SiS755 Reference Board: Athlon64 from SiS
by Wesley Fink on November 23, 2003 11:19 PM EST- Posted in
- CPUs
SiS755 Reference Board: Stress Testing
We performed stress tests on the SiS755 Reference Board in these areas and configurations:1. Chipset and motherboard stress testing, conducted by running the FSB at 215MHz.
2. Memory stress testing, conducted by running RAM at 400MHz with 2 DIMM slots filled and at 400MHz with all 3 DIMM slots filled at the lowest memory timings possible.
Front Side Bus Stress Test Results:
As normally done in our testing of production motherboards, we ran a full range of stress tests and benchmarks on the SiS755 Reference Board to test stability at an overclocked speed. This included Prime95 torture tests, and the addition of other tasks — data compression, various DX8 and DX9 games, and apps like Word and Excel — while Prime95 was running in the background. Finally, we ran our benchmark suite, which includes ZD Winstone suite, Unreal Tournament 2003, SPECviewperf 7.0, and Gun Metal Benchmark 2. At default voltage, 215MHz was the highest overclock that we were able to achieve with the SiS755 at 800 HT while running these tests.Unlike our experiences with some of the VIA K8T800 overclock, the SiS was completely stable when it was able to run an overclock. The PCI/AGP lock, at the very least, appears to improve stability of the overclocked speed.
Memory Stress Test Results:
This memory stress test simply tests the ability of the SiS755 Reference Board to operate at its officially supported memory frequency (400MHz DDR) at the lowest supported memory timings that our Mushkin PC3500 Level 2 memory will support:| Stable DDR400 Timings — 2 DIMMs (2/3 DIMMs populated) |
|
| Clock Speed: | 200MHz |
| Timing Mode: | N/A |
| CAS Latency: | 2.0 |
| Bank Interleave: | N/A |
| RAS to CAS Delay: | 2T |
| RAS Precharge: | 6T |
| Precharge Delay: | 2T |
| Command Rate: | N/A |
We had no problem running 2 DS 512MB DIMMs of our standard Mushkin PC3500 Level2 in the SiS755 Reference Board. We were also able to run the memory test suite with complete stability at 2-2-2-6 timings. Tests with some demanding games other than our tests suite, however, required RAS-to-CAS to be slowed to 3 for most stable operation. We also tested the SiS755 with just one DIMM and found timing requirements for one or two DIMMs to be the same.
Filling all available memory banks is more strenuous on the memory subsystem than testing 2 DIMMs, and, unlike some experiences with other Athlon 64 chipsets, 3 DS DIMMs worked just fine on the SiS755. We did have to lower memory timings to 2-2-3-6 for stable operation with 3 DIMMs, but this is a very small decrease.
| Stable DDR400 Timings — 3 DIMMs (3/3 DIMMs populated) |
|
| Clock Speed: | 200MHz |
| Timing Mode: | N/A |
| CAS Latency: | 2.0 |
| Bank Interleave: | N/A |
| RAS to CAS Delay: | 3T |
| RAS Precharge: | 6T |
| Precharge Delay: | 2T |
| Command Rate: | N/A |
We tested the memory timings with both 2 and 3 banks filled using several stress tests and general applications to guarantee stability. Prime95 torture tests were successfully run at the timings listed in the above charts. We also ran ScienceMark (memory tests only) and Super Pi. None of the three stress tests created any stability problems for the SiS755 Reference Board at these memory timings.
While it was good to see the SiS able to run with 3 DS DIMMs installed, we did notice unusual latencies when testing with 3 DIMMs in ScienceMark 2.0. While performance was completely stable with 3 DIMMs, latency with 3 DIMMs is almost twice the latency with 2 DIMMs. Since the latency numbers with the SiS had been quite good in tests, we looked a little deeper into what was going on here. Interestingly, latencies tested virtually the same with either one or two DS DIMMs installed. However, once a 3rd DIMM was added, whether single- or double-sided, the latency and memory bandwidth dropped to about half the values with 1 or 2 DIMMs. This may be an issue with just the SiS755 Reference board, and we plan to look at this further on production boards. For now, up to 4 sides (2 DS DIMMs) could be used with top performance. Using 5 or 6 sides (3 DS DIMMs or 2 DS and a single-sided) increased latency significantly. Our advice, until we can test this further on production boards, is to limit memory to 4 sides for best performance with the SiS755.
Facebook
Twitter
RSS
16 Comments
View All Comments
Googer - Thursday, December 16, 2004 - link
Lets Bench IT!MoronBasher - Saturday, December 6, 2003 - link
I have a sis 746, and 735 and i can't wait for this chipset to come out!!!mason - Thursday, November 27, 2003 - link
Great review!Two things -
1. The SIS reference board is almost always the fastest performing board with that particular chipset. I think it's premature to label it a winner until you can find a decent performing production board.
2. SIS has been making great chipsets for the Athlon since the 735, only no one knew about them because they typically ran second to VIA in memory banddwidth, which was what all the major review sites cared about. The HDD performance and overclocking headroom on the 735 and 745 both outclassed the VIA boards (KT266, KT266A, KT333) of the day, but they were hard to find and few manufacturers made anything better than cut-rate boards.
The MSI 745 Ultra and the Leadtek 735 board were possibly the best of the early generation SIS athlon boards.
The current SIS748 chipset for the athlon 400FSB is actually very good, though again slightly slower than the KT600 is memory bandwidth.
I just wish major sites would review the few boards that do use these chipsets, so we could start seeing higher quality boards like the DFI 748.
destaccado - Thursday, November 27, 2003 - link
buy now.....asus k8v is an awesome board, in 2 months you might as well wait another two months, and another.....jeeptreker - Wednesday, November 26, 2003 - link
Purchase timing is my concern. I'm considering the AMD 64 3200+ & the Chaintech ZNF-150 w/ nForce3-150 chipset for Christmas. Am I on the edge of a change? If I wait a month or 2 will the AMD 64 3400+ and SIS755 or nForce3-250 chipsets be out? If these items won't be out for 6 mos. then I'll just be chasing technology.Buy now or wait? Any suggestions will be appreciated.
Gimpus - Tuesday, November 25, 2003 - link
When should we expect motherboards based on the SiS755 chipset to reach retail chanels?NFS4 - Tuesday, November 25, 2003 - link
I'll wait til I see the performance of shipping boards before I jump ship...reference boards tend to be "tweaked" in some case. I'm currently lovin' my VIA K8T800 based Asus K8V Deluxe.justly - Tuesday, November 25, 2003 - link
"To our surprise, SiS755 also provided the best performance in AnandTech benchmarks of any Athlon64 chipset that we have tested."Did this realy surprise Anandtech? it didn't surprise me.
For quite some time SiS southbridge chips have proven to be well designed, in fact whenever disk preformance was tested SiS has had some of the best scores. The main problem SiS has had in the past was with the performance of their memory controller. So it would only make sence that when the memory controller is no longer a concern, as it is with the Athlon 64, SiS would be able to show of its southbridge performance.
Other than your surprise about the performance I thought it was a good article. I hope Anandtech doesn't ignore future SiS (or ALi) AMD solutions like it has in the past with socket "A".
Pumpkinierre - Monday, November 24, 2003 - link
I noticed your description of the erratic booting/stability behavior when overclocking which is similar to overclocking problems described in the ipkonfig article-today's AT news post:http://www.ipkonfig.com/Articles/AMDCastrated/Page... .Is this due to data corruption in the HT link as a result of CPU speed being linked to the HT link speed?
HammerFan - Monday, November 24, 2003 - link
I think performance will be influenced most by how "streamlined" everything in the chipset is (i.e. IDE drivers at maximum performance, things such as LAN and RAID taken off of the PCI bus and put on the Southbridge).