We all knew VIA has had a bad rap, but Damn...

[MultimediaMan]

It's seems that the Industry as a whole is turning it's back on VIA...NVIDIA are you listening? Read all about it here.

[Jazzz]

their products ain't crap but lets just say they could be whole lot better

[Dr Mordrid]

And that's exactly what a lot of board makers are thinking: why go with VIA and their troublesome quirks when you can go with SiS's stable chipsets as an Intel alternative?

Sounds like that logic is finally sinking in.....

Dr. Mordrid

[lecter]

I agree, in my opinion if money is no object, a P4 goes with an Intel chipset and an AMD goes with an Nforce...for budget users, it seems that SiS rulz.

[Dr Mordrid]

I'm not convinced of nForce yet. IF someone runs an RT.X100 and a big, fat RAID on one without it chocking on realtime exports to the HDD then I might reconsider.

Dr. Mordrid

[Ryu Connor]

TR collected some results while tracking down a bug in the PCI implemenation of the MSI K7N420 Pro.

http://www.tech-report.com/news_reply.x/3675/0/

Post #120 gives the cumulative results:

A max PCI read score of 116,624MB/s and a max PCI write 64,567MB/s have been observed/obtained.

Whether or not this would be enough bandwidth I can't say. I do know some people have been using Canopus DVs with success.

[Wombat]

I haven't been that impressed with nForce chips thus far. What makes it any better than the SiS?

[Dr Mordrid]

Especially with the new SiS 963 southbridges 1066 mb/s pipe vs. nForce's 800 mb/s.

Dr. Mordrid

[Ryu Connor]

http://www.tech-report.com/reviews/2002q1/via-kt333/index.x?pg=1

It's performance is a little stronger. 2-3% slower than the KT266A and 3-6% slower than the KT333 on average. There are some results where nForce comes out on top, but the margain of victory is particularly slim. In fact depening on who your perception the percentage values it is behind on the VIA boards wouldn't be all that awe inspiring.

The SiS chipsets on average are further behind than this. In many cases the AMD760 is even a stronger performer than the SiS chipsets.

Of course performance isn't everything and the impression I get from this forum is that compatiblity and stability is of more value than performance any ways.

I could tell you it is perhaps the most stable platform I've had the pleasure to deal with. I could certainly find other to say the same thing, hell four of my own close friends made the shift after having seen my success. But, eh, one trip is worth more than a thousand travel books.

Beyond performance and stability is it's support of APIC for IRQs up to 22.

http://www.unspacy.com/ryu/irq.gif

Beyond that are the integrated capabilties. The NIC performance wise is your garden variety 10/100 model, but the driver does offer a choice between performance or CPU usage. That's a nice addition usually not seen except on higher dollar cards.

The audio is pretty damn hard to beat in terms of features, CPU usage, and sound.

http://www.anandtech.com/showdoc.html?i=1535&p=16

The fact all those integrated items on a 800MB/s hypertransport connection hooked to 4.2GB/s worth of bandwidth doesn't tend to hurt either.

It's drivers are pretty nice as well. Naturally there is an AGP GART, but beyond that it's just material to configure the NIC, Sound, integrated video, and Twinbank.

No IDE drivers, no USB drivers, no weird patches for 4 way interleave and PCI latency. It just uses your common variety Standard PCI IDE and Standard OpenHCD USB controllers, which I suppose you can see from the IRQ listing I put up earlier.

I'm not going to snowball you though. It's not a perfect platform, there's no such thing. It's one of the more reasonable options available though.

[K6-III]

Yeah, but an NForce2 board made by Epox would really be enticing...

[Dr Mordrid]

And as noted elsewhere the software benchmarks mean nothing with it comes to PCI bandwidth. Put some high bandwidth cards like the RT.X100, SX4000 RAID5 etc. in there and see how it does, then we can talk about real world performance.

SiS's latest chips support APIC too.

Audio I couldn't care less about. The TBSC does that just fine.

Dr. M<ordrid

[Ryu Connor]

Yeah, but an NForce2 board made by Epox would really be enticing...

Actually. I find the coming Asus A7N8X infinitely more interesting.

http://www.tbreak.com/forums/showthread.php?postid=105654

I will be shifting over to it upon release.

[Ryu Connor]

And as noted elsewhere the software benchmarks mean nothing with it comes to PCI bandwidth. Put some high bandwidth cards like the RT.X100, SX4000 RAID5 etc. in there and see how it does, then we can talk about real world performance.

Those were RAID results. Adaptec 29160 with Cheetah X15 in RAID-0 at the top end and we got all sorts of other results along the way.

[Dr Mordrid]

I'm using a Fasttrak SX4000 RAID5 and am getting >80 mb/s writes and ~110mb/s reads using four Maxtor D740X 7,200 rpm IDE drives.

The question is can that RAID0 you're talking about do smooth reads/writes while a realtime editing board is pushing huge loads during a realtime export with a deep stack?

On my SiS 745 chipped system this is no problem at all.

Dr. Mordrid

[Ryu Connor]

In other words, will the PCI bus hold up under heavy arbitration.

That's a valid question for which I don't have a valid answer. Short of the one Canopus DV users I mentioned earlier I haven't seen the board hammered under such conditions. His rig works okay, but then again he's only using RAID-0 IDE.

I suspect we'll eventually get an answer to your question. Honestly I'd like to know myself. Since nForce2 has garnered more manufacturers it seems likely that Matrox will receive one to test. Then we'll perhaps know for sure one way or another.