EVGA 680i LT SLI: NVIDIA's 680i Cost Reduced
by Gary Key on March 28, 2007 4:00 AM EST- Posted in
- CPUs
Disk Controller Performance
The AnandTech iPeak test is designed to measure "pure" hard disk controller performance, and in this case, we keep the hard drive as consistent as possible while varying the hard drive controller. The idea is to measure the performance of each hard drive controller with the same hard drive.
We play back our raw files that are recorded I/O operations when running a real world benchmark - the entire Winstone 2004 suite. Intel's iPeak utility is then used to play back the trace file of all I/O operations that took place during a single run of Business Winstone 2004 and MCC Winstone 2004. The drive is formatted before each test run and a composite average of three tests on each controller interface is tabulated in order to ensure consistency in the benchmark.
iPeak gives a mean service time in milliseconds; in other words, the average time that each drive took to fulfill each I/O operation. In order to make the data more presentable, we report the scores as an average number of I/O operations per second so that higher scores translate into better performance. The resulting scores are useful for comparing "pure" performance of the storage controllers in this case.
The performance patterns hold steady across both Multimedia Content I/O and Business I/O with the SB600 outperforming the Intel ICH7R, Intel ICH8R, and NVIDIA 680i/650i chipsets in our non-RAID tests. The Promise controller on the DFI board is the slowest of all solutions with the Marvell controller on the Intel 975X performing only slightly better. The 680i and 650i MCP chipsets are different but their disk performance in our tests is similar. We also ran RAID 0 and RAID 5 tests on each controller with the results being in favor of the 680i MCP by around 2% over the 680i LT MCP in these particular benchmarks. Although our NVIDIA boards generally score lower in these "pure" throughput tests, we find their actual performance in disk intensive applications are generally equal to or better than the other solutions.
Power Consumption
Our power consumption numbers are based on our test setup for this article that consists of an MSI 8800GTX video card and our boards have C1E/EIST turned off. The 680i boards lead all in power consumption but the 680i LT SLI is close behind with the LT certainly not meaning Lite in the way of power consumption. The P965 consumes the least amount of power at idle but around 10% more than the RD600 at load. The 975X still offers a decent level of power conservation compared to the other chipsets even though it is getting long in the tooth.
Overclocking
As we mentioned earlier in the article, the overclocking capability of the EVGA board has been neutered to some degree by the limited voltages and options available in the BIOS. Maybe we were just unlucky with our processor choices but this board had enough FSB holes that we thought about using it to whip up a few Swiss cheese sandwiches. After speaking with NVIDIA, we expect the overclocking capabilities of this chipset to vary based upon the luck of the draw when purchasing a board. Of course this is true of any component but it will be interesting to see the results of this latest effort by NVIDIA once the boards are in the hands of owners.
We believe their target is for consistent performance in the 450FSB to 475FSB range and anything above that will be icing on the big green cake. This will be disappointing to users who end up with a board that performs lower than our retail box sample but overclocking is never guaranteed. We will be purchasing a retail board from Newegg shortly to compare against our current board and will update our results if they are significant. Considering that there's a good chance the 680i LT boards are simply using rejected 680i chipsets - chips that couldn't run at higher FSB settings - we would advise caution for anyone looking to get an optimal overclocking platform.
We also found that during overclocking we had to set our CPU voltages a little higher than on the other boards to attain the same or better overclocks. We experienced this with the MSI 650i board and it was generally due to tighter memory/chipset timings in order to improve the performance of the board at stock speeds. Overall, our overclocking experiences were very good and the results of our tests show a board that is still very capable although certainly less than what we believe it is capable of with an enthusiast level BIOS.
The AnandTech iPeak test is designed to measure "pure" hard disk controller performance, and in this case, we keep the hard drive as consistent as possible while varying the hard drive controller. The idea is to measure the performance of each hard drive controller with the same hard drive.
We play back our raw files that are recorded I/O operations when running a real world benchmark - the entire Winstone 2004 suite. Intel's iPeak utility is then used to play back the trace file of all I/O operations that took place during a single run of Business Winstone 2004 and MCC Winstone 2004. The drive is formatted before each test run and a composite average of three tests on each controller interface is tabulated in order to ensure consistency in the benchmark.
iPeak gives a mean service time in milliseconds; in other words, the average time that each drive took to fulfill each I/O operation. In order to make the data more presentable, we report the scores as an average number of I/O operations per second so that higher scores translate into better performance. The resulting scores are useful for comparing "pure" performance of the storage controllers in this case.
The performance patterns hold steady across both Multimedia Content I/O and Business I/O with the SB600 outperforming the Intel ICH7R, Intel ICH8R, and NVIDIA 680i/650i chipsets in our non-RAID tests. The Promise controller on the DFI board is the slowest of all solutions with the Marvell controller on the Intel 975X performing only slightly better. The 680i and 650i MCP chipsets are different but their disk performance in our tests is similar. We also ran RAID 0 and RAID 5 tests on each controller with the results being in favor of the 680i MCP by around 2% over the 680i LT MCP in these particular benchmarks. Although our NVIDIA boards generally score lower in these "pure" throughput tests, we find their actual performance in disk intensive applications are generally equal to or better than the other solutions.
Power Consumption
Our power consumption numbers are based on our test setup for this article that consists of an MSI 8800GTX video card and our boards have C1E/EIST turned off. The 680i boards lead all in power consumption but the 680i LT SLI is close behind with the LT certainly not meaning Lite in the way of power consumption. The P965 consumes the least amount of power at idle but around 10% more than the RD600 at load. The 975X still offers a decent level of power conservation compared to the other chipsets even though it is getting long in the tooth.
Overclocking
As we mentioned earlier in the article, the overclocking capability of the EVGA board has been neutered to some degree by the limited voltages and options available in the BIOS. Maybe we were just unlucky with our processor choices but this board had enough FSB holes that we thought about using it to whip up a few Swiss cheese sandwiches. After speaking with NVIDIA, we expect the overclocking capabilities of this chipset to vary based upon the luck of the draw when purchasing a board. Of course this is true of any component but it will be interesting to see the results of this latest effort by NVIDIA once the boards are in the hands of owners.
We believe their target is for consistent performance in the 450FSB to 475FSB range and anything above that will be icing on the big green cake. This will be disappointing to users who end up with a board that performs lower than our retail box sample but overclocking is never guaranteed. We will be purchasing a retail board from Newegg shortly to compare against our current board and will update our results if they are significant. Considering that there's a good chance the 680i LT boards are simply using rejected 680i chipsets - chips that couldn't run at higher FSB settings - we would advise caution for anyone looking to get an optimal overclocking platform.
We also found that during overclocking we had to set our CPU voltages a little higher than on the other boards to attain the same or better overclocks. We experienced this with the MSI 650i board and it was generally due to tighter memory/chipset timings in order to improve the performance of the board at stock speeds. Overall, our overclocking experiences were very good and the results of our tests show a board that is still very capable although certainly less than what we believe it is capable of with an enthusiast level BIOS.
Facebook
Twitter
RSS
16 Comments
View All Comments
Stele - Wednesday, March 28, 2007 - link
It boils down to the engineering headroom put into the PWM design. Let's say the maximum supply current anticipated from a generation of CPUs is 50A (usually from datasheets and/or design guidelines from the CPU manufacturer). Motherboard designers can design their PWM to be just sufficient for this - most notably, the MOSFETs chosen may be those that can handle around 75A - or they can build some headroom in and choose MOSFETs capable of, say, 100A.
The former design philosophy saves cost, and after all it gets the job done. However, since the components would be running near their design limit, they would generate quite a bit of heat... especially during spikes of load and/or when new CPUs with even higher current draw (e.g. quad-core CPUs) show up.
The latter design philosophy is more expensive, but because the components would be running well below their rated spec, they fare much better in terms of thermal dissipation (and hence efficiency, as less power is lost as heat). The lower running temperatures also help improve the reliability of the components since less thermal stresses are present. Furthermore, when current loads increase - be it due to sudden load spikes and/or power-hungry CPUs and/or overclocking - they still have a lot of headroom, and so are able to handle the extra load without breaking a sweat. This results in better stability and again, lower heat dissipation compared to lower-spec'ed components at the same load.
In view of this, perhaps one other area that Anandtech could look at when reviewing motherboards is to have an IR thermometer handy (the ones that you can point and measure temperatures of surfaces remotely with) and perhaps measure the temperatures of the chipset and PWM (or at least the temperatures of their heatsinks) at idle and load. It may not be perfect (especially when heatpiped, and a hotter heatsink could also mean that the heat flow from the component to the heatsink is good due to good thermal contact) but imho at least it would give a useful ballpark figure.
Stele - Wednesday, March 28, 2007 - link
Pretty good review, critically assessing the chipset in light of theory (paper specs) as well as reality (actual value for money based on real-world prices and competiting products).IMHO, Nvidia's attempt to create a lower-cost version of the 680i SLI by limiting BIOS options and tossing the odd feature out (like passive chipset cooling and two USB/one network port) seems a little clumsy at best... Instead, I think the hybrid chipset combination used in, for example, the Asus P5N32-E SLI Plus is a more elegant solution. As this review demonstrated, the BIOS options do not necessarily cap the LT's overclocking capability enough to prevent it from being a threat to the 680i SLI's exclusive turf. Indeed, it might well turn out that the latter's sales would not be jeopardised by the LT anyway - not because of the arbitrarily imposed limitations, but because of the almost non-existent price difference between boards based on the two chipsets, considering the lost features.
Meanwhile, other reviews of the P5N32-E SLI Plus have noted that the MCP used is that from the AMD-platform 590 SLI chipset. I wonder if, other than the slight rearrangement of PCI-Express lanes, there are any real differences (e.g. revised/improved networking/disk controller engines etc) between the two MCPs? Or are these blocks nothing more than carried over directly from the 590 SLI? It would be great if Anandtech could look into that.
Lastly, a tiny note - it's a little amusing to note how it must be a bit pedantic to have to spell out the full name of the 'solid' capacitors used... at least Anandtech strives hard to get it right! :) Really, though, for the kind of 'solid' capacitors that we're talking about on most motherboards, "aluminium solid electrolytic capacitors" or even just "solid electrolytic capactitors" (vs. 'regular' (liquid) electrolytic) would do perfectly. Conductive polymer capacitors are generally (though not restricted to) the little rectangular ones such as the ones seen on Asus RoG boards as well as the P5N32-E SLI Plus. Just a thought :)
yyrkoon - Wednesday, March 28, 2007 - link
Well, about the capacitors, I know as per some OEM, anandtech was calling them 'solid state capacitors', which you sound like you know enough about electronics to know this is wrong. Anyhow, several readers, including myself called 'foul', and there you have it . . .JarredWalton - Wednesday, March 28, 2007 - link
There were a few complaints when he used "solid capacitors", so I guess the full name is the safe way to go. LOLStele - Wednesday, March 28, 2007 - link
Oh very true! "Solid capacitors" and "solid state capacitors" (as yyrkoon rightly commented) are both commonly used on many hardware sites, and are both inaccurate. One reason for the repeated complaints, iirc, is that we were simply oscillating between one inaccurate term and another. :P"Solid electrolytic capacitors"... now that is perfectly acceptable, despite a difference of one word. The devil, as they say, is in the details. ;)
yyrkoon - Wednesday, March 28, 2007 - link
You did the right thing Jarred, you know how picky 'us' readers are ;)