My first and only DFI board: the Lanparty Jr. GF9400 T2RS. I scored this for ~$65 shipped, which is not cheap for LGA775 overall, but appears to be very inexpensive for a DFI Lanparty board.
@Mr.Scott helped me find an appropriate BIOS for this board (big thanks
nearly impossible otherwise... ), but in the end I didn't update it, as the one it came with appears to be slightly newer.
That thermal pad was mighty stubborn, luckily the paste was compliant. I used a screwdriver and other implements to no avail; some pad residue still remains. It's as good as I can get it for now.
As you may be able to glean from the photos, the cooling assembly is a little odd. The black Lanparty element is not a part of the nickel-plated copper (+Al ?) heatsink, just bolted on. It made thermal contact via thermal paste originally, now by a large 0.5mm APT2560 thermal pad. This seems inefficient to me, but for a relatively low heat output I don't think it matters too much.
MCH gets very hot without airflow, eventually reaching 75⁰c @ 15min idle. Luckily there are 50mm mounting holes on the black segment, if needed. During testing, I just pointed an Arctic P8 PWM PST CO at this area instead. Both the VRM and MCH area received 1.0mm APT2560 pads. I had to be very careful not to warp the board by cranking down those nuts and screws, I think I found the correct balance eventually. If anything,
Bumpgate would be what kills this board
MCH is dated wk44/2008.
Haven't heard of Vitesse before, pretty interesting to see.
"Linne_out" in the top right corner, never heard of that either.
This appears to be an exceedingly uncommon board, and I can guess why: SFF (for the day) gaming motherboard, $140 new, nVidia chipset (specifically, their
last for Intel desktop CPUs), solid OC features such as a post code and onboard pwr/rst buttons. Thus I only really have
user experiences from Newegg, as well as a
review from TweakTown to go off of. The consensus is that everyone had extreme difficulty doing either of the following:
•stabilize DDR2-1066
•boot >1600FSB reliably
And, well, I'm not special
1700MHz FSB is completely out of the question. The absolute highest I could get the RAM was 933MHz, and the highest I could get the FSB was 1680MHz QDR.
CPU-Z Validation
RAM overclocking was really messed up. HWiNFO reads the exact settings that I input in BIOS, but CPU-Z reads something completely different, which I may not at all change. I'm wondering if it's the motherboard being odd and the chipset not having 100% support. I'll talk about this more at the end, but basically I just don't think nVidia cared about chipsets anymore, so BIOS+Windows interactions are just plain wonky. For instance:
Considering how the below settings look (yes, this is the final validated OC), something's just not right with this BIOS. Those subtimings are concerning
clearly these are imaginary only.
For voltages, I assumed that "Core Aux +1.2v Dual" = vMCH, but I cannot verify this.
So this begs the question: why does this chipset suck at overclocking? DFI clearly put effort into it (vDroop < = 25mv for vCore), all power delivery and hot components are cooled adequately. Well, as I said before, I really don't think nVidia wanted to make chipsets or put a single dollar of R&D into them.
Last chipsets for AMD:
•Chipsets GF8100, 8200, 8300 = MCP78.
- MCP78 = 710a / 720a / 720d. Only change for GF8_00 is mGPU, per SKU.
•Chipset 980a
- Rebranded 780a, just with nVidia pushing for more DDR3 motherboards.
Last chipsets for Intel (desktop):
•Chipsets GF9300,
9400 = MCP7A
- MCP7A = 730i / 740i / 760i. Again, GPU is the main change. Also, MCP7A-ION was used for GF9400M ION.
So, excluding GeForce 320m (MCP89 one-off, ca. 2010), nVidia didn't put much further effort into new chipsets for the last ~2 years of producing them. I would assume they really gave up once
their legal skirmish with Intel started to look bad.
DFI and I both tried our best with this chipset, it just seems nVidia wasn't willing to fully bake it.
EDIT:
Here is a repository of info in case anyone wants to read more about nVidia Bumpgate.
May complete this post at a later date with more details.
Basically issue is a combination of thermal dissipation of the chip with hotspot, generating mechanical stress due to thermal expansion, wrong underfill material and wrong bump material.
It's not the solder balls between the chip and the board.
Chips affected are mostly from 2006-2008.
MCP and GPUs are affected, both in laptops and desktops.
Some articles list G86, G86A2, G84, C51, G72, G72M, G73, G72A3, MCP67 and NV42. But these aren't the only ones, some G9x and other MCPs were affected too.
That's some later GeForce 6000, almost all GeForce 7000, almost all GeForce 8000, some early GeForce 9000. Affected MCPs are MCP5x (northbridge only, not southbridge), MCP6x and early MCP7x.
NVidia started producing fixed revision for some affected chips in the summer of 2008. Some defective ones were still produced a bit after that.
Fixed revision of affected chips got a new underfill material that's white rather than dark, so it should be easy to recognize them. They'll typically have a datecode after 0830, and some of them have an odd digit instead of an even digit at the end of the model number (eg. G84-603-A2 instead of G84-602-A2).
Note that some chips were still using the bad bump material with the good underfill material (early 9600M) but they're much less prone to failure than earlier series.
Fixed chips include most G84/G86, most of them are still available at see-ic.
According to this information, it appears this motherboard is unaffected, however my XFX & EVGA 750i boards are affected.
