AM2 Motherboards-Part 4: ATI Crossfire Xpress 3200
by Wesley Fink on August 21, 2006 7:00 AM EST- Posted in
- Motherboards
Overclocking
MSI reached one of the higher overclocks in our roundup, topping out at 335x9. At stock 12x the board maxed out at 258. Both of these are a bit lower than the ASUS with its 8-phase power regulation, but results are still remarkably close to the ASUS considering this is a cheaper board and a much simpler design. Most of the credit for the overclocking performance should go to the ATI RD580 chipset. The K9A was very stable at the fastest 335 speed. Perhaps just as important, the MSI recovered gracefully from bad overclocks - allowing a reboot and entry into the BIOS to fix the settings from a bad OC.
Memory Testing
Optimum tRAS
DDR2 memory behaved a little differently in RD580 tRAS testing than we have seen on other AM2 boards. As you can see from the standard chart below, a 2GB kit of Corsair 8500 (DDR2-1066) achieved the best bandwidth at tRAS settings of 12 to 14. As a result we used 13 as our tRAS setting for all benchmarking.
Since the optimum tRAS timing seems to vary with DDR2 depending on chipset, motherboard, and CPU, the best way to achieve maximum bandwidth with any configuration is to run your own series of tRAS timing tests with MemTest86, which is a free program. It only takes a few minutes to find out what works best for your individual configuration.
Memory Stress Testing
The MSI K9A supported a wide enough range of memory voltages and timings to allow memory stress testing at 3-3-3-13 timings at DDR2-800. This has become our standard test setup for Socket AM2 and Core 2 Duo memory testing. This is mainly because these aggressive timings at DDR2-800 almost always yield the best overall performance on these two processors. Conroe may be a little faster at aggressive DDR2-1066 timings, but AM2 does not normally support DDR2-1067, leaving DDR2-800 as the highest performing common memory speed.
As found in our Conroe Buying Guide: Feeding the Monster, most DDR2-800 memory is rated at conservative 5-5-5-15 timings, because that assumes you only run with the default 1.8 V DDR2 voltage. We have found in our DDR2 testing that many modules are easily able to run 4-4-4-12 timings at 2.0 V, and most high-end modules like the Corsair modules we are using for benchmarking (rated at 5-5-5-18 DDR2-1066) are generally able to run with 3-3-3 timings at DDR2-800 and approximately 2.2V. Even value ram based on Elpida memory chips can manage 4-3-3 timings at DDR2-800 with 2.2V to 2.3V memory voltage.
With two DIMMs installed, testing was completely stable at 3-3-3-13 1T timings at DDR2-800. This is very rare since AMD does not really support 1T timings in the current on-chip memory controller. However, MSI joins Gigabyte and ASUS as the only AM2 boards to support Command Rate 1 at DDR2-800. We did find, with the Auto or 2T setting, that stability could be achieved at a lower 2.2V setting. This is important to know when trying to squeeze the most from motherboards with limited memory voltage options.
Several benchmarks were run to compare performance at 1T and 2T timings. As in past testing, we could not find any real-world performance improvements with the 1T timings compared to 2T. This will likely remain the case until AMD releases a new AM2 memory controller that will officially support 1T Command Rate.
Installing four DIMMs stresses the memory subsystem further, but the MSI K9A did handle 4 DIMMs at the same timings and a 2T command rate. While the system was stable at the same 3-3-3-13 timings that worked best with 2 DIMMs we needed to adjust to the top of the available voltage options and did wish we had a bit more DIMM voltage.
The MSI and ASUS are the only AM2 motherboards we have tested so far that are able to run 4 1GB DIMMs at 3-3-3 timings. That performance speaks very well for the stability of both designs, since the memory controller in all our motherboard tests is the same initial release AM2 CPU.
MSI K9A Platinum Overclocking Testbed |
|
Processor: | AM2 4800+ (X2 2.4GHz 1MB L2 Cache per core) |
CPU Voltage: | 1.525V (default 1.4V) |
Cooling: | AMD Stock Heatpipe FX62 Cooler |
Power Supply: | OCZ Power Stream 520W |
Memory: | Corsair Twin2x2048-PC2-8500C5 (2x1GB) (Micron Memory Chips) |
Hard Drive | Hitachi 250GB 7200RPM SATA2 16MB Cache |
Maximum OC: (Standard Ratio) |
258x12 (Normal, 3-3-3-13) 3096MHz (+29%) |
Maximum FSB: (Lower Ratio) |
335 x 9 (For Overclocking, 3-3-3-13) (3015MHz, 2 DIMMs in DC mode) (+68% Bus Overclock) |
MSI reached one of the higher overclocks in our roundup, topping out at 335x9. At stock 12x the board maxed out at 258. Both of these are a bit lower than the ASUS with its 8-phase power regulation, but results are still remarkably close to the ASUS considering this is a cheaper board and a much simpler design. Most of the credit for the overclocking performance should go to the ATI RD580 chipset. The K9A was very stable at the fastest 335 speed. Perhaps just as important, the MSI recovered gracefully from bad overclocks - allowing a reboot and entry into the BIOS to fix the settings from a bad OC.
Memory Testing
Optimum tRAS
DDR2 memory behaved a little differently in RD580 tRAS testing than we have seen on other AM2 boards. As you can see from the standard chart below, a 2GB kit of Corsair 8500 (DDR2-1066) achieved the best bandwidth at tRAS settings of 12 to 14. As a result we used 13 as our tRAS setting for all benchmarking.
Memtest86 Bandwidth MSI K9A Platinum with AM2 4800+ |
|
6 tRAS | 2047 |
7 tRAS | 2047 |
8 tRAS | 2047 |
9 tRAS | 2093 |
10 tRAS | 2093 |
11 tRAS | 2141 |
12 tRAS | 2192 |
13 tRAS | 2192 |
14 tRAS | 2192 |
15 tRAS | 2047 |
16 tRAS | 2047 |
17 tRAS | 2047 |
Since the optimum tRAS timing seems to vary with DDR2 depending on chipset, motherboard, and CPU, the best way to achieve maximum bandwidth with any configuration is to run your own series of tRAS timing tests with MemTest86, which is a free program. It only takes a few minutes to find out what works best for your individual configuration.
Memory Stress Testing
The MSI K9A supported a wide enough range of memory voltages and timings to allow memory stress testing at 3-3-3-13 timings at DDR2-800. This has become our standard test setup for Socket AM2 and Core 2 Duo memory testing. This is mainly because these aggressive timings at DDR2-800 almost always yield the best overall performance on these two processors. Conroe may be a little faster at aggressive DDR2-1066 timings, but AM2 does not normally support DDR2-1067, leaving DDR2-800 as the highest performing common memory speed.
As found in our Conroe Buying Guide: Feeding the Monster, most DDR2-800 memory is rated at conservative 5-5-5-15 timings, because that assumes you only run with the default 1.8 V DDR2 voltage. We have found in our DDR2 testing that many modules are easily able to run 4-4-4-12 timings at 2.0 V, and most high-end modules like the Corsair modules we are using for benchmarking (rated at 5-5-5-18 DDR2-1066) are generally able to run with 3-3-3 timings at DDR2-800 and approximately 2.2V. Even value ram based on Elpida memory chips can manage 4-3-3 timings at DDR2-800 with 2.2V to 2.3V memory voltage.
MSI K9A Platinum DDR2-800 Timings - 2 DIMMs (2/4 slots populated - 1 Dual-Channel Bank) |
|
Clock Speed: | 200MHz |
Timing Mode: | 800MHz - Default |
CAS Latency: | 3 |
RAS to CAS Delay: | 3 |
RAS Precharge: | 3 |
RAS Cycle Time: | 13 |
Command Rate: | 1T |
Voltage: | 2.3V (2.2V with 2T) |
With two DIMMs installed, testing was completely stable at 3-3-3-13 1T timings at DDR2-800. This is very rare since AMD does not really support 1T timings in the current on-chip memory controller. However, MSI joins Gigabyte and ASUS as the only AM2 boards to support Command Rate 1 at DDR2-800. We did find, with the Auto or 2T setting, that stability could be achieved at a lower 2.2V setting. This is important to know when trying to squeeze the most from motherboards with limited memory voltage options.
Several benchmarks were run to compare performance at 1T and 2T timings. As in past testing, we could not find any real-world performance improvements with the 1T timings compared to 2T. This will likely remain the case until AMD releases a new AM2 memory controller that will officially support 1T Command Rate.
MSI K9A Platinum DDR2-800 Timings - 4 DIMMs (4/4 slots populated - 2 Dual-Channel Banks) |
|
Clock Speed: | 200MHz |
Timing Mode: | 800MHz - Default |
CAS Latency: | 3 |
RAS to CAS Delay: | 3 |
RAS Precharge: | 3 |
RAS Cycle Time: | 13 |
Command Rate: | 2T |
Voltage: | 2.3V |
Installing four DIMMs stresses the memory subsystem further, but the MSI K9A did handle 4 DIMMs at the same timings and a 2T command rate. While the system was stable at the same 3-3-3-13 timings that worked best with 2 DIMMs we needed to adjust to the top of the available voltage options and did wish we had a bit more DIMM voltage.
The MSI and ASUS are the only AM2 motherboards we have tested so far that are able to run 4 1GB DIMMs at 3-3-3 timings. That performance speaks very well for the stability of both designs, since the memory controller in all our motherboard tests is the same initial release AM2 CPU.
28 Comments
View All Comments
Wesley Fink - Monday, August 21, 2006 - link
We plan to test the Abit, but we have not yet received the board.yyrkoon - Monday, August 21, 2006 - link
Good to hear :)mendocinosummit - Monday, August 21, 2006 - link
Sweet. I was hoping that was the case.Wesley Fink - Monday, August 21, 2006 - link
It's not the end of the AM2 board testing. We fully expect great RD580 boards from DFI and Asus. However, after testing 9 AM2 boards in-depth it was time to assess where we were.goinginstyle - Monday, August 21, 2006 - link
Finally a review that does not have Conroe plastered all over it. I have the ECS RD480 board based upon an earlier review of it here. The board has performed well and is stable as a rock. Do you think you maybe have a bad board or is it really as bad as it was described? I was looking forward to going to AM2 with a cheap X3800x2 but did not want a nvidia solution since I already have crossfire. Will there be other RD580 boards from Asus or DFI or maybe someone like Abit?I was excited about this board after reading the review at HardOcp where it received an editors award. Your comment "we have to wonder how ECS could manage to produce such a mediocre motherboard." is completely opposite of the other review. But after reading both reviews again I noticed your test results were very consistent when compared to the other boards that were not reviewed in that article. Looking at the results I have to wonder what board they tested.
Just a few questions if you have time. What would cause the cold boot issue? Does the board have the SB600 or SB460 as listed in the chart. How is the performance of the jmicron chipset? I can not find any test results with it and it seems like everyone is using it now. Thanks for providing some very useful AMD information as they still exist even after Conroe.
Wesley Fink - Monday, August 21, 2006 - link
The ECS has the SB600. ECS included literature that stated the soutn bridge was SB460, but we took off the heatsink and looked. It is definitely an SB600. We will change the ECS spec chart.The cold boot issues can be poor power regulation, the board's boot sequence, or even just a poor BIOS. It IS possible it might be corrected in the furure with a BIOS update, but there are plenty of other AM2 boards that work well today and cost about the same or less.
We noticed in the other website review you reference that they made a passing mention of the missing CAS adjustment, but they still gave it an Editors choice. They also noted the very poor overclocking, but still gave the ECS an Editors Choice. I'm sorry, but we don't give Editors Choice awards to the worse overclocker of 9 competing AM2 boards that does not even offer CAS adjustments.
allnighter - Monday, August 21, 2006 - link
...work comp, win2k+explorerPatrese - Monday, August 21, 2006 - link
Great review, I only missed Oblivion results. I'd like to see more Anantech reviews about motherboards directed to other audiences, like HTPCs, for instance. There are a lot of motherboards directed to that segment, like the Asus M2NPV-VM, and I think it would be great to see a review as detailed as you generally do for the entusiast boards. A MicroATX roundup would be perfect... :)Wesley Fink - Monday, August 21, 2006 - link
A few HTPC motherboard reviews are already in process. You will be seeing these from Gary Key in the near future.jackylman - Monday, August 21, 2006 - link
I'd really love to see some hardcore Sempron undervolting on those reviews. (hint,hint)