ATI Radeon Xpress 200 PCI performance

Introduction

There seems to be a problem with ATIs Radeon XPRESS 200 PCI performance. I test it here with an Intel PRO-1000 MT Desktop adapter Gigabit Ethernet performance.

It started when I discovered a very high CPU usage and low transfer speeds with a Gigabit Ethernet PCI adapter.

Test systems

Server
Workstation
Reference

Processor
Sempron 2800+ 1.6GHz
Athlon 64 3200+ 2.2GHz
Duron 1.6GHz

System bus
800MHz HyperTransport
800MHz HyperTransport
400MHz FSB

Motherboard
MSI RS482M-IL
MSI K8N Neo Platinum
MSI K7N2 Delta-L

BIOS revision
1.4
?
?

North bridge
Radeon Xpress 200
nForce3 250Gb
nForce2

South bridge
Radeon Xpress 200
N/A
nForce2

Chipset drivers
South Bridge 6.2
?
5.10

Memory
512MB cas 2.0
1GB cas 2.0 1T
512MB cas 2.0

On-board NIC
RealTek 100Mbit
nForce Gigabit
nForce 100Mbit

Intel PRO-1000 MT drivers
10.2
N/A
10.2

Hard drive
Seagate Barracuda 7200.8 400GB SATA
Western Digital Raptor 74GB SATA
IBM 80GB IDE

Graphics
Integrated Radeon Xpress 200, Catalyst 6.2
GeForce 6800GT 256MB AGP, ForceWare 81.98
GeForce 3 Ti 500, ForceWare 78.01

OS
Windows XP Professional, SP2, DirectX 9.0c

PCI Gigabit Ethernet Performance Test

Machine
Copy from
CPU usage

Server on-board RealTek 100Mbit NIC
9MB/s 33%

Server Intel PRO-1000 MT in PCI slot 3
17MB/s 90%

Server Intel PRO-1000 MT in PCI slot 1
26MB/s 15%

Reference on-board nForce 100Mbit NIC
10MB/s 13%

Reference Intel PRO-1000 MT in PCI slot 4
50MB/s 20%

References

A forum post that claims it is an issue with the ATI Radeon Xpress 200:
http://episteme.arstechnica.com/groupee/forums/a/tpc/f/77909774/m/883001067731

Yes. It's an old problem. I posted a while ago that I couldn't understand why the Ars System Guide recommended the Xpress 200 with this and other known problems. (USB)

Another post:
http://episteme.arstechnica.com/groupee/forums/a/tpc/f/77909774/m/813006824731

They did not plan to address the PCI bus throughput issues, a problem they said only affects GigE cards (mostly). I wouldn't put a RAID card on the PCI bus on that thing, though. (PCI-E seems OK.)

The Inquirer tests reports a USB performance problem:
http://www.theinquirer.net/?article=20039

What we were hearing is that chipset has USB issues, but technically meets the spec. There is a re-spin coming in Q1, but ATI declined to be more specific than that. While they would not say what is being reworked, the south bridge is the obvious target for their affections.

The Tech Report tests USB performance and CPU usage of a board with this chipset:
http://techreport.com/reviews/2004q4/radeon-xpress200/index.x?pg=17

Here we have a bit of a problem. The USB controller in the Radeon Xpress 200 south bridge is about half the speed of the Intel and VIA solutions, and it uses twice as much CPU time. This may not be a showstopper, but I wouldn't recommend using this chipset extensively with an external hard drive or the like.

Conclusion

ATI - Customer Care said: (10.02.2006)

It is not a limitation of the chipest.
It may be a motherboard issue.
I would recommend contacting MSI to see what they advise.

Is ATI Customer Care a bit short on knowledge, or do they simply not want customers to know about problems with their products? Maybe they are afraid of Class Action lawsuit.

I am returning this motherboard and probably request an Asus K8N-VM for replacement. I've troubleshooted this GigE PCI performance issue for weeks now... not very happy about it ATI!

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ width space workaround _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _