Asus P4P800s PAT = Memory Turbo?

[Thor86]

Link

*Edit* - for Title change.

 

[gramboh]

Any way to confirm this? If it's true, I wonder if Intel will try to force Asus to disable it in future BIOS or not, I thought it was supposed to be limited to 875.

 

[extro]

Kyle at www.hardocp.com says it's not so. Check out the 9th Edition for 5/28/03.

"This simply is not true. Intel's Performance Acceleration Technology is very much a function of hardware. The resources to enabled PAT on a 865PE chipset would simply be too great if it is possible at all. So get it out of your head that Asus is enabling PAT on their Springdale chipset, as that simply is not true. I will have to categorize this as simply bad marketing and not an outright lie. What is true is that Asus has enabled some incredibly aggressive memory optimizations on their P4P800 865PE mainboard."

 

[Rand]

Well whatever it is they are doing it's managed to boost the P4P800 benchmarks above that of the Canterwood in almost every review I've seen.... besides the one at THG but that seems to be the exception.
I'd really like to know what exactly they are doing, and whether it would be viable to pull it off with the P4C800.

Linky to Asus Press Release

Until I see some proof otherwise I'm going under the assumption that it does indeed enable PAT as Asus claims.
It would certainly explain the performance, and I'm more inclined to believe Asus over HardOCP.

 

[oldfart]

Its not PAT, its "MAM", Memory Acceleration Mode. Which is much like PAT. The thing is, like PAT, it only works @ 1:1 mem ratio. 5:4 or 3:2, it doesn't. Most overclockers have to use the 5:4 ratio unless they have some super high end memory, or a higher multiplier CPU.

Prometheus did some testing. Read about it here.
and also here

 

[Tom]

From a quick perusal of those articles the only question I would have is about the memory speed at the 5:4 ratio being DDR396. This test was of MAM not PAT, so I don't know if it matters, but at least some people have reported that PAT only works when memory is running at DDR400. At least what I saw Prometheus did not test the 5:4 ratio at a high enough FSB for the 5:4 ratio to give a memory speed over DDR400.

Probably not significant, I don't know how accurate the info about PAT requiring DDR400 is, and if MAM would have the same requirement.

Oh, the other thing is apparently the 3:2 ratio wasn't actually tested, I guess the assumption is that it would give the same results as 5:4 but it would be nice to see some actual evidence. So far the only analysis I've seen that compared 1:1, 5:4, and 3:2, was in Zroc's big post, which didn't address PAT in that part of his analysis. The interesting thing was his results indicated app. 2% penalty for 5:4 and 5% penalty for 3:2, as compared to 1:1, all at 200FSB. The important point to remember though is this penalty isn't necessarily because of any overhead associated with using a memory ratio, but could just be the fact that by using a ratio at the same FSB obviously the result is slower memory.(DDR333 and DDR266 as opposed to DDR400)

What would be interesting would be a comparison of say a 3.0c running at stock speed, 200FSB, 1:1 ratio giving DDR400 versus a 2.4c overclocked to 3.0g, 250FSB using 5:4 ratio also giving DDR400. In this case I would suspect that the increased FSB would outweigh any penalty from using a 5:4 ratio and actually perform better. Maybe such a test is out there and I've just overlooked it.

Anyway the reason this interests me is I read a lot of posts about 1:1 being much better than using 5:4 and 3:2, but I don't see much evidence that it's right.

 

[ronss]

don,t believe it has pat. but whatever asus has done, this board is fast-p4p800 deluxe . it tangles with the p4c800 board in lot of the benchmarks. could be that asus has tweaked the bios