Foxconn and Gigabyte Tackle Socket AM2
by Jarred Walton on June 22, 2006 1:30 AM EST- Posted in
- Motherboards
Overclocking
As is typical of Gigabyte motherboards, you have to press Control+F1 in order to reveal many advanced options within the BIOS. We don't particularly like this mechanism, and would prefer that Gigabyte simply embed any settings that they feel are dangerous under additional submenus. Still, as long as you know how to access the advanced functions, it's not a major problem. The major area that you can access to by pressing Control+F1 is the advanced chipset menu, where the most important items are memory timings and HyperTransport multipliers. It's a bit odd that voltage adjustments and bus speeds can be accessed in the overclocking submenu, as tweaking the memory settings usually plays a critical role in achieving maximum performance and stability out of an overclock. That said, let's see how the system does in our stress testing.
We initially ran into a severe difficulty with the Gigabyte board and their original BIOS. All attempts to adjust the CPU multiplier had no affect once we were running Windows. This limited the amount of overclocking testing we could perform, but we were eventually able to track down the cause of the problem to a conflict with the AMD Cool-n-Quiet driver. The early BIOS versions we've tested lack the ability to disable AMD's Cool-n-Quiet feature, so once Windows is loaded the driver takes over and any BIOS CPU multiplier settings are lost. Certain applications (like PCMark05) require the driver in order to generate correct results, so hopefully Gigabyte will have a final fix for this issue by the time this board becomes available for purchase.
Once we figured out the problem and hacked the AMD driver, we were able to reach a maximum CPU overclock of 12 x 256 MHz resulting in a CPU speed of 3072 MHz. This was done using the same 4800+ CPU as the previous AM2 motherboard article, and almost matches the results of the Biostar board. Once we switched to a lower CPU multiplier, we were able to run at a maximum HTT bus speed of 308 MHz. This is slightly lower than other premium AM2 motherboards we have tested, and it could easily be explained by the early nature of our particular board. We should be getting a final retail sample in the near future, so we will update our maximum HTT results at that point in time. In the meantime, a 308 MHz HTT bus should still be enough to get the most out of nearly any dual core AM2 processor, unless we start seeing X2 chips that only come with a 9X multiplier.
Memory Stress Testing
The Gigabyte board does not offer the fine granularity of memory voltage controls that the Foxconn board provides, as you are only able to increase voltage in 0.1V increments. We would prefer at the very least 0.05 V increments, and even better are the 0.025V increments that Foxconn and others offer. However, the end result of our memory stress testing with the Gigabyte board turned out to be extremely positive.
With two DIMMs installed, testing was completely stable at 3-3-3-8 2T timings at DDR2-800. (Note that in actual practice we have determined a tRAS setting of 13 to be slightly better overall in performance with our memory modules.) After further tuning, we were also able to run 3-3-3-8 1T timings at DDR2-800, with the Bank Cycle Time (tRC) set to 18. That makes the Gigabyte board the first AM2 motherboard in our labs to successfully run with a 1T command rate at DDR2-800. This results in slightly faster performance, though as with many memory tweaks the difference between 1T and 2T is not something you would typically notice without running benchmarks -- it's on the order of maybe 1-2% faster.
Installing four DIMMs stresses the memory subsystem further, and it should come as little surprise that we had to drop the command rate back to 2T. We were able to continue to run with 3-4-3-9 2T timings, however, once again matching what the best socket AM2 motherboards are able to achieve at DDR2-800 with our Corsair PC2-8500 DIMMs.
As is typical of Gigabyte motherboards, you have to press Control+F1 in order to reveal many advanced options within the BIOS. We don't particularly like this mechanism, and would prefer that Gigabyte simply embed any settings that they feel are dangerous under additional submenus. Still, as long as you know how to access the advanced functions, it's not a major problem. The major area that you can access to by pressing Control+F1 is the advanced chipset menu, where the most important items are memory timings and HyperTransport multipliers. It's a bit odd that voltage adjustments and bus speeds can be accessed in the overclocking submenu, as tweaking the memory settings usually plays a critical role in achieving maximum performance and stability out of an overclock. That said, let's see how the system does in our stress testing.
Gigabyte GA-M59SLI-S5 | |
Processor: | AMD Athlon 64 X2 - 4800+ (AM2) |
Chipset: | nForce 590SLI |
CPU Voltage: | 1.55V (1.300V default) |
Memory Settings: | 5-5-5-18 2T 12x Multiplier 4-4-4-12 2T 9x Multiplier |
DDR2 Results: | 12 x 256HTT = DDR2 1025 9 x 307HTT = DDR2 790 |
HT Multiplier: | 5x below 300HTT, 4x above 300HTT |
Memory Voltage: | 2.2V |
Chipset Voltage: | 1.6V |
HT Voltage: | 1.35V |
Cooling: | Zalman CNPS9500 AM2 |
Power Supply: | OCZ GamexStream 700W |
Maximum CPU OverClock: | 256HTT x 12 (3075MHz) +28% |
Maximum FSB OverClock: | 307HTT x 9 (2764MHz) +53% |
We initially ran into a severe difficulty with the Gigabyte board and their original BIOS. All attempts to adjust the CPU multiplier had no affect once we were running Windows. This limited the amount of overclocking testing we could perform, but we were eventually able to track down the cause of the problem to a conflict with the AMD Cool-n-Quiet driver. The early BIOS versions we've tested lack the ability to disable AMD's Cool-n-Quiet feature, so once Windows is loaded the driver takes over and any BIOS CPU multiplier settings are lost. Certain applications (like PCMark05) require the driver in order to generate correct results, so hopefully Gigabyte will have a final fix for this issue by the time this board becomes available for purchase.
Once we figured out the problem and hacked the AMD driver, we were able to reach a maximum CPU overclock of 12 x 256 MHz resulting in a CPU speed of 3072 MHz. This was done using the same 4800+ CPU as the previous AM2 motherboard article, and almost matches the results of the Biostar board. Once we switched to a lower CPU multiplier, we were able to run at a maximum HTT bus speed of 308 MHz. This is slightly lower than other premium AM2 motherboards we have tested, and it could easily be explained by the early nature of our particular board. We should be getting a final retail sample in the near future, so we will update our maximum HTT results at that point in time. In the meantime, a 308 MHz HTT bus should still be enough to get the most out of nearly any dual core AM2 processor, unless we start seeing X2 chips that only come with a 9X multiplier.
Memory Stress Testing
The Gigabyte board does not offer the fine granularity of memory voltage controls that the Foxconn board provides, as you are only able to increase voltage in 0.1V increments. We would prefer at the very least 0.05 V increments, and even better are the 0.025V increments that Foxconn and others offer. However, the end result of our memory stress testing with the Gigabyte board turned out to be extremely positive.
Gigabyte GA-M59SLI-S5 Stable 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: | 8 |
Command Rate: | 1T |
Voltage: | 2.2V |
With two DIMMs installed, testing was completely stable at 3-3-3-8 2T timings at DDR2-800. (Note that in actual practice we have determined a tRAS setting of 13 to be slightly better overall in performance with our memory modules.) After further tuning, we were also able to run 3-3-3-8 1T timings at DDR2-800, with the Bank Cycle Time (tRC) set to 18. That makes the Gigabyte board the first AM2 motherboard in our labs to successfully run with a 1T command rate at DDR2-800. This results in slightly faster performance, though as with many memory tweaks the difference between 1T and 2T is not something you would typically notice without running benchmarks -- it's on the order of maybe 1-2% faster.
Gigabyte GA-M59SLI-S5 Stable 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: | 4 |
RAS Precharge: | 3 |
RAS Cycle Time: | 9 |
Command Rate: | 2T |
Voltage: | 2.2V |
Installing four DIMMs stresses the memory subsystem further, and it should come as little surprise that we had to drop the command rate back to 2T. We were able to continue to run with 3-4-3-9 2T timings, however, once again matching what the best socket AM2 motherboards are able to achieve at DDR2-800 with our Corsair PC2-8500 DIMMs.
36 Comments
View All Comments
JarredWalton - Friday, June 23, 2006 - link
Must be - I have a Canan Rebel EOS now, and I haven't come to grips with the manual yet. LOLsprockkets - Friday, June 23, 2006 - link
Rubycons I believe are Taiwaneese caps, but they had the proper formula for the electrolyte in them so they never "blew up", whereas others tried to copy the formula and got it wrong and did "blow up." I still have boards with them from the 2000-2002 time frame still working whereas the other boards with other caps are leaking all over and the boards failed.That is no longer an issue, but still, they are good.
About those new connectors, anyone use a normal cable on them? I did on one, the MSI GF6150 board, and litterally had to crush the connector cable to get it out because it was not one of the new style ones. Anyone else have this happen to them?
JarredWalton - Friday, June 23, 2006 - link
I haven't had a problem with any cables in the new style connector, but here's a shot of the GB cables:http://images.anandtech.com/reviews/motherboards/r...">Gigabyte (Pre-Release) Accessories
Basically, there's a small metal latch that you depress to allow the cable to disconnect. I've seen SATA connections that were so loose they could literally fall off with a slight bump, so the latch avoids that. If you have a cable with a really fat connector, it could cause problems, but all of the other SATA connectors I've seen are about the same, minus the metal latch.
Phiro - Thursday, June 22, 2006 - link
On the first page matrix, you list the Gigabyte as having the Realtek 883, then from about then on, you switch to the 888. Which one is it?Gary Key - Friday, June 23, 2006 - link
The board has the ALC-888 but Gigabyte had planned on the ALC-883 in the beginning, unfortunately the first set of drivers we used incorrectly reported the chipset but after looking at the chip itself (magnifying glass) and running the latest drivers confirmed our sample had the ALC-888. Sorry we missed the update on the first chart.JarredWalton - Thursday, June 22, 2006 - link
It's ALC888, though that's about the same as 883. 888T would be the interesting option, as that adds some support for Skype and VoIP.photoguy99 - Thursday, June 22, 2006 - link
Why would anyone invest in AM2 now when we can see the price/performance shakeout of Conroe in 1 month?Maybe AM2 will come out as a good choice - If a 4800+ drops to $199 I can see getting on board.
If anyone out there is jumping now for an AM2 I would honestly be interested in the logic behind it.
glennpratt - Friday, June 23, 2006 - link
I just bought an AM2 3500 for $100. I doubt Conroe will be at that price point anytime soon and my budget isn't flexible, plus you can't get a geforce 6150 for intel (HTPC). So for $300 I have an excellent upgrade to my media center IMO. New Case, A64 3500, 1GB DDR2, gerforce 6150.I would see your point if I was shooting for the latest and greatest, but personally I never spend more then ~$100 for a component in my PC's (usually < $80). And in my experience things don't change too quickly in that price range. Regardless of hype.
photoguy99 - Friday, June 23, 2006 - link
That makes sense Glen, good luck with your new system.Mant - Thursday, June 22, 2006 - link
This is off topic...does anyone know where can I buy one of those cool new copper heatpipe coolers that are on these boards to replace the Turbo-charged-extreme-jet-turbine fan on my 939 motherboard? I'm dying to retire that POS and this would be just the thing to replace it!