• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

VIA chipset mobo's - Poor disk performance - Is it just me?

M

Mac

Senior member
I have recently built three VIA KT133a Socket A PC's using both Abit and MSI mobo's. Even after installing the "4-in-1" drivers, the disk performance was sub par compared to an old Abit BX6.r2 mobo. In fact the newer PC's used the newer ATA-66 and ATA-100 7200 RPM HD's. The old Abit with ATA-33 7200 RPM HD simply hammered these newer mobos. On top of that, two machines used AMD T-Birds 1000/133 and the other a T-Bird 1200/100. The Abit used an oc'd Celeron 566/850.

Taking everything into account, the new computers should have left the old BX board in the dust. If it were raw compute power, yes they do. But when doing any type of large disk activity, the BX board ran anywhere from 20-40 % faster. The most recent and frustrating task was installing a 45GB hard drive and copying the files from an older smaller hard drive. It took forever.

Am I'm missing something or is this one of those issues people don't talk about?

 
What benches are you using to compare? My Kt7-Raid scored 42000 on Sandra in Raid 0 and 24000 with a single ATA66 hdd. My new KR7A-Raid scores 26500 with a single ATA133 hdd on IDE3. I am gathering a couple new ata133 hdd's to Raid soon (1 down, 1 to go)
 
and to show that this affects scsi as well as ide drives look here😉

to be more specific:

With the MSI motherboard the RAID-0 sequentially reads data with up to 114 MBytes/s. This performance is maintained over nearly the whole capacity of the RAID: the minimum speed only drops to 106 MBytes/s. Even with burst transfers from the cards cache we found the same results.

This looks much worse with the VIA motherboard. Our RAID there only delivers a maximum of 72 MBytes/s with sequential reads. Although the system also sticks with this speed over almost the whole capacity, a lot of the potential of the RAID is wasted. To be more precise: If you spend more than 2300 Dollars for a RAID system like this, you loose 34 per cent of the performance when this RAID has to deal with a VIA chipset.


hope this helps
 
I played RTCW, loading the levels is pretty heavy HD activity, but I'm always the first person there...
I'm more than happy with VIAs IDE performance.
 
The performance I am measuring is real world...not a synthetic bench mark. When I first suspected this, I set up large file transfers under different scenarios...

1. copy from folder to folder - same drive same partition
2. copy from folder to folder - same drive, different partition
3. copy from folder to folder - different drive - same IDE channel

Measured the actual elapsed time. In every instance, the old BX board (celeron, ATA 33 HD), wiped the newer and more powerful VIA/T-Bird/ATA 66 or 100 rigs.

What was even stranger, Win ME ran noticeably slower than a Win98SE machine.
 
You guys need to apply VIA's PCI performance patch ... visit www.viaarena.com to get it.
VIA tweaked the PCI behavior of the chipset to give more throughput, and until this change in
chipset programming filters through to the people who make BIOSes, we apply this little program.

This applies to any VIA chipset down to the MVP3 socket-7, at least.

regards, Peter
 
even without the patch, you'll notice sis and nvidia are not in the comparisions, why? Simply, they are slower then via, and the comparision is biased.
 
AA0,

...and the comparision is biased...

What are you referring to? The examples I posted are real world. When attempting to copy large folder/files (200MB+), the Via mobo's simply start to bog down.
 
You know, I had not noticed this until I had received my Asus P3C-L with the i820 + 128MB + CRIMM. I have also a Slot 1 Asus (propietory motherboard from HP)VIA Apollo Pro with 256MB RAM PC133. I only have 1 Pentium 3. So I swap the CPU when needed, until I can buy another CPU. I installed Windows 2000 with the same configuration( DVD, WD 20GB 7200RPM, NIC, ATI Radeon) and the INTEL chipset really went FASTER than the Apollo Pro. I mean the VIA chipset thats for ever. I know now, I dont really want a VIA chipset anymore, I'll get an AMD chipset if I want AMD CPU's
 
to the link of the IDE performance, SiS and nvidia both have lower performance than VIA, but they don't get any bad feedback from it.

Out of the 5 chipset makers, they are right in the middle. Stop making them look like they are the worst. I'm not including AMD here, but I've never seen anyone bench their IDE.
 
I'm not saying that Via is the worst, but I am saying that according to my experience, the onboard IDE performance is significantly slower, anywhere from 20-40% than a BX mobo that is over three years old with an underpowered CPU".. This is outrageous and nobody seems to talk about it. It becomes very noticeable when doing large file transfers. You can actually watch the status window halt regularly.

As mentioned previously, it becomes very apparent when upgrading multiGB hard drives. For example, I had to move about 25GB of files off of a primary drive in order to repartition it, and then transfer the files back to the resized partitions. The primary drive was a a 45GB 7200 RPM ATA-100 hard drive copying to 40 GB 5400 RPM ATA-66 and then back to the 45GB drive. Both drives were on the same channel. The entire operation took approximately 4.5 hours.

Unless there is something major wrong in my system setups (possible but I would be surprised), it is making me rethink my strategy of bringing up a new Via KT266A based system to do video editing. Video files are huge.

I am not an Intel bigot and have nothing against AMD or Via...in fact I have two brand new AMD Tbirds sitting on my shelf ready for installation. (It's winter time and I need the extra heat...sorry, just couldn't resist😉 ). But I am starting to wonder. Which brings me back to the original question...am I'm missing something or is this one of those issues people don't talk about?
 
No, it isn't just you.

All VIA-chipset based mainboards suffer from lowest-in-the-class (perhaps save for a particular ALi chipset 😛) IDE subsystem performance.
 
Does anyone know if the new KT333 boards will have the patch already or do they need applied? I still haven't bought a mobo/cpu yet and I'm still considering an AMD solution. I really want AMD, but the Via RAID and Sounblaster problems spooked me. I'm not comfortable with chipsets from anyone else either, with the exception of AMD. I will be running RAID so it's important...
 
Last I read it should have only been the 133 chipsets and earlyer and only those with spacific north and south bridges together (The last patch FAQ I got from Via claims this) Not Non-Via norths with Via souths or Visa Versa.

I have a Epox 8KTA3+Pro that has the formentioned problem but when I installed the patch my system ran about 10f deg hotter and my SCSI U160 drives ran at half speed and my LAN cards stop working completely so I uniinstalled it and normal again.

Something I noticed in the patch faqs is that they adjust the latency and Quing of the PCI slots but in sandra and other hardware info programs mine are correct and I think the patch screws things up if all is working right already.
The FAQ also claims its not just Via's fault since its MB MFG who change/use reserved registries in the chipsets that cause this so nobody takes blame.

Try the patch, If it helps you then ok, If not then uninstall it and look elswhere for the problem.

 


<< ..All VIA-chipset based mainboards suffer from lowest-in-the-class (perhaps save for a particular ALi chipset 😛) IDE subsystem performance. >>


Actually, ALi chipsets have quite good IDE performance, thanks to a very robust southbridge. It's the northbridge that has usually been criticized for its speed issues, although the C-revision of the ALiMAGiK 1 chipset seems to be keeping up pretty well with things on all fronts.
 
AA0, that's BS (as usual, I'm tempted to say).

SiS have the highest performing IDE unit out there - simply because it's directly connected to the chipset
northbridge (!), on its own internal bus. It has been that way ever since the 5571 socket-7 single chip from six years ago -
in fact, they already did that on their last 486 chipset, the SiS 496/497.

Just look at a couple of HDTach or other benchmark comparisons, and you'll see that the SiS IDE engine is the finest one out there.
Watch for UDMA-133 capable updated P4 and K7 chipsets from them soon.


MrBumpy, what has been holding the VIA IDE back (in the xx133 chipsets) isn't the IDE unit in the southbridge itself,
but rather the poorly chosen bus arbitration scheme on PCI - VIA had way too much fairness going on, hampering single device
throughput. (This is also why those hybrid chipsets AMD761 plus VIA 686B have good IDE performance.)
Their latest "patch" leans that more toward throughput (allowing longer data bursts). The xx266 chipsets don't
need that patch for their builtin IDE, but any PCI card will benefit from the improved PCI strategy.

This leaves us with the worst performing PCI bus out there - which is on Intel's 850 chipset. They admitted it (see their
"Specification Update" document), and said they won't do anything about it. AA0, here's your company to be picked on.

regards, Peter
 
SiS benchmarks show them to come under VIA with their 735 chip.

Just because the north and south bridge are close together, doesn't mean its definitely going to bring higher performance. The two bridges still exist, they are just in one chip.
 
... with a 1.2-GByte-per-second multithreaded bus, compared to VIA's 266-MByte/s half duplex V-Link.
And the SiS IDE isn't even on that bus, it has its own direct northbridge (!) connection.

Sure, overall application and gaming performance of KT266A might be superior to SiS 735
thanks to the excellent DDR RAM controller in the former, but IDE and PCI performance absolutely is not.

regards, Peter
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top