NVIDIA 680i SLI: Official 1333MHz FSB CPU Support Arrives
by Gary Key on July 20, 2007 12:01 AM EST- Posted in
- CPUs
QX6850 Quad Core Overclocking:
We need to state up front that our QX6850 sample is not the best of the lot when it comes to overclocking. In fact, it seems downright disappointing to us as it does not matter what voltages or board we utilize: this processor refuses to POST once you dial in 4GHz.
We are utilizing the recently released P30 BIOS that features additional tuning for 1333MHz FSB processors. We did notice some improvements over the P29 Beta but to be honest, the majority of improvements are with the 1066FSB based processors. These improvements range from slightly higher overclocks with lower voltage requirements to improved compatibility with our RAID cards.
Our first test is designed to maximize our CPU speed while also overclocking the FSB as much as possible. We ended up with a final 10x395 setting with our memory operating in Sync mode at DDR2-790 with timings at 3-4-3-9 1T. We noticed with this CPU and board combination that throughout testing we had a choice of either running tight timings near DDR2-800 or having to really loosen up the timings at DDR2-1066 or above in order to maintain stability.
Our OCZ PC2-9200 will normally do 4-4-4-12 timings up to DDR2-1100 but we had to run 5-5-5-15 or higher above 1000 with this CPU. We determined our best memory throughput and latencies were around DDR2-800 so we remained in this range throughout testing. We also tried additional memory modules from Corsair, G.Skill, and SuperTalent with the same results. It appears to us that some fine tuning is still required with the P30 BIOS at this time.
We dropped the multiplier on our QX6850 to eight and were rewarded with a final 8x435FSB setting compared to a 8x475FSB on the ASUS Blitz Extreme board. This board did make it to 8x470FSB with greatly increased voltages, but it was not stable over the time it takes to run our benchmark test suite. The main culprit is heat and even with additional air cooling, the system is just not stable on a 24/7 basis.
We should also note that we experienced one of the infamous 680i "holes" with our component combination that started around 440FSB and stayed until the 465FSB range. That left us with a choice of 435FSB or 470FSB; we really tried to make the 470FSB work on a consistent basis but we just could not get the system stable enough to sign off on it for daily usage. Once again, we had a degree of difficulty running our memory in the DDR2-1066 to 1170 range without resorting to very loose timings with 5-7-7-24 required at DDR2-1170 for stability. This did not occur with our E6850 CPU so the problem seems to be the quad core setup.
While our current EVGA 680i SLI sample tops out around 40FSB lower than our two ASUS P35 boards, we still consider 435FSB with the 680i chipset to be very good. We have a fairly strong opinion about the current FSB race and the detriments it brings to overall system performance and stability, but we'll save it for the roundup.
| EVGA 680i SLI Overclocking Testbed |
|
| Processor | Intel Core 2 Extreme QX6850 Quad Core, 3.00GHz, 8MB Unified Cache 1333FSB, 9x Multiplier |
| CPU Voltage | 1.4625V (default 1.2750V) |
| Cooling | Tuniq 120 Air Cooling |
| Power Supply | OCZ ProXStream 1000W |
| Memory | OCZ Reaper PC2-9200 (2x1GB) 3-4-3-9 |
| Video Cards | 1 x MSI 8800GTX |
| Hard Drive | Western Digital 150GB 10,000RPM SATA 16MB Buffer |
| Case | Cooler Master CM Stacker 830 |
| Maximum CPU OC | 395x10 (3-4-3-9 1T, 790MHz, 2.15V), CPU 1.4625V 3950MHz (+32% CPU) |
| Maximum FSB OC | 435x8 (3-4-3-9, 870MHz, 2.15V), CPU 1.3625V 3480MHz (+31% FSB) |
| . | |
We need to state up front that our QX6850 sample is not the best of the lot when it comes to overclocking. In fact, it seems downright disappointing to us as it does not matter what voltages or board we utilize: this processor refuses to POST once you dial in 4GHz.
 |
| Click to enlarge |
We are utilizing the recently released P30 BIOS that features additional tuning for 1333MHz FSB processors. We did notice some improvements over the P29 Beta but to be honest, the majority of improvements are with the 1066FSB based processors. These improvements range from slightly higher overclocks with lower voltage requirements to improved compatibility with our RAID cards.
Our first test is designed to maximize our CPU speed while also overclocking the FSB as much as possible. We ended up with a final 10x395 setting with our memory operating in Sync mode at DDR2-790 with timings at 3-4-3-9 1T. We noticed with this CPU and board combination that throughout testing we had a choice of either running tight timings near DDR2-800 or having to really loosen up the timings at DDR2-1066 or above in order to maintain stability.
Our OCZ PC2-9200 will normally do 4-4-4-12 timings up to DDR2-1100 but we had to run 5-5-5-15 or higher above 1000 with this CPU. We determined our best memory throughput and latencies were around DDR2-800 so we remained in this range throughout testing. We also tried additional memory modules from Corsair, G.Skill, and SuperTalent with the same results. It appears to us that some fine tuning is still required with the P30 BIOS at this time.
 |
| Click to enlarge |
We dropped the multiplier on our QX6850 to eight and were rewarded with a final 8x435FSB setting compared to a 8x475FSB on the ASUS Blitz Extreme board. This board did make it to 8x470FSB with greatly increased voltages, but it was not stable over the time it takes to run our benchmark test suite. The main culprit is heat and even with additional air cooling, the system is just not stable on a 24/7 basis.
We should also note that we experienced one of the infamous 680i "holes" with our component combination that started around 440FSB and stayed until the 465FSB range. That left us with a choice of 435FSB or 470FSB; we really tried to make the 470FSB work on a consistent basis but we just could not get the system stable enough to sign off on it for daily usage. Once again, we had a degree of difficulty running our memory in the DDR2-1066 to 1170 range without resorting to very loose timings with 5-7-7-24 required at DDR2-1170 for stability. This did not occur with our E6850 CPU so the problem seems to be the quad core setup.
While our current EVGA 680i SLI sample tops out around 40FSB lower than our two ASUS P35 boards, we still consider 435FSB with the 680i chipset to be very good. We have a fairly strong opinion about the current FSB race and the detriments it brings to overall system performance and stability, but we'll save it for the roundup.
Facebook
Twitter
RSS
18 Comments
View All Comments
TA152H - Saturday, July 21, 2007 - link
I see this Asus Blitzkrieg Extreme, but what it is? It kind of gets thrown in there with no explanation, but I can't find anything that describes it, or any mention of it on Asus's website.Don't you think you should have at least mentioned what this thing is? I mean, I gathered it's a P35 DDR3 motherboard, but then so is the P5K3. So, obviously something is different, but no explanation is given.
Chunga29 - Sunday, July 22, 2007 - link
What's an ASUS Commando? Originally it was a P5B Deluxe plus $150. I would guess the Blitzkrieg is about the same: tuned for a bit higher OCing, but really just a more expensive mobo that isn't necessary for 99% of users.Review forthcoming, I'd assume, Gary?
TA152H - Saturday, July 21, 2007 - link
Show me someone with a Nvidia chipset for Intel processors, and I'll show you an idiot. Why would anyone in their right mind even consider a non-Intel chipset, except for real low power uses (which even Intel does for some of their motherboards, from SIS). Intel easily makes the best chipsets and more importantly they are the standard and almost as importantly are extremely well supported by the parent company.Nvidia is fine for a minor company like AMD that can't make a decent chipset, or hasn't chosen to, but for Intel that makes a much broader line of products, which are not only excellent but extremely well supported, Nvidia makes no sense at all. It's not only that, they don't manufacture their products, and now that Intel is moving their chipsets to more modern manufacturing processes, it's obvious they are taking them very seriously.
It's a good review though, just so everyone can see there is no point in Nvidia. Otherwise, there would be those lingering doubts.
ATWindsor - Saturday, July 21, 2007 - link
Theese are extremly seriuos problems with a motherboard, why not focus abit more on functionality and stability and less on overclocking in the reviews, so the users won't get theese nasty suprises?AtW
rjm55 - Friday, July 20, 2007 - link
nVidia can crow all they want about 1333 support, but the fact is their dhipsets don't support DDR3. With the new DDR3-1600, DDR3-1666, and DDR3-2000 low latency parts that have started shipping, nVidia has nothing to crow about. DDR2 has just become obsolete with these new DDR3 parts and nVidia can only handle DDR2. Guess these expensive boards will soon be on closeout at Newegg.TA152H - Saturday, July 21, 2007 - link
I agree, but they probably needed to do nothing to this part, just say it runs at 1333 MHz and thus is compatible with the newer products.Clearly they need a DDR3 chipset, but then again, what is the point of Nvidia chipsets for Intel processors anyway? It's not like Intel's aren't fantastic, or that they overprice them. I don't see any value-add to Nvidia in this space, they take too much power and offer essentially nothing. But, if they insist on staying, they need to get DDR3 out and fast. DDR2 is obsolete, and their chipsets thus are as well.
DigitalFreak - Friday, July 20, 2007 - link
ROFLMAO! You have become obsolete.Here's your sign.
Owls - Friday, July 20, 2007 - link
We lined up several 600i motherboards on the test bench, popped our QX6850 in each one, flipped the switch, and then took several deep breaths as board after board failed to run properly. A couple of boards would not even POST, a few worked fine, and some would boot into Vista and then act strange - not Britney Spears strange mind you; more like Paris Hilton behind bars. You just knew that QX6850 wanted to escape its confines and party all night long, but instead it was limited to a few whimpers and constant pleas for help from its socket induced hell.Get to the point. You don't get +1s for being funny in a professional article, what a joke.
gigahertz20 - Friday, July 20, 2007 - link
Go somewhere else then if you don't like a little humor in the articles. I thought it was funny and plus it's not like we pay to read Anandtech articles, so stop bitching about something you get for free.JKing76 - Friday, July 20, 2007 - link
I notice that this is not a micro-ATX roundup. I notice that a week has elapsed from the specific day we were told the roundup would start. And of course, everyone in the uATX Results thread has noticed that we've been jerked around for three months about this supposed review.