MSI SLI Boards and 90nm Parts...

[Merovingian]

I remember a problem with MSI SLI board and 90nm chips, has this been resolved? If not, how reliable are the DFI boards?

 

[lostorbit]

MSI has just released BIOS ver. 3.4 for the SLI boards, I have the feeling it didn't help much, but I'm not sure.

I have both boards at home, a Winchester 3000+ and a Venice 3200+, and I can tell you from experience:

both boards are nice and have their pros and cons:

with the DFI, you can really feel that this board is there for you, incredible overclocking features, can save 4 different CMOS settings plus the last successful boot. A lot of tweaking and control features, Memtest+ built into the BIOS and all the voltage limits you can ever imagine. A lot of people had reported stability issues, especially with Corsair RAM, but I think it's all in the settings, the board is not for noobs to play with, but online support on the DFI forum has very informative info on settings and tweaks. Also, the forum seems to have a different crowd from the rest of the forums (more knowledgeable) with very helpful techs.
I used the board with two Corsair Valueselect sticks with no problems whatsoever with the latest BIOS, super stable at 260 FSB x 10 (RAM set to 166, it's a 'value').
What I did not like about the DFI was the lack of a serial port on the I/O panel (has a header on the board though), lack of a parallel port, only two PCI slots! (compensates with 4 PCI express slots, but that's still ahead of its time, with no PCIe cards out there but grapnhics cards), the MSI had 3 PCI and 2 PCIe for SLI functionality (no extra PCIe slots though).
You also have to feed the DFI with two extra power connectors, annoying (molexs like the ones you'd use with your drives), seemingly to drive the fans when power is scarce (unverified), maybe you can run stable without connecting them if you've got a good power supply, don't know, everyone was recommeding to 'feed the board' for extra stability.

The DFI board had numerous jumpers (especially to switch to dual graphics mode or SLI), the MSI was, with a couple of exceptions, virtually jumperless.

I use my PC as an HTPC, so noise and sound quality were important factors for me. The NForce chipset fan was slightly noisier than the MSI's when supplied with 5v (DFI's had a superior chipset fan though) and the onboard Realtek ALC850 sound solution, although implemented much better than with any other board with a similar chip, wasn't as good as the SBLive 24 on MSI (I know, I know, creative, much like MSI, suck big time with driver updates and issue resolutions seem to take forever, if ever resolved). the Zalman 7700 was too close to my 6600GT's heatpipe on the DFI (the fins were slightly touching), had me moving the card to the other (8x) PCIe slot to resolve.

I am now back to using the MSI with the Venice, because of the onboard sound and the I/O ports(didn't want to spend money on a standalone PCI audio solution and needed all three PCI slots for other cards, I still use the parallel interface for JP1 interface remote programming). It does not suffer from the problem of booting beyond 219 FSB like with the newer Winchesters (but I am using the unofficial beta 5.06). Also worth mentioning that the Venice does not overclock as well as with the DFI, but for me, running on lower temps and lower noise was more important than 'overclockability'.

MSI had released ver. 3.2 & 3.3 after a pretty good amount of time and had not resolved the issue (in fact most people reported more problems with the newer BIOSes, and most reverted to using ver. 3.1 or even 3.0). I did not get to try 3.4 yet but, I do remember someone having the same complaints as with the 3.3 (MSI's website does not mention the resolution of the FSB issue when detailing what's new with this release).
Fruthermore, even if MSI gets to solve the issue, in my opinion, it has taken it way, way too long to acknowledge and start working on the issue (at least 3 months now, maybe more, don't remember). MSI has lost all credibility in my eyes. If Anandtech and some other website had not put the pressure on them, I would've expected them to just ignore the issue. DFI on the other hand, from what I've read, was very quick to come out with a solution.
the MSI board mostly works fine if you're not overclocking, but why the hell would MSI boast about overclocking features and leaves the Winnie users biting the dust? just take a look at the forums, especially MSI's, people are angry and left helpless, and that's unacceptable. It's true that it's mainly not MSI's fault that AMD changed the design of it's Winnies, but come on guys, almost every other manufacturer had resolved the issue, that should tell you something about this company:
Either they are way too slow with their development/support (highly unlikely) or they have a serious design issue that prevented them from fixing the problem with the Neo4s, and of course, they're not going to acknowledge it nor offer a recall (total financial fiasco). In both cases, It's outrageous, and personally, I'm never buying MSI again. I think I'll be looking out for the next gen DFIs when switching to dual core, I really liked my experience with that board, hopefully they'll offer an even better bundle in the future.

I think if you're supplementing the board with a decent audio card and don't suffer from my same limitations (the I/O ports) and overclocking is a requirement, you should look no further than the DFI.

 

[js9600]

Very good description of MSI. But at least 1 Venice 3500 in this world also have booting problem, no matter bios version/settings. Considering how AMD64 works, ie. MSI bios have build in dynamic overclocking up to 11%, how other boards perform, and how their bios "development" seem to work there is no way you should buy this model. May be it will be fixed, may be not - there is not much confidence left. They are wasting peoples time and money. Their forums or other support I will not comment but deleting posts about this issue is part of it so you can figure out the rest.

Btw, with 3.4 my Venice boots up at 1000mhz, 1.1volt and forced 5xmultiplier regardless of bios settings. Another challenge though fixed if I use default HT frequency of 5. Since Im about to do some RAID, also a shutdown hanging bug there, I could end up not being able to either boot properly or shut down properly, heh.

 

[lostorbit]

What you are describing here, js9600, is the safe boot situation I believe. I've seen it implemented a few BIOS versions earlier (counting betas). My 5.06 Beta used to exhibit similar behaviour with the Winchester. I don't really remember the solution, but a fix to BIOS settings, a CMOS Clear, a power disconnect or the last two options combined cleared the situation.

By the way, with the BIOS I'm using, I used to be able to boot @ 239 Mhz FSB, 1x HTT with the Winchester. The system once in Windows, can be pushed easily to 300/320+ Mhz FSB (using clockgen), with no stability issues (I've heard of people pushing much higher), but 300 /320 is more than sufficient to achieve maximum overclock with the current CPU line offering.

With people using clockgen to counteract the FSB limit bug (or the FSB 219 bug), another issue was that the BIOS would reset the FSB to boot settings after waking up from S3 state, so if you were in the habit of putting your computer to sleep, you had to re-run clockgen after every waking cycle.
DFI on the other hand, showed erratic temperature and other vital readings, so I guess many nforce4 boards have flawed S3 implementations (although if you do not suffer from the FSB bug on the MSI, you should be able to resume with no apparent problems)

Dynamic overcloking is a waste of time (so is agressive timings and the hidden graphics card overclocking options), you've got to manually to achieve a stable setup.

My apologies, TheMerovingian, for going off-topic.

 

[js9600]

Yes something to remember while software oc´ing. There is allways Core Center or what MSI call their oc-program, may be it can handle problem. I like Clockgen but not best way of oc´ing.

Hmm, not convinced about safe boot but on the other hand not really sure what you mean, heh. I have moved back to 3.3 but sure I could repeat problem over and over - any other HTT than 5 gave forced multiplier and low volt. Change to 5 and everything was normal.

Well all mobos have some uselss gadgets and bios settings, if MSI throw in some bios-oc that is fine with me. At least it shows they know what oc is...

Aggressive timings are not bad, just very aggressive See some of the changes with A64Tweaker. Speaking of the tweaker there is still a difference between some values in bios and the ones reported by program - bug no. X perhaps. Think I read it was worse with first bios versions. Anyway the preset is proof MSI even knows what latency tweaking means. Naturally highly undocumented. Now they need to make it work on other than their lab-computers.