
Posts: 68
Joined: Thu Jan 18, 2007 4:26 pm

Post by LuckyNV » Sun Mar 16, 2008 3:30 am

Immortals wrote:Wow. I replaced my "silent pipe 2" gigabyte 2600xt with one of these. temps are about 12-15 degrees cooler at idle and 20 - 25 degrees cooler at load. Obviously I have rev. 2 and I had no problems sticking on the memory heatsinks. I don't know whether it was because they have upgraded their solution or because my Gigabyte didn't have any cooling on them thus making it unnecessary for me to clean off
most likely the latter, stock thermalpads tend to leave a very thin greasy layer on the RAM chips, even thorough cleaning it can still be present.

ErickS
Posts: 11
Joined: Mon May 19, 2008 8:17 am
Location: Israel

My experience with the Accelero S1

Post by ErickS » Fri Jul 11, 2008 6:22 am

I must say that it wasn't easy at all for me to install the cooler on my Gigabyte Geforce 8800gt, due mainly to its enormous size.
I skipped adding those plastic L + T parts, because they just didn't fit in well at all. But, they seemed to be pulling on the fins, and moving the entire heatsink diagonally, instead of it being aligned with the card beneath it. That was when the screws were only loosely attached to the plate, like the instructions said. Fortunately, seems like it's ok without them.
I also had a hard time fastening the screws diagonally like the instructions say. It was fairly easy doing it on a Zalman vf-900cu because of the different build, screw types, and size.
The Accelero, on the other hand, felt too 'free floating' - until I had at least a couple of the screws tightened, but I had to tighten the screws to improve my grip on it, which of course led to an uneven turning of each individual screw.
I felt like I did it all wrong, and that the contact just wasn't as uniform as it is supposed to be.

I also had to take out the screw-less PCI holders that my case came with, so the heatsink could fit in. I actually enjoyed that feature - not having to mess around with any screws every time I take out a PCI card. Well, too bad.
It also seems like It will be difficult to pull out the PCIE power connection, as there's no way I could fit a finger to lift the plastic part that secures it in place. Hopefully the card won't give me problems anytime soon.

Now for the good part - I was actually pleasantly surprised that despite my less than adequate job at installing this enormous heatsink, it remains relatively cool.
Bear in mind that it is summer, I live in a hot country, and that my room is ever hotter due to its poor location.

The card came originally with a Zalman vf-830 cooler, which was terribly noisy to my liking. On its full speed:
Idle: 49-50C | Load: 65-67C

Swapping it for my trusty old vf-900, I was disappointed to get:
Idle: 47-50C | Load: 67-72C

With the Accelero and the turbo module, I'm getting:
Idle: 45-46C | Load: 55C

I'm yet to see it get to 56C outside of Furmark, but these temperatures are good enough for me.

Now, all I have to do is cool down my Northbridge with a fan, and I'm set :)

P.S - It's really quiet, even with the turbo module. A little too quiet, perhaps :)

Post Reply