MSI RS480M2-IL benchmarked: lethargic southbridge
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
MSI RS480M2-IL benchmarked: lethargic southbridge
Since I can see that this board is gaining popularity, I want to send you a warning about the performance. The magazine C't tested two mainboards with the ATI chipset.
MSI RS480M2-IL with on-board graphic, preproduction sample
TUL AX480-A-GF with out graphic
Both showed a real slow southbridge. Example benchmarks against
Asus A8n-SLI Deluxe:
Ethernet onboard RX/TX 32kbyte packets Mbyte/s
Asus 112/113 75/79
MSI 8/11/15/59
TUL 14/46/0/0
Ethernet 2nd onboard or Intel PCIe RX/TX 32kbyte packets Mbyte/s
Asus onboard 75/79
MSI Intel PCIe 15/59
TUL Intel 0/0 (no idea why 0)
SATA-RAID onboard RX/TX
Asus 111/99
MSI 73/18
TUL 61/18
SATA reading out of HDD buffer MByte/s
Asus 112
MSI 94
TUL 62
USB2 RX/TX MByte/s
Asus 35/32
MSI 16/7
TUL 14/14
Conclusion: Lame southbridge, bad PCIe performance. This chipset is not competitive against a Nvidia or VIA.
MSI RS480M2-IL with on-board graphic, preproduction sample
TUL AX480-A-GF with out graphic
Both showed a real slow southbridge. Example benchmarks against
Asus A8n-SLI Deluxe:
Ethernet onboard RX/TX 32kbyte packets Mbyte/s
Asus 112/113 75/79
MSI 8/11/15/59
TUL 14/46/0/0
Ethernet 2nd onboard or Intel PCIe RX/TX 32kbyte packets Mbyte/s
Asus onboard 75/79
MSI Intel PCIe 15/59
TUL Intel 0/0 (no idea why 0)
SATA-RAID onboard RX/TX
Asus 111/99
MSI 73/18
TUL 61/18
SATA reading out of HDD buffer MByte/s
Asus 112
MSI 94
TUL 62
USB2 RX/TX MByte/s
Asus 35/32
MSI 16/7
TUL 14/14
Conclusion: Lame southbridge, bad PCIe performance. This chipset is not competitive against a Nvidia or VIA.
The MSI mobo has been around for like six months and still no improvement. AFAIK, this is the actual performance of the XP400 southbridge, that's why the XP450 have showed up. The ATI Xpress northbridge can be used together with ULI southbridges as well. Both these SB's are better performing.
Just saw that Gigabyte have removed their ATI/ULI µATX mobo from their site, too bad. Komplett have the two Sapphire ATX models in stock, and ebuyer got two µATX and two ATX models, all Sapphire . Still don't know what southbridge Sapphire uses.
Just saw that Gigabyte have removed their ATI/ULI µATX mobo from their site, too bad. Komplett have the two Sapphire ATX models in stock, and ebuyer got two µATX and two ATX models, all Sapphire . Still don't know what southbridge Sapphire uses.
-
- Posts: 37
- Joined: Fri May 13, 2005 8:33 am
For those who haven't come across them yet, here are some links about the USB performance of the new ATI chipset:
http://www.techreport.com/reviews/2004q ... ex.x?pg=17
http://www.gamepc.com/labs/view_content ... so&page=13
I'm still hoping to incorporate socket 939 µATX into my summer silent-pc build. Unfortunely, decent USB transfer is important to my workflow, as I frequently transfer larges chunks of data (up to 2GB) to my PC; this rules out available ATI Xpress boards, all of which have the crippled XP400 southbridge. That leaves the nForce4, with which I have heat concerns given the demanding nature of my build (I'm going for small, quiet, and cool, a.k.a. having my cake and eating it too).
At this point I am waiting to see what the summer brings while preparing for the worst; Come August, I may downgrade to nForce3 µATX or its µATX brethren (if they are cooler) or design my rig around the relatively hot-running nForce4, which, at least via the foxconn boards, seems like an otherwise solid contender.
http://www.techreport.com/reviews/2004q ... ex.x?pg=17
http://www.gamepc.com/labs/view_content ... so&page=13
I'm still hoping to incorporate socket 939 µATX into my summer silent-pc build. Unfortunely, decent USB transfer is important to my workflow, as I frequently transfer larges chunks of data (up to 2GB) to my PC; this rules out available ATI Xpress boards, all of which have the crippled XP400 southbridge. That leaves the nForce4, with which I have heat concerns given the demanding nature of my build (I'm going for small, quiet, and cool, a.k.a. having my cake and eating it too).
At this point I am waiting to see what the summer brings while preparing for the worst; Come August, I may downgrade to nForce3 µATX or its µATX brethren (if they are cooler) or design my rig around the relatively hot-running nForce4, which, at least via the foxconn boards, seems like an otherwise solid contender.
I can test the board I have, I found the C'T website, it's pretty crap, and as jojo said it doesnt have the review.
Can anyone who has got the article thell me what was tested and under what conditions, I can then try to reproduce those tests, and see what happens.
I have only tried it with a USB flash drive, but they are'nt very quick, the LAN did'nt seem to fly, but that could just be because of the particular files, and the fragmentedness of the drives.!!!
But as mentioned, seems fine to me.
Andy
Can anyone who has got the article thell me what was tested and under what conditions, I can then try to reproduce those tests, and see what happens.
I have only tried it with a USB flash drive, but they are'nt very quick, the LAN did'nt seem to fly, but that could just be because of the particular files, and the fragmentedness of the drives.!!!
But as mentioned, seems fine to me.
Andy
I am currently testing this board, and comparing it to an MSI KT6 Delta which uses a VIA KT600 NB + VIA VT8237 SB.
I was shocked when I saw the results of the USB test, which were conducted using a Maxtor DM9 2MB 120GB drive in a Belkin USB 2 Caddy. I downloaded and installed the latest ATI SB drivers for XP, I checked the USB drivers before, and after, they were exactly the same, Microsoft USB drivers!!! Not ATI Drivers. Just to double check, I used a different USB 2 cable incase it was a fluke, it was'nt, I got the same result again.
I performed the tests using the free version of HDTach using the long test, and my results are.
KT6 Delta - Random Access 14.4ms, CPU Util 8%, Average Read 30.7MB/s, Burst Read 30.9MB/s.
RS480M2-IL - Random Access, 14.2ms, CPU Util 6%, Average Read 17.1MB/s, Burst Read 17.5MB/s.
PATA Hard drive performance seems to be on-par with the norm. Also using HDTach long test, with a ExcelStor JS680 80GB 2MB.
Random Access 13.5ms, CPU Util 3%, Average read 48.8MB/s, Burst Read 87.8MB/s.
Unfortunatley I don't have a SATA drive available to test, or any firewire tests.
I can test the LAN, but I don't have any software to test with, I am open to suggestions, I have a Gigabit LAN, with KT6 Delta's + DM9 and DM10 drives for shunting data around.
In the meantime I am going to investigate a little further, and see if I can dig anything up.
Andy
I was shocked when I saw the results of the USB test, which were conducted using a Maxtor DM9 2MB 120GB drive in a Belkin USB 2 Caddy. I downloaded and installed the latest ATI SB drivers for XP, I checked the USB drivers before, and after, they were exactly the same, Microsoft USB drivers!!! Not ATI Drivers. Just to double check, I used a different USB 2 cable incase it was a fluke, it was'nt, I got the same result again.
I performed the tests using the free version of HDTach using the long test, and my results are.
KT6 Delta - Random Access 14.4ms, CPU Util 8%, Average Read 30.7MB/s, Burst Read 30.9MB/s.
RS480M2-IL - Random Access, 14.2ms, CPU Util 6%, Average Read 17.1MB/s, Burst Read 17.5MB/s.
PATA Hard drive performance seems to be on-par with the norm. Also using HDTach long test, with a ExcelStor JS680 80GB 2MB.
Random Access 13.5ms, CPU Util 3%, Average read 48.8MB/s, Burst Read 87.8MB/s.
Unfortunatley I don't have a SATA drive available to test, or any firewire tests.
I can test the LAN, but I don't have any software to test with, I am open to suggestions, I have a Gigabit LAN, with KT6 Delta's + DM9 and DM10 drives for shunting data around.
In the meantime I am going to investigate a little further, and see if I can dig anything up.
Andy
The C't uses netio and I am also aware of iperf.andyb wrote:I can test the LAN, but I don't have any software to test with, I am open to suggestions, I have a Gigabit LAN, with KT6 Delta's + DM9 and DM10 drives for shunting data around.
http://freshmeat.net/projects/netio/
http://dast.nlanr.net/Projects/Iperf/
Good news, Western Canada is getting the Gigabyte uATX RS480 mobo next week!
(well at least the sales contact at MemoryExpress.net informed me of that)
http://mirror.memoryexpress.net with locations in Edmonton and Calgary.
Hopefully the ULi southbridge USB is better than the MSI's ATI ass-end.
(well at least the sales contact at MemoryExpress.net informed me of that)
http://mirror.memoryexpress.net with locations in Edmonton and Calgary.
Hopefully the ULi southbridge USB is better than the MSI's ATI ass-end.
So it's still there, good! Haven't found any links for it lately...
This mobo looks good, only thing missing is video out.
All mobos with this kind of onboard graphics should have that!
This mobo looks good, only thing missing is video out.
All mobos with this kind of onboard graphics should have that!
waitaminit
doesn't the msi rs480 use the uli southbridge as well?
http://techreport.com/reviews/2005q2/a6 ... dex.x?pg=1
according to this, the uli has ass usb performance.
doesn't the msi rs480 use the uli southbridge as well?
http://techreport.com/reviews/2005q2/a6 ... dex.x?pg=1
according to this, the uli has ass usb performance.
correction: msi uses sb400 ati southbridge which apparently is ass. the tech-report roundup uses an ati xpress200 chipset with uli southbridge, which is AS ass as the sb400 :/
xpress200 southbridge usb benches here:
http://techreport.com/reviews/2004q4/ra ... ex.x?pg=17
correction: uli's southbridge isn't TOO shabby compared to original sb400 ati southbridge but cpu utilization is still over double that of nforce :/
xpress200 southbridge usb benches here:
http://techreport.com/reviews/2004q4/ra ... ex.x?pg=17
correction: uli's southbridge isn't TOO shabby compared to original sb400 ati southbridge but cpu utilization is still over double that of nforce :/
I have never used (or even considered) an Ali chipset, I'm still not going to now they have changed a letter, the ATI's USB performance is bad enough.
I will pass on the results of my LAN performance tests, (not scientific), I cant be bothered with netio, or Iperf, so I have made my up my own. Basically coppying a 4ish gig folder with 2000 files in it, and then test it VS the trusty old KT6 Delta, the only difference is THIS KT6 Delta has 1GB of RAM and a 3200+, I will also compare it whilst using Gigabit, and if I get time, a 2GB or so Zip file to test streaming performance vs small file performance.
On the upside, my tests should tell us all the real world implications. On the downside, I cant be bothered to test CPU usage.
Andy
I will pass on the results of my LAN performance tests, (not scientific), I cant be bothered with netio, or Iperf, so I have made my up my own. Basically coppying a 4ish gig folder with 2000 files in it, and then test it VS the trusty old KT6 Delta, the only difference is THIS KT6 Delta has 1GB of RAM and a 3200+, I will also compare it whilst using Gigabit, and if I get time, a 2GB or so Zip file to test streaming performance vs small file performance.
On the upside, my tests should tell us all the real world implications. On the downside, I cant be bothered to test CPU usage.
Andy
For network bandwidth tests I use a program that I think is from NetIQ (at work now, can verify later). I think the download is free if you fill out their form:
http://www.ixiacom.com/support/chariot/ ... netiqsrc=1
If you cant get it, I will try to email it if requested...if i can find it.
You would need three systems on the network to accurately test. One as the endpoint, and your two test systems.
http://www.ixiacom.com/support/chariot/ ... netiqsrc=1
If you cant get it, I will try to email it if requested...if i can find it.
You would need three systems on the network to accurately test. One as the endpoint, and your two test systems.
I have finished my "home made" LAN test, as mentioned before, not scientific, but should give a fair indication of "real life" usage for most people.
Ok here are the tests.
The server, KT6 Delta, XP3000+, 512MB RAM, SATA 200GB DM10 8MB, W2K3 Server.
The stalwart, KT6 Delta, XP3200+, 1024MB RAM, PATA DM9 2MB, XP SP2.
The Contender, RS480M2-IL, A64 3000+, 448MB RAM (512MB - 64MB Shared), PATA Excelstor J680 80GB 2MB, XP SP2.
The two tests, test 1 = 2004 Files a total of 4,768,844,594 bytes. Test 2 = 1 Zip file a total of 4,046,134,444 bytes.
The results, KT6 Delta @ Gigabit, Test 1 = 4:39 (279 seconds), Test 2 = 2:41 (201 seconds).
KT6 Delta @ 100Mb/s, Test 1 = 8:22 (502 seconds), Test 2 = 6:48 (408 seconds).
RS480M2-IL @ 100Mb/s, Test 1 = 8:42 (522 seconds), Test 2 = 6:35 (395 seconds).
As already mentioned this test is not scientific, the machines are only vaguely similar in specification. As Such the conclusions can be drawn in many ways.
Personally, I think the LAN performance of this motherboard, the MSI RS480M2-IL is good enough for most people, and most uses, and no-one should be complaining.
Unfortunatley I dont have a firewire device to test on this board, it seems that the Southbridge is split down the middle, PATA, SATA and LAN performance is fine, Firewire performance is unknown, USB performance is really lame, but will only affect people using external hard drives.
All in all, considering the benefits, and ignoring the bad points this board is still on my list of good boards, it wont be for everyone, but neither is the nForce solution, many people dont need blazing graphics, and could do with a bit more cash in their pocket.
Andy
Ok here are the tests.
The server, KT6 Delta, XP3000+, 512MB RAM, SATA 200GB DM10 8MB, W2K3 Server.
The stalwart, KT6 Delta, XP3200+, 1024MB RAM, PATA DM9 2MB, XP SP2.
The Contender, RS480M2-IL, A64 3000+, 448MB RAM (512MB - 64MB Shared), PATA Excelstor J680 80GB 2MB, XP SP2.
The two tests, test 1 = 2004 Files a total of 4,768,844,594 bytes. Test 2 = 1 Zip file a total of 4,046,134,444 bytes.
The results, KT6 Delta @ Gigabit, Test 1 = 4:39 (279 seconds), Test 2 = 2:41 (201 seconds).
KT6 Delta @ 100Mb/s, Test 1 = 8:22 (502 seconds), Test 2 = 6:48 (408 seconds).
RS480M2-IL @ 100Mb/s, Test 1 = 8:42 (522 seconds), Test 2 = 6:35 (395 seconds).
As already mentioned this test is not scientific, the machines are only vaguely similar in specification. As Such the conclusions can be drawn in many ways.
Personally, I think the LAN performance of this motherboard, the MSI RS480M2-IL is good enough for most people, and most uses, and no-one should be complaining.
Unfortunatley I dont have a firewire device to test on this board, it seems that the Southbridge is split down the middle, PATA, SATA and LAN performance is fine, Firewire performance is unknown, USB performance is really lame, but will only affect people using external hard drives.
All in all, considering the benefits, and ignoring the bad points this board is still on my list of good boards, it wont be for everyone, but neither is the nForce solution, many people dont need blazing graphics, and could do with a bit more cash in their pocket.
Andy
Having just built 2 systems using the RS480M2-IL board I agree with Andyb's results and opinions. Both systems have 1Gb of Kingmax DDR 400 dual channel ram, Maxtor 10 Sata 200Gb HDD and LiteOn DVD writers.
iTunes will crash when connected to an iPod via USB, but the firewire to iPod works very quicklyand flawlessly. Other USB devices (Canon printers, scanner, downloads from Nikon D70 and D50, mouse etc) work without a hitch.
A lot of reviews of this board lament the lack of gigabit ethernet, but I don't think gigabit ethernet will speed up a standalone pc connected to the internet via an ADSL modem and a 512k line.
The board satisfied my requirements spec for both HTPC (Antec Overture2 case and X-Card) and desktop (Antec Sonata case). The desktop is used for Photoshop and limited Word & Excel and the internal graphics is fine.
Conclusion - cheap way to get a quiet AMD 64 3000+ system up that supports all the current peripherals.
iTunes will crash when connected to an iPod via USB, but the firewire to iPod works very quicklyand flawlessly. Other USB devices (Canon printers, scanner, downloads from Nikon D70 and D50, mouse etc) work without a hitch.
A lot of reviews of this board lament the lack of gigabit ethernet, but I don't think gigabit ethernet will speed up a standalone pc connected to the internet via an ADSL modem and a 512k line.
The board satisfied my requirements spec for both HTPC (Antec Overture2 case and X-Card) and desktop (Antec Sonata case). The desktop is used for Photoshop and limited Word & Excel and the internal graphics is fine.
Conclusion - cheap way to get a quiet AMD 64 3000+ system up that supports all the current peripherals.