NVIDIA 680i SLI: Official 1333MHz FSB CPU Support Arrives
by Gary Key on July 20, 2007 12:01 AM EST- Posted in
- CPUs
Quick Thoughts:
At first glance, all facts indicate the NVIDIA 680i chipset fully supports the new 1333MHz FSB processors. Dare we say that at times the performance capability of the 680i chipset even excels when utilizing the latest and greatest CPU from Intel? This is in stark contrast to several of our test results when the P35 was first launched and indicates NVIDIA has been busy tweaking the BIOS to gain some additional speed. If only NVIDIA had spent some additional time tweaking their Vista drivers then we would be truly happy - not really, but close enough.
There was nothing really surprising in the initial test results from a historical viewpoint. Both chipsets showed their strengths in certain areas and held on tight in the others. In applications that are GPU or storage system sensitive, the 680i consistently finishes first in most of our benchmarks. It is only in the memory bandwidth, CPU throughput, or latency sensitive applications that we see the Intel P35 chipset pull away, although the differences for both are minimal in nearly all cases.
As far as the new P30 BIOS goes for the EVGA 680i SLI board, there are not any additional settings or features when compared to previous releases. However, official 1333FSB support is present and working properly with a wide range of E6x50 processors that we have tried to date. In fact, the advantages we noticed in lower voltage requirements when overclocking and improved memory throughput with the 1066FSB processors is what impressed us the most about this BIOS. However, not all is perfect in the land of 1333 and that brings us to our concerns about this BIOS.
We had a very difficult time running our memory at settings above DDR2-1000 unless we really relaxed our timings with the QX6850. This did not occur with our E6850 or E6550 processors for the most part - we still had to relax the memory sub-timings compared to the 1066FSB cousins but nothing like we had to do with quad core. We have had some luck running our memory at 1T above DDR2-800 on this board in the past but not this time. We still noticed a couple of "holes" where the straps change, something that is present in all 600i chipsets, but with the QX6850 the frequency and spread seemed worse.
Our 435FSB overclocking result with the QX6850 processor is very good for a 680i based board although we firmly believe it could be better if it were not for the location of the strap changes and memory timing problems. That said, the FSB increases would not be much higher unless additional cooling is present on the MCP/SPP and associated chips around the CPU area. The voltages required to reach the 470+ FSB range with the QX6850 created significant thermal issues with our board when testing 24/7. The heat buildup did not allow us to keep the board at our 8x470FSB setting for more than a few hours during benchmark testing.
We also had problems when setting our memory to Linked and utilizing any sub-setting other than Sync when overclocking the QX6850. The board was not always stable, would not post, or required greatly relaxed timings with the 5:4, 3:2, or even the auto ratio settings once we passed the 385FSB mark. At this time we believe some additional BIOS tuning will solve most of these problems as we have already noticed strap and ratio changes in the latest ASUS, MSI, and Gigabyte BIOS releases that have cured some of their early ills.
Overall, we were pleasantly surprised with the performance of the 680i chipset and the QX6850. In early testing we are seeing very good results with other 600i based boards and have no doubt performance will continue to improve over time. Our advice if you have a 600i based board and want to purchase one of the new 1333FSB based processors is to go for it. Just realize chances are your current 1066FSB Core 2 Duo will probably already run at 1333FSB without too much trouble, so the more likely reason to upgrade CPUs is if you're looking at moving to quad core. Even then, we'd likely stick with the Q6600, especially next week when the new prices kick in.
At first glance, all facts indicate the NVIDIA 680i chipset fully supports the new 1333MHz FSB processors. Dare we say that at times the performance capability of the 680i chipset even excels when utilizing the latest and greatest CPU from Intel? This is in stark contrast to several of our test results when the P35 was first launched and indicates NVIDIA has been busy tweaking the BIOS to gain some additional speed. If only NVIDIA had spent some additional time tweaking their Vista drivers then we would be truly happy - not really, but close enough.
There was nothing really surprising in the initial test results from a historical viewpoint. Both chipsets showed their strengths in certain areas and held on tight in the others. In applications that are GPU or storage system sensitive, the 680i consistently finishes first in most of our benchmarks. It is only in the memory bandwidth, CPU throughput, or latency sensitive applications that we see the Intel P35 chipset pull away, although the differences for both are minimal in nearly all cases.
As far as the new P30 BIOS goes for the EVGA 680i SLI board, there are not any additional settings or features when compared to previous releases. However, official 1333FSB support is present and working properly with a wide range of E6x50 processors that we have tried to date. In fact, the advantages we noticed in lower voltage requirements when overclocking and improved memory throughput with the 1066FSB processors is what impressed us the most about this BIOS. However, not all is perfect in the land of 1333 and that brings us to our concerns about this BIOS.
We had a very difficult time running our memory at settings above DDR2-1000 unless we really relaxed our timings with the QX6850. This did not occur with our E6850 or E6550 processors for the most part - we still had to relax the memory sub-timings compared to the 1066FSB cousins but nothing like we had to do with quad core. We have had some luck running our memory at 1T above DDR2-800 on this board in the past but not this time. We still noticed a couple of "holes" where the straps change, something that is present in all 600i chipsets, but with the QX6850 the frequency and spread seemed worse.
Our 435FSB overclocking result with the QX6850 processor is very good for a 680i based board although we firmly believe it could be better if it were not for the location of the strap changes and memory timing problems. That said, the FSB increases would not be much higher unless additional cooling is present on the MCP/SPP and associated chips around the CPU area. The voltages required to reach the 470+ FSB range with the QX6850 created significant thermal issues with our board when testing 24/7. The heat buildup did not allow us to keep the board at our 8x470FSB setting for more than a few hours during benchmark testing.
We also had problems when setting our memory to Linked and utilizing any sub-setting other than Sync when overclocking the QX6850. The board was not always stable, would not post, or required greatly relaxed timings with the 5:4, 3:2, or even the auto ratio settings once we passed the 385FSB mark. At this time we believe some additional BIOS tuning will solve most of these problems as we have already noticed strap and ratio changes in the latest ASUS, MSI, and Gigabyte BIOS releases that have cured some of their early ills.
Overall, we were pleasantly surprised with the performance of the 680i chipset and the QX6850. In early testing we are seeing very good results with other 600i based boards and have no doubt performance will continue to improve over time. Our advice if you have a 600i based board and want to purchase one of the new 1333FSB based processors is to go for it. Just realize chances are your current 1066FSB Core 2 Duo will probably already run at 1333FSB without too much trouble, so the more likely reason to upgrade CPUs is if you're looking at moving to quad core. Even then, we'd likely stick with the Q6600, especially next week when the new prices kick in.
18 Comments
View All Comments
FireTech - Saturday, July 21, 2007 - link
Seconded.
Actually, it's over four months now. First post in http://forums.anandtech.com/messageview.aspx?catid...">mATX Roundup was on 15/03/07....
No matter, we do have another Intel focussed article to read while we wait for it ;)
DigitalFreak - Friday, July 20, 2007 - link
Good job! Want a cookie?
JKing76 - Friday, July 20, 2007 - link
No thanks, the thrice-promised uATX roundup would be fine.CrystalBay - Friday, July 20, 2007 - link
It must get really boring to keep testing MB's without much more than 2% performance differences.... Parity blows.....poohbear - Friday, July 20, 2007 - link
nice article, but the above was a bit much for a "professional" article. no offense, i like anantech for their technical knowledge not their ability to ramble on w/ half-hearted jokes. do i really need to read all that before understanding all u needed was a P35 board?
Chunga29 - Friday, July 20, 2007 - link
I actually thought the snarky comments at the beginning was the most enjoyable part. All NVIDIA's marketing hype kicking in bragging about their platform, when the reality is that it's nothing really special. The more pressing concern was the next sentence:Okay, so that's just one more reason to stick with XP for me, but seriously are you telling me that NVIDIA's launch Vista drivers were so perfect that no updates are necessary? I'd say their driver teams are probably overworked right now, but regardless if I were looking at Vista that situation wouldn't make me very happy.
BitJunkie - Friday, July 20, 2007 - link
I've had this board and a 7950GX2 card from nvidia since December 2006 with an E6600 plugged in to it. It's been stable at a 1333 FSB with memory running at 1066 since then all be it with slack 5-5-5-15 timings.On the whole it's a fast(ish) and stable system BUT....and it's a BIG but the drivers have been the weak link since I installed Vista in January. Not real change for me in that time, I STILL get BSOD'd regularly, the SLI functionality is borked for the gfx card and up until new BIOS my raid array keept locking up on boot.
Expecting them to deliver the goods for the newer 1333 fsb quad core CPUs is kind of like pissing in to the wind. I so hope amd / ati get their act together soon, this is getting old.
Sunrise089 - Friday, July 20, 2007 - link
Agreed - I can find benchmarks anywhere, and this sort of article (unlike say the memory wall one) doesn't require a lot of technical expertise - so why not liven it up a bit.