The 680i based motherboards

If you speak to anyone in Nvidia they will tell you that to get the best out of the 8800 series cards in SLI consumers should use motherboards based on their new 680i chipset. They also claim it is one of the best enthusiast chipsets ever made, so when we were offered one to test we gladly accepted.

What follows would normally be a short section within the review (basically a mini motherboard review) in which we highlight the good and bad points of a particular retail sample. However as we started testing 680i systems in mid November and it is now mid January it’s clear that there is something not right here. So what’s the deal?

The short version is that for us, the reference design 680i motherboards have been a disaster. We'll give a few thoughts on our overall feelings shortly, for now let’s go through some of the more major issues we have experienced. (NOTE: These are not all of our issues, for that we could talk all day).

Our initial sample came direct from Nvidia and was an unbranded model, though identical in look and feel to the Evga and BFG retail models. When we first installed this board and booted into the bios we found there were a number of great options which set the board apart from the competition such as the ability to more easily configure memory speeds in relation to CPU speeds. So we made our way through the bios and configured it to our tastes, including setting our memory to its recommended specifications. Next of course was the install of Windows XP, with our Quad Core CPU installed this should have taken 20-30mins and as we have an automated install disc we were able to head off to the TV and let the system install. When we returned we expected to see the standard XP desktop pic however were greeted by the install screen frozen with about 16 minutes remaining.

At this point we figure it’s a “one off” and the install was restarted. This time we made it to about 7mins remaining and again a freeze. Clearly something wasn’t right and our next step was changing from our automated install to a longer manual XP install. This time we made it through to the windows desktop...unfortunately, any time we tried to install a program or basically do anything the system would blue screen. In addition to this we noted that our second GTX was not recognised by the system. It would turn on with the system (fan on etc) however in Windows we could neither see it nor enable SLI.

Looking into the reboots and freezes further we eventually found that on random boots the board was ignoring our memory timings (but not speeds) and was attempting to boot the memory at 1000MHz with 1T (for example). This happened regardless of memory brand and resulted in complete instability on all sticks.

Luckily there was a new bios released shortly afterwards that fixed this issue and we were back into solving the other issue with the “missing” 2nd Card. To cut an (extremely) long story short our reference sample had a broken slot and eventually ended up in the bin and a retail Evga board was ordered.

Out of curiousity when the board arrived we left the shipping BIOS installed and tried our memory again. Again this board featured the same memory timings bug and we had to continually monitor the timings on each boot. With this bug confirmed we updated the bios to the latest available and installed Windows XP. To our complete shock we found that again our second GTX was not identified. Surely we couldn't have been unlucky enough to have received two boards with broken second slots! Time for some internet research to see if other users were experiencing the same issues. What we found was that many quad core CPU users were experiencing the same issue of 8800GTX SLI not working. The solution? revert to an older bios ... which we did.

At this point we had to contend with the memory issues and in addition the bios (which had to be used to get quad CPU and SLI working) had a new bug, random CMOS resets where the board would boot with everything set to default. This “safe mode” is meant to kick in when overclocking the board too high and will result in the end user being able to enter the bios and fix their last change. Unfortunately this was happening to us at stock speeds. We also noted that a further bug was evident, no display on POST with two cards installed. The system would start up but on random occasions it would not result in a display on our monitor and we had to hit reset until it properly kicked in.

After a couple of weeks of having to work around the memory and bios issues a new bios became available which fixed the Quad core/SLI bug and the memory/safe mode bug.

Now up and running with the SLI we began to think about performance testing on the board and installed the various games and applications required for us to do so. Our first test was Oblivion and we fired it up at 1920x1200 4aa, 16af and max detail (with HDR). This should have been easily playable for two GTX's in SLI because one card handles it reasonably well. Well in theory … The game was doing something which looked very close to stuttering. Changing the game resolution or detail didn’t fix this and so it was off to the net again to look into this. On Evga's forum we found the solution, some PS/2 mice will have "judders" when playing games. Our review system does use a PS/2 mouse, purely to free up USB ports and so we changed to a USB model and the stutter/judder was gone and back in to the world of Oblivion we went. Again though, all was not well ... when playing Oblivion it was clear that there were other random issues, as when moving forward via the keyboard, random event calls were being made within the game, randomly calling up spell screens etc. This was just about the oddest bug we have experienced when testing a component. We rebooted, and following this, retried Oblivion with the same result. Next we tried Need for Speed Carbon and again after a short while (less than a minute) we ended up with a beeping system and the game began to receive random key presses which made it unplayable.

So back off to the net again we discover that some PS/2 keyboards don’t like to work with the board. The keyboard we were using was a standard Logitech model, and so we changed to a standard Microsoft PS/2 keyboard, only to be greeted by the same issue. Installing a USB Logitech keyboard cured the problem. Hardly what we would class as an ideal cure however as the only way to get this board to work was to sacrifice two USB ports to use a keyboard and a mouse. USB hub anyone?

We decided to take a look at the overclocking potential of the Evga 680i as that is meant to be one of its strengths. Our quad core Qx6700 was able to reach 360 MHz on the Intel D975XBX2 without issue at stock multiplier and so we started at 350 MHz on the 680i. The result, no post. After a few resets we were back in safe mode and tried at 340, no luck. 330, nope, 320, nein, 310 non, 300.... post! Really our joy at having a post was cancelled out by the fact that the board just wouldn’t pass 300 with a quad core CPU installed and this was regardless of the memory settings or voltages used.... especially appalling as we were using Vapochill! Clearly the board has an issue with quad core and FSB overclocking and so we decided to try the multiplier overclock method. Results from this were much better and we eventually settled for 15x and a CPU speed of 4 GHz. We could boot at 4.2Ghz (16x) and the system was 99% stable however there were a couple of random reboots so we backed off a little. In the process of multiplier overclocking we also found another bug which is that the safe mode option only works for FSB overclocking, if you get Multiplier overclocking wrong it’s a case of resetting the CMOS and trying again. More on that later...

We'll be honest here and say that all the little quirks of the board were really getting annoying and so we arranged a sample of a different board, the Asus P5N32-E SLI which also uses the 680i chipset and is similar to their Striker board.

Now begins the Asus memory nightmare.

When you receive a board which has a sticker on the front of the box reading “due to the 680i chipset ASUS only guarantee DIMMs on the qualified vendor list” you begin to worry. We feel that when buying a performance motherboard, or even a basic one, as long as we choose to use quality memory we should be able to do so without issue on any motherboard. And barring a few exceptions where particular boards don’t like one single memory product, this has always been possible. Not the case with the Asus though.

In our short time with the board we tried various specification memory in the 800 MHz, 1000 MHz and 1066 MHz ranges. The brands used were Crucial, Corsair, Geil and OCZ. None would allow us to successfully boot with two sticks installed regardless of the voltages, slots or timings used. Basically the board would not function even at a basic level, we noted that there was a bios available on the Asus site which enhanced memory compatibility however our board had shipped with that and so in disgust we removed the board, packed it up and went back to the Evga branded version.

By this point the P23 bios was out for the Evga board and we were quick to try it, unfortunately it fixed none of the issues we were still experiencing. The PS/2 keyboard and mouse were still unusable. The CPU wouldn’t FSB overclock past 300Mhz and other assorted bugs were still apparent such as no display on post. The flash also introduced reduced memory compatibility with some sticks resulting in no posts (when it would before). So much for progress. In order to use our preferred sticks (OCZ or Supertalent) after clearing the CMOS we had to install lower specification memory modules , boot, then set the timings for our OCZ or Supertalent sticks … Save the settings, switch the memory and then boot. This occurs also with the P24 bios released just last week, which doesn't fix any of the quad core or keyboard and mouse issues...or anything else that other users are experiencing from what we can see.

Whilst we are on the topic of bios flashes and CMOS resets we should also mention that some aspects of the board are really not well thought out at all. There is no avoiding the fact that this board is designed for enthusiasts, and is marketed as one which will outperform competitor’s solutions in the overclocking arena. Enthusiasts who purchase the board will more than likely have Quad core CPUs or SLI'd cards, some may very well have both. To create a board where the CMOS reset jumper and battery are directly below the second Graphics card is the work of a complete moron. If a SLI user gets into a situation where they have to reset the CMOS (and for Quad Core users, this can prove to be a regular occurrence during overclocking procedures) the end user has to dismantle their system to clear the CMOS because the second GFX card blocks all access to it. Doh!

The lack of thought doesn’t stop there however because the people behind the writing of the manual can’t even note down the various ports without getting them wrong. The Sata ports for example are labelled incorrectly in the board’s documentation and whilst this isn't a big issue, it’s another thing which degrades the overall user experience ... if by this stage that is possible.

We're on our fourth page of listing issues within MS Word so anyone still reading, firstly, well done. We could go on about some other issues such as the board losing its NICs on some boots or random post codes which don’t really mean anything, or the fact that Nvidia’s own Ntune software seems incompatible with this product but we won’t, instead lets summarise our feelings on the 680I boards we have tested (this applies to reference design boards only).

It’s pretty clear to us that the 680i was not ready for retail when it was launched and that end users are basically beta testing the boards for Nvidia. We also note, that all of our issues detailed above (other than the original broken slot) have been experienced by end users and documented on various manufacturers support forums. We can only assume the product was launched to tie in with the 8800 launch and that someone in Nvidia decided the issues which plague the board could be fixed “live” and were not serious enough to hold back the product for. We disagree vehemently.

We can categorically class this as the worst chipset product we have ever used and the decision to launch it in the original state is a complete insult to the high end enthusiasts who will end up with these in their systems. We appreciate that these are strong words however they are needed, especially when you consider we haven't even touched on the Sata issues which were incredibly serious but documented ad nauseam elsewhere or the memory incompatibility issues many people are experiencing which were worse than ours.

So where do we stand today? Well with the latest bios, P24 there are still too many niggling issues (keyboard and mouse etc) and major issues (Quad Core Support) for us to recommend it to anyone as a potential purchase. The Quad core support is a particularly worrying aspect because future Quad Core CPU's are going to be released using 1333FSB (333 MHz) and with the current revision of the board we cannot guarantee that any board purchased will work with those CPU's. This could be fixed by a bios update; however that's a big chance to take.

We've completed our tests with the board, working round the issues however upon completing the tests we were ecstatic to remove the board and move back to the complete stability of other products.

For those interested, outstanding issues can be tracked in this forum thread on eVGA's website and quad core overclocking issues/fixes can be tracked here