SPCR

Nice review, as usual.

Considering the raise in CPU temperature caused by the Asus card, how beneficial would it be to use a duct to provide fresh outside air to the processor in a system with a CPU close to it's thermal borderline?

Tephras wrote:Considering the raise in CPU temperature caused by the Asus card, how beneficial would it be to use a duct to provide fresh outside air to the processor in a system with a CPU close to it's thermal borderline?

Quite beneficial, obviously.... but I think you have to remember the load the system was on -- both the GPU and the CPU were maxed out for 2 hrs. This is highly unlikely, especially for a HPTC. Also, we're deliberately using a CPU whose thermal characteriscs will soon be much higher than the norm for desktops.

Very good article, in particular when comparing it with the previous one on Gigabyte (for AGP card)..thanks.

In addition to the VGA card review, it's also interesting (at least for me) to note the very different baseline results primarily due to Northwood/Prescott cores.

A small nitpicking on typo - on page 4 about half way where the procedure to estimate power consumption is described, at the end of #1 it refers to Gigabyte card which is not the subject of this review. I guess you mean to say VGA card.

great review! Shame about the 6600 - I'd imagine alot of people will be disappointed it performed so badly. Will you get any feedback from the manu about the cause of failure?

is there anyway to measure the GPU temps on any of these cards? I know some cards don't allow this so it would lose it's value for comparative purposes, but it would be useful direct (rather than inferred) information. It'd also give a better idea of how much leeway there is for lowering airflow through the case, especially if a lower power CPU was to be used.

is there anyway to measure power consumption directly, in a similar way to Xbitlabs?

keep up the good work!

Excellent review(s).

I actually own the AOpen 6600 card, and it's been sort of a love/hate relationship. Getting it to work with WoW was a royal pain (to get it to work, you need to run WoW with the -opengl flag. Otherwise, expect crashes 10-20 seconds into gameplay. From researching the problem, I believe it has more to do with the 6600 chipset than AOpen's implementation, but I could be mistaken). I've also had issues with it overheating and restarting when gaming; direct airflow helped that quite a bit.

I would not recommend the AOpen based on the experiences I've had with it unless someone absolutely needed a passive low-profile card. It works fine for Media Center and playback in general (no problems whatsoever), but gaming has been perilous, to say the least.

Lastly, I would love to see a review of this card!. Either that, or just a normal 7600gs.

mattthemuppet wrote: is there anyway to measure power consumption directly, in a similar way to Xbitlabs?

I'd second that request as well--directly would be preferable to estimated.

The Asus EAX1600XT Silent is a very nice card--I know, I've had two and RMA'd both of them. So be aware of the potential compatibility problems with nForce4 chipsets, particularly on Asus boards.

The problem is with some X1600Pro and X1600XT cards and some nForce4 implementations--the problem mostly manifests itself by freezing the system in 3D apps, though for some (including me), 2D was affected as well.

There's a 30 page thread about this on the Sapphire forums. http://www.sapphiretech.com/en/forums/s ... php?t=2413

Additionally, there are similiar comments on some of these X1600 boards on Newegg.

The Sapphire Forum thread suggests that Sapphire recalled their X1600Pro cards and have added higher value caps to newer boards to fix the problem.

Some people have managed to solve the problem by increasing voltage to the PCI-E slot, but many (most?) BIOS's don't have this option. Others report success by running the PCI-E at 8X instead of 16X. Again, not everyone can do this.

There is some indication that the problem is due to "slightly weak 12v power rail to PCI-Express" on the affected motherboards. (I have no idea, but it sounds logical.)

In my case, after extensive troubleshooting and multiple installs, I was finally told by Asus tech support that there was probably an incompatibility with the motherboard.

The motherboard was an Asus A8N5X.

There are a few passive ones out there and they should run cooler than the 6600 due to the smaller fab process. Both EVGA and XFX have passive 7600GS cards. Might be a good followup to compare the EAX1600XT Silent to a pair of 7600GS cards.

-Derek

mattthemuppet wrote:is there anyway to measure the GPU temps on any of these cards? I know some cards don't allow this so it would lose it's value for comparative purposes, but it would be useful direct (rather than inferred) information. It'd also give a better idea of how much leeway there is for lowering airflow through the case, especially if a lower power CPU was to be used.

is there anyway to measure power consumption directly, in a similar way to Xbitlabs?

Short answers: No, and no.

After the testing was completed, somebody brought it to my attention that it is possible to hack the BIOS on 6600 based cards to enable thermal monitoring. However, since our 6600 sample was no longer running, it was obviously not possible to do a re-test.

The Radeon based card requires a specific chip (called RS-53?) to monitor the temperature, which was missing from the Asus card.

For direct measurements of power consumption, we would need to custom build a PCIe pass-through that can tap the power. This is hardly a simple undertaking. It may happen eventually if there's enough demand and we find some free time, but don't hold your breath.

Page 2:

The heatsink is quite light, yet does not sacrificing cooling surface area.

I believe that should be "does not sacrifice."

For direct measurements of power consumption, we would need to custom build a PCIe pass-through that can tap the power. This is hardly a simple undertaking. It may happen eventually if there's enough demand and we find some free time, but don't hold your breath.

Perhaps something like so?

MikeC wrote: My comment about current VGA cards in general:

1) The graphic card form factor is NOT designed to handle components that generate such incredble amounts of heat. It's basically a silly little "daughter" card with no way to anchor proper heatsink/cooling devices, especially with any real mass. On top of that, the exposed die is on the "wrong" side for the typical ATX tower style case.

2) Two things should be forced on the GPU makers --
-- some way to shame/limit them to.... say, 60W.
-- force them to adopt some form of dynamic clock/voltage adjustement, like CnQ, SpeedStep, etc. and dictate a maximum idle state power -- like 25W.
-- create a new form factor in conjunction with mobo, case and PSU makers that makes better thermal sense of the whole shebag.

3) Anything even approaching 100W is obscene, really, especially when they're encouraging two of them for one box.

I see two trends occuring in the PC industry right now. The first is a trend towards CPUs being measured in terms of power consumption verses performance. Intel's power consumption was/is so gluttonous that it became totally obvious that change had to be made, hence the core/conroe movement. I think in the next several years, CPUs are going to really move towards being much more efficient.

The second trend is towards video cards (plural now) actually bringing 300+ watt PSUs to their knees. So, with that in mind, I think the first trend is going to be less of a factor in "silent" computing than the second, especially since Mike's comments make it pretty clear that PC architecture was never designed to have daughter-boards dissipating 90+ watts.

Not that I'm trying to rag on you--I'm sure there are real engineering hurdles to measuring video card consumption directly (never know--maybe there's a few double-E's in SPCR.com's readership that could pitch in a hand). That being said, video cards are where I see the most egregious power consumption occuring in the next several years.

that's cool Devon, I figured there were good reasons, I was just curious. I also realise that the testing methodology will evolve with time, in the same way as everything else on here (the PSU testing methodology springs to mind) does.

still great to see that graphics cards have arrived on SPCR in a big way and without endless reams of benchmarks too

These are essentially quite HTPC oriented boards right? It would have been interesting if the Asus one was mounted in a horizontally oriented case. Maybe the convection would have been less optimal from the cooler?


Johan

We did this in the review for the NSK2400. No problems were observed.

Devonavar wrote:We did this in the review for the NSK2400. No problems were observed.

Yeah, read that review right after and spotted the Asus card. Sorry

Why isn't there any powerful graphics cards made with passive cooling? The olnly exception was the 7800GT from Asus, that was cooled by a massive heatpipe. Otherwise, it is only low- or mid-rang cards like X1600 and 7600GS.

It is possible to cool a 7900GT without active cooling, they don't produce too much heat. I think the manufacturers should produce more passive-cooled cards.

Now when that isn't the case, I think SPCR should test the passive VGA-coolers.

First of all, I can not understand why a low profile card does not use the DMS-59 connector which can be splitted into dual VGA or dual DVI. There are -- as far as I know -- four low profile Nvidia cards that use this connector: the PNY Quadro 280/285 (AGP/PCI-E), they are on Geforce4 MX level (many HP systems uses these). The Dell G0170, based on the FX5200. And finally, EVGA has a 6200TC modell, I advise Googling on dms-59 geforce 6200. These are slow but passive in the traditional sense -- no bulky heatpipe coolers. A faster card with heatpipe would have been a nice addition.