Via KT-600

[KRSESQ]

FWIW here is Via's info on the KT600 chipset. Doesn't say anything about PCI bandwidth, so more independent testing is going to be needed. If some brave soul wants to try it out with an RTx, be advised it still relies on the Hyperion 4 in 1 drivers. :P

http://www.via.com.tw/en/k-series/kt600.jsp

Kevin

[funky-d-munky]

I have been reading up trying to answer my own question. I have found out that the Southbridge includes SATA raid. This takes these HD's right out of the PCI bandwidth equation on this board. Considering the PCI bus has a maximum bandwidth of 133mb/s removing the SATA off onto the Southbridge has to be a huge advantage. The RAID controller now has the full bandwidth between North & Southbridge which is 533 MB/S for the KT600.
Reviews of the RAID have given it high praise with benchmarks showing allmost full usage of the SATA 150mb/s maximum. As well the IDE chanells are showing benches that almost max out the 133mb/s maximum for those devices.

http://www.hardocp.com/article.html?art=NTA4LDQ=

I will keep on investigating but this board is starting to shape-up as something to consider. Especially at the price.

[Nowhere]

funky, the thing that the hdd controller doesn't use pci now doesn't mean much...the same was true earlier after all, although only with PATA 133mb/s (and assuming that some KT333 board for example didn't have SATA chip), but so what - this didn't use pci neither, but probems with pci were present.
And you said earlier that perhaps this have changed with this new generation...IMHO it's still practically the same generation.

[funky-d-munky]

"...the same was true earlier after all, although only with PATA 133mb/s (and assuming that some KT333 board for example didn't have SATA chip), but so what - this didn't use pci neither, but probems with pci were present."

I am having a hard time trying to understand what you are saying. Can you explain further? PATA has always used PCI bus.

[Nowhere]

AFAIK, no. Unless VIAs (and others) announcements about advantages of using dedicated bus to connect southbridge to northbridge instead of pci were in best case not telling whole truth or, in worst, lying. And to me they've stated that...in this (this diagram is everywhere...so one can assume that it's from via itself), or this (in case of which we can be almost certain that it's from via) or this (you gonna argue even with block diagram from instruction?).
Google image search didn't find a single one that looks like diagram of kt133, where southbridge (and thus pata controller) is clearly connected via pci.
On every one of diagrams I've linked to pata controller is visualised as NOT being connected to pci bus...of course there's still possibilty about which I've mentioned at the beginning (and my conclusions come mainly from those diagrams etc. sources, so I don't consider them 100% certain), so perhaps there's here somebody who has the certainity about the subject?

[Nuno]

Assuming they didn´t fix the PCI controller issue from KT400 to KT600 (why would they, it´s in there since KT133 or so) here are some numbers:



Many people won´t like to ear this, but if your´re using PCI cards that really tax the bus, SiS is the only option on the AMD platform. Their PCI controllers rival on performance and compliance with Intel chipsets only. But that doesn´t give you the extra 5% performance on quake3

[funky-d-munky]

All IDE drives use PCI bandwidth. A few new boards Intel 875P / ICH5R (awsome but expensive) and the Kt-600 (cheap) have put SATA on the southbridge. This gives them more bandwidth but more importantly they will not use up valuable PCI bandwidth. As you can see from your chart Nuno, if your RAID array didn't take up PCI bandwidth the KT-600 would by far have the most bandwidth left over. Second only to Intel.
funky

[Elie]

Puting the SATA on the southbridge really doesn't mean anything dude!

The PCI performance with or without the IDE still remains low for cards like the RT.X100/RT2000 Storm etc.

You can buy the board if you really like VIA for some reason or another, but if your intent is video editing be warned.

And I'm saying this from experience believe me.

Cheers,
Elie

[Nuno]

Precisely. Aren´t we talking about a high-bandwidth video editing card here? Even if IDE is not taking any PCI bandwith (I think all chipsets now have IDE out of PCI, it has to be, as ATA133 actually equals the full PCI bandwidth), VIA PCI controller still is too flaky to deal with a Matrox RT.X, IMHO.

That test was done with a PROMISE Ultra 133 TX2 PCI card, in order to test PCI performance, not disk troughoutput. It´s actually a test that´s rarely done on motherboards review, I wonder why.

Here´s the link to the full review if you can read german (I can´t ) http://www.k7jo.de/reviews/mainboard...7s8x/index.htm

[Nuno]

Oh, and this is maybe also worth reading

http://www.tecchannel.de/hardware/817/0.html

[funky-d-munky]

What do you mean Elie?
Say both systems have a system drive, and a raid0, and a RT.X100.
In that equation the KT-600 board could minus the raid0 out of the equation. That leaves plenty for the RT.X100. Based on the graph it would have far more PCI bandwidth leftover than the SiS board. Of course the I875 has the best of both worlds.

Nuno, very old article there using a KT266 for heavensake. They have made many improvements. But all the more reason to run your raid off the southbridge regardless of your chipset.

[Elie]

funky-d-munky,

OK your analogy is correct however the amount of bandwidth required by the RT.X100 itself is so high that the RT series of cards are starving for more bandwidth regardless of where the IDE's are.

I used to have the MSI KT333 motherboard and although rated very well and it was stable, everytime I playback a simple transition like a cross dissolve I would get a black flash and stuttering during the transition.
As soon as I switched to the Asus A7S333 all was fine, now I have the following systems...

Asus A7M-266-D with dual AMD 2000+ and 1GB ram running RT.X100 (see sig) without any issues.

Second system Asus P4PE running RT2500 again no issues at all.

My SIS based system is running nicely with my RT2000.

Now like I said why don't you try it out first, if it causes you some problems, then return it or sell it on eBay and buy yourself something else not based on VIA.

Regards,
Elie

[funky-d-munky]

Yes I understand what your saying. But when the SiS boards only have 15% more PCI bandwidth over a KT-600 and then you remove the hole RAID (thats a massive, massive chunk of PCI bandwidth) that has to at the very least equalize and probably over take that 15% deficency. Wouldn't you say?
Again all of this is guesswork as reviewers can't seem to do much other than Sandra and Quake 3. Useless.
funky

[Dr Mordrid]

15-20% is FAR MORE than enough to make the difference between an editing system falling on its arse and running smoothly, especially if a RAID is involved, and when dealing with VIA's limitations you're far more likely to hit the wall.

Dr. Mordrid

[funky-d-munky]

"especially if a RAID is involved"

Well thats my point!
The RAID would NOT be involved on the KT-600 as it would be on the southbridge, not taking up any PCI bus bandwidth.
This would easily erase that 15-20% SiS advantage, and then some.

Of corse a i875p with RAID on suthbridge owens all.