Options
VIA KT880 Review
Tom's Hardware: [link=http://www.tomshardware.com/motherboard/20040308/index.html]VIA KT880 Dual Channel Chipset Review[/link]
[blockquote]Based on the benchmark results, only a user with an incredibly discerning eye for graphics could pick a clear winner. Otherwise, it's a true toss-up between NVIDIA's nForce2 Ultra 400 and the new KT880 from VIA.
At the end of the day, however, we tend to give the nod to the VIA chipset compared to the nForce2 Ultra 400 from NVIDIA, based on the number of extras VIA offers. These include a Gigabit-capable network interface and two Serial ATA ports integrated in the Southbridge, which NVIDIA only offers as extras. [/blockquote]
Tom's Hardware: [link=http://www.tomshardware.com/motherboard/20040308/index.html]Article[/link]
[blockquote]Based on the benchmark results, only a user with an incredibly discerning eye for graphics could pick a clear winner. Otherwise, it's a true toss-up between NVIDIA's nForce2 Ultra 400 and the new KT880 from VIA.
At the end of the day, however, we tend to give the nod to the VIA chipset compared to the nForce2 Ultra 400 from NVIDIA, based on the number of extras VIA offers. These include a Gigabit-capable network interface and two Serial ATA ports integrated in the Southbridge, which NVIDIA only offers as extras. [/blockquote]
Tom's Hardware: [link=http://www.tomshardware.com/motherboard/20040308/index.html]Article[/link]
0
Comments
The Inq: Via KT880 for K7: dual channel chipset benched
Also Tom's Tests were with an AXP 3200+
Seriously tho, it's still a Via chipset. It's bound to have problems with certain PCI cards, Creative sound cards in particular (although it's not limited to that). I didn't read the article because I refuse to visit Tom's for any reason, but I assume it has no PCI/AGP lock, right? If so, I can't take it seriously from an overclocking standpoint. Is the onboard audio on par with the MCP-T's?
The PT880, KT880 and K8T800 Pro are all PCI/AGP Locked. Also with the Envy 24, VIA has one the best Audio controller available
Based on the benchmarks I've seen in this thread, it looks to me like the Via chipset is not significantly faster than the nForce2 in the best of cases. And it's STILL a Via chipset.
Usually for worse :o
I've used the KT-333 chipset as well as the KT-400 chipset and I never had a problem with the way either one performed.
Why the attitude Geeky? Did VIA steal your G/F? Maybe shoved you around on the schoolyard your freshman year of highschool?
Who's the best is cyclical, Remember when ATI was a Joke......
All it would take for Geeky1 to change his mind about VIA chipset is for Abit (or Epox) to make a KT880 chipset mobo that runs 250mhz+ stable (fingers crossed). Then his anti VIA stance will be long forgotten
I have one Via-based board that I actually use (or did, until it died...), and that is my MSI Pro266TD Master-LR. In this case, I had no choice as to what chipset to use though. The board is a dual socket 370 board that takes ddr ram. There are very, very few Dual PIII Tualatin boards with DDR support (three, in fact... the MSI, a Supermicro, and an Iwill), and they all use Via's Pro266 chipset, since it's the only P3/DDR chipset ever made.
It offers no performance advantage to speak of over regular SDRAM, but DDR is a hell of a lot cheaper than SDRAM is now, and that was my primary reason for wanting a board with DDR support.
If Intel had made a DDR P3 chipset though, I'd be using that instead.
Omega, it'd take a hell of a lot more than that, actually. Besides, I've NEVER used anything other than ATi cards in my personal systems, starting with a Rage 128 in the late '90s.
Unfortunately, I can say quite categorically that the KT880 chipset does *not* feature a PCI/AGP lock.
I have received confirmation of this directly from VIA :-(
Chip.
Yes, it a stupid mistake on their part. But the thing that you guys have missed; the *crucial* thing that sets the KT880 apart from anything nVidia have on offer; the thing that ultimately will make me buy it - is that the KT880 features the VT8237 southbridge, which itself has integral sata disk support, including raid.
Why does this matter? Well nvidia's efforts so far depend on a PCI based disk controller. On the motherboard, sure, but PCI based nevertheless.
The 133MB/s of theoretical bandwidth through PCI (circa 110~115 in practice) wasn't a problem in the days of 5400 rpm ATA33 disks. But nowadays, with ATA133 and SATA150 and even 10,000 rpm disks and raid0 being common, its a *major* problem.
My 2 x Western Digital Raptor drives are performing nowhere near their best. Burst rates of up to 200MB/s are seen with the VIA Southbridge vs a miserable 110MB/s with nVidia. And the sustained reading and writing rates can be up in the 140's and 150's with VIA. Again vs 100-ish on nvidia.
So, will I put up with no PCI lock, perhaps slightly less memory bandwidth, questionable soundcard support etc.? In return for a big hike in disk performance? YUP!
Chip
Well check if you like. But its really not so hard to understand. Single sata disks have burst rates up to 150MB/s. And the PCI bus won't let more than about 110MB/s of disk i/o through. Take 2 sata disks in a raid configuration, where the burst rates will be much higher and you can already see its a problem.
So OK, burst rates only apply to data in the drives' cache (although 2 x 8MB is quite a lot of data!). But what about sustained reading and writing? Well my Raptors can do 72MB/s sustained. Times two. Doesn't fit in 110MB/s, does it.
Chip
http://storagereview.com/articles/200303/20030320WD360GD_2.html
2 of them in RAID 0 would be somewhere in the range of 100-120MB/s. PCI is 133MB/s. Even with overhead, I doubt switching to a faster interface would net you more than a few MB/s, if that.
//Edit
Posted without seeing your above post
I really need the internal transfer rate of the drive (which I can use to calculate the actual maximum sustained data transfer rate), which WD doesn't publish, even though Maxtor, Seagate, and Hitachi all do...
Well you can doubt all you like. (And apart from that, I have the larger 74GB Raptors that can do 72MB/s, much more than the smaller Raptors.) But anyway....
The *facts* are, you can't get more than about 110MB/s of throughput through a nVidia PCI-based disk controller. And you can get much more than that - like 125MB/s or more - through the VIA controller running 2 disk raid0. And maybe 150MB/s or more with more disks.
You wanna see some benchmarks?
Chip
OK, well I'll have to hunt around for a few minutes to find the review I am thinking of. But in the meantime, here's an ATTO (if you know ATTO) of 2 x Raptor 74GB's in Raid0:
There's no way that will fit through PCI, obviously!
Chip
EDIT: The review I was thinking of doesn't show what I had wanted (I must have remembered it incorrectly). It show the Intel ICH5 southbridge managing 140MB/s (with a 2 disk Raid0 setup) and the VT8237 only about 130MB/s at best. But there is something odd about this: For example, the SIL3114 does very badly - I would expect it to manage around 110MB/s reads, even through 32bit PCI - and its coming out at a miserable 90-something. Anyway, there are other sources showing the VT8237 doing much better than that - I will have to hunt around.
Having said that, even the Intel southbridge numbers prove my point in principle, i.e. 2 Raptors can do 140MB/s. And that won't fit through PCI.
I knew they were fast, but I didn't think they were THAT fast. I may have to consider using them instead of the SCSI RAID array I was thinking about...