Investigations into Socket 939 Athlon 64 Overclocking
by Jarred Walton on October 3, 2005 4:35 PM EST- Posted in
- CPUs
RAM Latency
We talked about memory latency a bit before, and we estimated the total latency. Using CPU-Z and the included Latency.exe utility, we can get an actual real value for memory latency - or at least, more real than our estimate. There are different ways of measuring memory latency, but we simply took the highest reported value from the table that was generated. For every system, that value was in the bottom-right corner of the table, indicating a data set of 32MB and a stride of 512. While the values may or may not be entirely accurate, they should at least be consistently measured across all the tested configurations. Here are the results, in CPU cycles as well as nanoseconds. Remember that at higher CPU speeds, each cycle is faster, so pay more attention to the nanoseconds graph.
While the value RAM is clearly slower than the expensive RAM, the big question is: how much performance do you actually lose by opting for more economical RAM? This is one of the areas that we are interested in testing for this article, and we'll comment on the results throughout the benchmarks.
RAM Possibilities
One area of the BIOS that's missing (for now) is support for additional RAM ratios. The reality is that the RAM speed is derived from the CPU clock with a divider. Ideally, we'd like to see a BIOS that gives direct access to the CPU divider rather than hiding it behind approximate memory speeds. That would allow for the selection of a larger range of options, but we're not sure if that's something that AMD controls in the CPU or if the BIOS programmers can do this. Here's a list of the standard RAM Settings, CPU multipliers and the resultant memory dividers that were available in the Infinity BIOS.
We have a LanParty UT nF3 250Gb that includes support for many in-between options from DDR200 through DDR500. Some of the interesting inclusions are DDR240, DDR280, DDR300, DDR360, DDR440, and DDR500. (CPU-Z actually failed to report the divider on a couple of those settings, but the resultant RAM speed was still read properly.) While additional memory dividers on the high end won't really help tweakers looking to get the most from the TCCD, BH5, or CH5 RAM, they can be particularly useful when using value RAM. All you need to do is select the appropriate divider to get your RAM under DDR400 - assuming standard value RAM. That way, you wouldn't take as much of a performance hit by running something like 10x250 (2.5 GHz). You could select a 13X divider rather than the standard 15X divider.
Additional ratios can even be useful for tuning performance RAM. For example, OCZ VX would not run stable above DDR500 (10x250), and we had to use 2-3-3-8-1T timings even then. (The 3.2V limit of the motherboard was at least partially to blame.) Running at 10x270, we had to drop to DDR333 (CPU/12), which resulted in the RAM running at DDR450 rather than the DDR540 result that would have been required for the normal CPU/10 ratio. However, if we could have selected a CPU/11 ratio, we could have run the RAM at DDR490 and gained a bit more performance. The additional ratios aren't required, but they would be nice to have.
If the last two paragraphs didn't make sense, then you can guess why we don't get additional access to RAM dividers. Experienced users might know how to make use of the settings, but many people would simply get confused; a "Catch-22".
We talked about memory latency a bit before, and we estimated the total latency. Using CPU-Z and the included Latency.exe utility, we can get an actual real value for memory latency - or at least, more real than our estimate. There are different ways of measuring memory latency, but we simply took the highest reported value from the table that was generated. For every system, that value was in the bottom-right corner of the table, indicating a data set of 32MB and a stride of 512. While the values may or may not be entirely accurate, they should at least be consistently measured across all the tested configurations. Here are the results, in CPU cycles as well as nanoseconds. Remember that at higher CPU speeds, each cycle is faster, so pay more attention to the nanoseconds graph.
While the value RAM is clearly slower than the expensive RAM, the big question is: how much performance do you actually lose by opting for more economical RAM? This is one of the areas that we are interested in testing for this article, and we'll comment on the results throughout the benchmarks.
RAM Possibilities
One area of the BIOS that's missing (for now) is support for additional RAM ratios. The reality is that the RAM speed is derived from the CPU clock with a divider. Ideally, we'd like to see a BIOS that gives direct access to the CPU divider rather than hiding it behind approximate memory speeds. That would allow for the selection of a larger range of options, but we're not sure if that's something that AMD controls in the CPU or if the BIOS programmers can do this. Here's a list of the standard RAM Settings, CPU multipliers and the resultant memory dividers that were available in the Infinity BIOS.
| Standard AMD Memory Ratios | |||
| RAM Setting | CPU Multiplier | RAM Divider | True RAM Speed |
| DDR200 | 9X | CPU/18 | DDR200 |
| DDR266 | 9X | CPU/14 | DDR257 |
| DDR333 | 9X | CPU/11 | DDR327 |
| DDR400 | 9X | CPU/9 | DDR400 |
| DDR200 | 10X | CPU/20 | DDR200 |
| DDR266 | 10X | CPU/15 | DDR267 |
| DDR333 | 10X | CPU/12 | DDR333 |
| DDR400 | 10X | CPU/10 | DDR400 |
| DDR200 | 11X | CPU/22 | DDR200 |
| DDR266 | 11X | CPU/17 | DDR259 |
| DDR333 | 11X | CPU/14 | DDR314 |
| DDR400 | 11X | CPU/11 | DDR400 |
| DDR200 | 12X | CPU/24 | DDR200 |
| DDR266 | 12X | CPU/18 | DDR267 |
| DDR333 | 12X | CPU/15 | DDR320 |
| DDR400 | 12X | CPU/12 | DDR400 |
We have a LanParty UT nF3 250Gb that includes support for many in-between options from DDR200 through DDR500. Some of the interesting inclusions are DDR240, DDR280, DDR300, DDR360, DDR440, and DDR500. (CPU-Z actually failed to report the divider on a couple of those settings, but the resultant RAM speed was still read properly.) While additional memory dividers on the high end won't really help tweakers looking to get the most from the TCCD, BH5, or CH5 RAM, they can be particularly useful when using value RAM. All you need to do is select the appropriate divider to get your RAM under DDR400 - assuming standard value RAM. That way, you wouldn't take as much of a performance hit by running something like 10x250 (2.5 GHz). You could select a 13X divider rather than the standard 15X divider.
Additional ratios can even be useful for tuning performance RAM. For example, OCZ VX would not run stable above DDR500 (10x250), and we had to use 2-3-3-8-1T timings even then. (The 3.2V limit of the motherboard was at least partially to blame.) Running at 10x270, we had to drop to DDR333 (CPU/12), which resulted in the RAM running at DDR450 rather than the DDR540 result that would have been required for the normal CPU/10 ratio. However, if we could have selected a CPU/11 ratio, we could have run the RAM at DDR490 and gained a bit more performance. The additional ratios aren't required, but they would be nice to have.
If the last two paragraphs didn't make sense, then you can guess why we don't get additional access to RAM dividers. Experienced users might know how to make use of the settings, but many people would simply get confused; a "Catch-22".
Facebook
Twitter
RSS
101 Comments
View All Comments
JarredWalton - Monday, October 3, 2005 - link
It's tough to say how things will pan out long-term. 1.650V seems reasonably safe to me, but I wouldn't do it without a better HSF than the stock model. The 1.850V settings made me quite nervous, though. If you can get your CPU to run at 1.600V instead of 1.650V, that would be better, I think. There's also a possibility that slowing down your RAM slightly might help the CPU run at lower voltages. I'd sacrifice 5% to run what I consider a "safer" overclock, though really the thought of frying a $140 CPU doesn't concern me too much. That's less than any car repair I've had to make....cryptonomicon - Monday, October 3, 2005 - link
well for most overclocks a reasonable ("safe") increase of voltage is 10-15%. however that is just a guideline, it may be more or less. there is sort of a way to find out: if you work on overclocking to the maximum of your chip while scaling the voltage, you will eventually hit a place where you have to increase the voltage dramatically just to get up the next FSB bump. for example if you are at 2500mhz and 1.6v, then it takes 1.75v just to get to 2600mhz, then you have hit that boundary and should go back down immediatly. when the voltage to cpu speed ratio is scaling consistently, then things are fine. but once the voltage required becomes blatently unbalanced, that is the logical time to stop... unless you have no concern for the longetivity of the chip.Ecmaster76 - Monday, October 3, 2005 - link
Finally goaded me into overclocking my P4 2.4c. I had been planning for a while but never bothered too.So I got bored and set the FSB to 250mhz (I went for my goal on my first try!) with a 5:4 (still DDR400) memory ratio. It works great at stock cooling + stock voltage. I will have to do some long term analysis of stability but since I am building a new system before the years end I don't really care if it catches on fire. Well as long as it doesn't melt some of my newer nerd toys that are attached to it.
lifeguard1999 - Monday, October 3, 2005 - link
I am running an AMD Athlon 64 3000+ Processor (Venice) @ 2.7 GHz, stock HSF; 1.55V Vcore; DFI LANPARTY nF4 SLI-DR. It was cool seeing you run something similar to my setup. I run value RAM and it seems that I made the right choice for me (giving up at most 5% performance). You ran at Vcores much higher than I do, so it was interesting to see the CPU handle that.The only thing I would add to this article is a paragraph mentioning temperatures that the CPU experienced.
mongoosesRawesome - Monday, October 3, 2005 - link
yes, i second that. temps at those volts using your cpu cooler as well as with maybe a few other coolers would be very helpful. also, if you could do a few tests using different coolers to see when temps hold you back.JarredWalton - Monday, October 3, 2005 - link
I've got some tests planned for cooling in the near future. I'll be looking at CPU temps for stock (2.0 GHz) as well as 270x10 with 1.750V. I've even got a few other things planned. My particular chip wouldn't POST at more than 2.6 GHz without at least 1.650V, but that will vary from chip to chip. The XP-90 never even got warm to the touch, though, which is pretty impressive. Even with an X2 chip, it barely gets above room temperature. (The core is of course hotter, but not substantially so I don't think.)tayhimself - Tuesday, October 4, 2005 - link
Good article, but your Vcore seems to scale up with most of the increments in speed? Did you HAVE TO raise the vcore? Usually you can leave the vcore until you really have to start pushing. Comments?JarredWalton - Tuesday, October 4, 2005 - link
2.20GHz was fine with default 1.300. 2.40GHz may have been okay; increasing the Vcore to 1.40V seemed to stabilize it a bit, though it may not have been completely necessary. 2.60GHz would POST with 1.450V, but loading XP locked up. 1.550V seemed mostly stable, but a few benchmarks would crash. 2.70GHz definitely needed at least 1.650V, and bumping it up a bit higher seemed to stabilize it once again. 2.80GHz was questionable at best even at 1.850V with the current cooling configuration. It wouldn't load XP at 2.80GHz at 1.750V, though.JarredWalton - Tuesday, October 4, 2005 - link
My memory on the voltages might be a bit off. Personal experimentation will probably be the best approach. I think I might have erred on the high side of required voltage. Still, past a certain point you'll usually need to scale voltage a bit with each bump in CPU speed. When it starts scaling faster - i.e. .1V more to get from 2700 to 2800 MHz - then you're hitting the limits of the CPU and should probably back off a bit and call it good.tayhimself - Tuesday, October 4, 2005 - link
Thanks a lot for your replies. Looks like there is a fair bit of overclocking even if you dont increase the Vcore too much to help save power/noise etc.Cheers