awesome review
interesting to see how the lack of heatpipes can make so much difference, i would have thought they have tested it before producing them, oh well.
i wonder how it would compare if i slap on my accelero twin turbo on a reference card...

Thanks for the review Mike and Lawrence!

Seems like a bizarre decision to omit the heatpipes.

For anyone who's interested in the comparison, my shiny new Sapphire 5870 VaporX using the default BIOS profile idles at around 1,050rpm and under FurMark load this rises to around 2,200rpm (max load temp is about 80C).

Great review. Glad to see the Accelero S1 fits. It doesn't fit my 5770 =(

Did you measure idle/load temps for this card with the Accelero installed? I'm curious to see whether the Accelero and a modest 120mm fan can keep the card cool...

I too am curious about the S1.

+1 for S1 results. The picture of it mounted is a massive tease! I'm guessing that a good 120 at 900rpm will work nicely.

Thanks for the review!

I`m curious as to how the load VRM and RAM temps are on these vid cards are, as they are often a lot worse than core temps.

Nice review; it's nice to see the high quality reviews applied to products that MIGHT not be the best for sliencer purists However, I agree with others that seeing the S1 mounted but not reviewed was quite a tease! But I personally would probably still get a 5850 to save on power. I need Snow Leopard to support the HD 5xxx first, tho

Thanks again for the review of a "high powered" product. Combined with JamieG's experience with the Vapor-X, maybe a 5870 wouldn't be so bad afterall...

Just be warned that other users are reporting higher idle rpms than mine with their 5870 Vapor-X here.

I'm going to check my BIOS version tonight and try to confirm whether it is different to WR304's.

The other possible culprit is the revision that WR304 mentioned. After doing a bit of googling:

Rev 1 (Black PCB) 100281VXSR has the reference PCB.

Rev 2 (Blue PCB): 100281VX-2SR has a custom PCB that is apparently shorter than the reference PCB.

Based on the info here, I think my 5870 Vapor-X might be a Rev 2 card as I seem to remember the PCIe connectors facing away from the GPU and not from the top of the card, as well as 875/2,500 clocks. However, the info at the bottom of that link suggests that the Rev 1 card should be quieter in idle.

I will check and post back.

@MikeC: maybe you should include some video power tests with OCCT and MSI Kombustor too, as FurMark is the primary "power virus" target for ATI drivers and that might bias the results.

My 5870 Vapor-X is the blue PCB Revision 2 version, with 'rear' facing PCIe connectors as shown in the link above, with clocks of 875/2,500.

BIOS version is VER012.019.000.002.00000.

FWIW, I'm using Catalyst 10.3 on Vista Home Premium 32bit.

Earlier (I think) MikeC has stated, that reviews start by a filtering step, where non-quiet products are sweeped out of the review pipe by a quick test.

I'd love to see a Sapphire Vapor-X 5870 (or 5850) review, EVEN though you have already reviewed a Powercolor 5850 and now this HIS 5870. But these GPUs are the best of the best at the moment that are still viable for a quiet enthusiast rig.

After all, HTPC-grade GPU functions are nowadays trivially achieved by integrated chipsets.

It looks like maybe Sapphire haven't consistently used the same fan model for all the HD5870 Vapor-X cards. My card is a blue rev 2 model with the same BIOS revision as JamieG's.

The higher idle fan speed of 1455 rpm on my card isn't a room/ case temperature issue as it's been quite cold. That's a shame as I'd be very happy with my card if it idled at 1,050 rpm. If I manually put the fan speed of my card up to 100% using MSI Afterburner the fan spins at 3,810 rpm.

I'm still building up to modding mine as that's going to mean cutting through the original fan cable to add a resistor which will void the warranty.

Apart from the idle fan speed the Sapphire HD 5870 Vapor-X cooler seems to work well with some impressively low idle temperatures. The fan does ramp up a bit under load however.


My Sapphire HD 5870 Vapor-X idle temperatures and fan speed (room temperature 13c approx)

Another ATI HD 5870 card with a non-standard cooler that looks quite interesting is the Powercolor PCS+ HD5870.

http://www.hardocp.com/article/2010/03/ ... d_review/1

.

Good review, but the power levels for furmark might be a bit misleading. Artifical benchmarks need to be taken into context. If you measure the FPS for furmark, you might see that the 5870 is generating more graphical computing power than the other cards, for this benchmark anyway. Hence, even with the increased power load, it might actually be quite power efficient.

This is actually the same problem some sites had when measuring the i7 efficiency. The i7 is actually more power efficient than the old 775 CPU's. But because some reviews did not measure the "results" and just the power load against time, they were showing the i7 as less power efficient. In reality, the i7 was more power efficient because it used less power to generate the same result.

To measure power consumption more fairly, you need to try to measure the same thing as much as possible, and preferably in real world usage situations. And if relying on benchmarks instead, include the results not just the power consumption and heat levels.

I agree. I'd like a more 'quiet' oriented 5870 or 5850 product reviewed as well.

These new 5XXX series have power (and heat) efficiency advantages over the previous GPU generation from both on the nVidia as well as ATI. It was actually a bit disappointing that this review didn't focus more on the improved efficiency levels over the previous generation.

Because increased power and heat efficiency "should" translate into more quiet computing in comparable real world situations if taken advantage of, something the HIS cooler clearly does not.

Regarding the improved power efficiency of the 5000 series ATI cards -- I'm not sure I agree that this means lower power consumption in the end, at least not in the normal use context of these cards: Gaming is what they're used for, and a better video card doesn't mean the game is completed quicker. Most likely, it probably means the game is played even longer because it's more fun, plays more smoothly, etc. Similarly, in the context of a gaming PC, the higher performance/power ratio of an i7 does not result in lower energy consumption.

Gaming is quite different from a singular task such as encoding a 1hr video file or applying a photoshop filter script to 100 photos where, when the task is over, the system falls to idle.

No argument there, but that's because it's a different topic.

The point was that furmark is an artifical benchmarking tool and does not reflect actual real world power consumption.

The second point was that if you choose to use an artifical benchmark to reflect power usage rather than real usage, simply showing the resulting power usage of this benchmark, without showing the results in the form of GPU processing power, shows the former out of context, and will penalize cards which do exceptionally well here. This is still fine unless you then decide to show the power usage figures of this artifical benchmark in isolation and infer that this applies to real world power usage.

Case in point: the 5870 is "good" at furmark, far better than the previous generation GPU's, and as a result shows extremely high power consumption compared to the previous generation cards. Taken out of context, this can be misleading as it indicates power efficiency has gone backwards, which is not the case.