ASUS Maximus II Formula - new P45 king?

[Grnfinger]

I'm thinking this board is a huge waste of cash
My X38 Maxmius would kick the out of the bandwidth shown here lately
1300MHz ram and you can pull mid 9's in everest, and sadly a board half in price is beating the ROG line. Asus has really dropped the ball here, they should have spent more on engineering and less on the eye candy.
1175MHz on a X38 with a 8400 broke 10k easy, I'm thinking this board is going to ebay and a Rampage X48 is in order.
Clearly P45 is not the answer

[Cable_Tie_Guy]

This board is good!
I only have some average ram (teamX8500) but on this board it can fly!
I am doing 10.000 MB/s in everest with DDR2-1166 5-5-5-15 at only 1.98v real on this board!

1.98v real DDR2-1166 5-5-5-15 with Maximus II Formula:

[Cable_Tie_Guy]

Grnfinger,

You are totally right about the bandwidth problems with P45... Most of it is because of its PL limitations... For sure, a P45 board will have lower numbers on everest then a good x48 or even P35. But this is a P45 chipset issue... not really a MIIF issue. Other P45 boards will show poor numbers on high clocked DDR2 too.

So I just don't agree with you, when you say it is "this" board problem. It is a P45 thingy.

Also, check my results on the last pic. Of course you shouldn't consider the "write / copy" numbers because those will depend much more on the CPU clock, and I am at only 3.4GHz (you are at 4.4GHz). For 1.98v real, that is a good deal.

For some reason, I find that it needs less volts to clock the ram on P45, especially with the Maximus II Formula.

I am trying now 1.99v and DDR2-1200 5-5-5-15 on MIIF. I can load windows on that, but then when I stress the ram, it will crash. Will try 2.01 - 2.02v.

Also, I wouldn't consider much those CryptiK DDR2-1300 at mid 9K and poor latency results. There must be something wrong on that because his numbers are really bad, even for a P45.

[Grnfinger]

Same divider, ram speed close enough not to argue. Your running what PL10?
On my Maximus X38 I'm running PL7, your latency sucks and your wirite and copy are mediocre for that speed. agrred the cpu does play a major roll so comparing my 8400 to your quad FSB is slighly unfair.
Board is nice, runs cool and looks fantastic, but thats where it ends.
Performance wise this board is sub par. It performs like a workstation board not a performance board. 16 phase power or not the board is not worth the money.

CryptiK posted 1300MHz+ and look at the bandwidth, rather poor score for 1300+MHz dont you think?
Excellent job btw CryptiK was not taking away from your achievement was using the score as an example for my argument.

[Cable_Tie_Guy]

I am on PL9.
Yes, those DDR2-1300 result are bad, can't argue with that.
I will try a everest run on DDR2-1174 5-5-5-15, witch is exactly what you are running. So we can compare it better ok?

I am sure your x38 will still win, but lets see by how much... Also, if the score is close enough, you could lower your cpu at my speeds and run it again, that would be a perfect compare scenario. I am sure I can't push my 65nm B3 to 4.4GHz on air LOL!

[Grnfinger]

I am on PL9.
Yes, those DDR2-1300 result are bad, can't argue with that.
I will try a everest run on DDR2-1174 5-5-5-15, witch is exactly what you are running. So we can compare it better ok?

I am sure your x38 will still win, but lets see by how much... Also, if the score is close enough, you could lower your cpu at my speeds and run it again, that would be a perfect compare scenario. I am sure I can't push my 65nm B3 to 4.4GHz on air LOL!

Post your score and I'll clock my CPU to your speed on X38 and compare results. This might be interesting.

[Cable_Tie_Guy]

Ok.
I don't really know how to tweak mem, so I will only change 5-5-5-15 on bios and that is it. I am not going to change anything beyond that.
Lets add a memset to the screen so the other parameters are visible too ok?

[Zucker2k]

Someone pass me the popcorn, somebody's about to get raped...

[mikeyakame]

With the P45's PL is roughly set 1 higher to the equiv of X48. PL6 on x48 = PL7 on p45, PL7 on x48 = PL8 on P45 and so on. So PL9 is equiv of PL8 on X48. Which is actually kinda slow. PL7 on P45 is blazing fast at 450MHz FSB. I could never get PL6 to post on a P45 board with < 1.3V vNB

[Grnfinger]

Someone pass me the popcorn, somebody's about to get raped...

[Zucker2k]

Lol, Cable_Tie_Guy makes it interesting with all the little rules. It's like an old western, a duel; Maximus vs Maximus II.

[bait]Mikeyakame and myself are ready to provide moral support, but won't hesitate to jump in the fray if need be. should be like letting lions into a pen. Intel Xpress Chipsets FTW[/bait]

[zlojack]

Hmm... this should be fun.

On the DDR3 front: My board hates my RAM so far and nothing works except AUTO

I'm waiting on some G.Skill PI's which supposedly work much better.

[Grnfinger]

I'm not the sheriff being run out of dodge here but I'll post my results using X38 and E8600 in the early morning.
Got some work stuff to do and its getting late if I dont get to it soon I'll be up all night.
On the weekend I'll do a quad comparison as well to see if it make's much of a difference in results.
I will also post some Q6700 runs I have from b4.

[Cable_Tie_Guy]

P45 M2F
3.42GHz
489FSB
DDR2-1174 5-5-5-15

[mikeyakame]

Not bad at all mate. Try making these changes in Memset. Refresh Period set to 8K or 16K for those clock frequencies. Should improve latency somewhat. Are they 1 or 2GB dimms, 2 rank, 4 rank or 8 rank? Everest should give you this info.

[Cable_Tie_Guy]

mikeyakame,

Thans man! I really don't know about those info but I will try to find them for you.
I am not changing the settings. Now Grnfinger have all the info to clock his x38 on my clocks and them we will be able to tell how much better the x38 is against P45 on the memory thingy.

Grnfinger,

It is all there. Just clock your system (CPU, FSB, MEM, MEM TIMINGS, ETC)respecting the numbers on my pic and then we will have an orange to orange compaison!

Thanks.

[mikeyakame]

mikeyakame,

Thans man! I really don't know about those info but I will try to find them for you.
I am not changing the settings. Now Grnfinger have all the info to clock his x38 on my clocks and them we will be able to tell how much better the x38 is against P45 on the memory thingy.

Grnfinger,

It is all there. Just clock your system (CPU, FSB, MEM, MEM TIMINGS, ETC)respecting the numbers on my pic and then we will have an orange to orange compaison!

Thanks.

Well that Refresh Period is incorrect for that memory frequency. Asus bios don't always adjust it automatically properly or give manual adjustment. Refresh Period is 7.8uS, and tREF == ( tRFC[ns] * rows * banks ) / IO frequency

For 1GB dimms with 256Mbit memory chips, 4 banks , 16K rows and tRFC of 45 and DDR2-1066 which has IO frequency of 533.

tCK = 1000ns (1ms) / 533 (which is the bus clock of 1066) = 1.86ns
tRFC = 45 clocks = tCK * 45 clocks = 83.7ns.

so tREF for this would be:

( 83.7 * 16384 * 4 ) / 533 == 10291 clocks
Optimal Refresh period in clocks = closest value to 10291. Higher is always better.

[Cable_Tie_Guy]

Well that Refresh Period is incorrect for that memory frequency. Asus bios don't always adjust it automatically properly or give manual adjustment. Refresh Period is 7.8uS, and tREF == ( tRFC[ns] * rows * banks ) / IO frequency

For 1GB dimms with 256Mbit memory chips, 4 banks , 16K rows and tRFC of 45 and DDR2-1066 which has IO frequency of 533.

tCK = 1000ns (1ms) / 533 (which is the bus clock of 1066) = 1.86ns
tRFC = 45 clocks = tCK * 45 clocks = 83.7ns.

so tREF for this would be:

( 83.7 * 16384 * 4 ) / 533 == 10291 clocks
Optimal Refresh period in clocks = closest value to 10291. Higher is always better.

WTF was that? You better talk some english or I will ask some admin to ban you.

HAeuHAeu! Joking... I don't know about that level of detail.... Basically, you are saying I should change something to improve the numbers, but I really can't figure what it is! LOL!

I am running 2x1GB rams... dual channel. I just changed BIOS to 5-5-5-15. Every other setting is on auto. Loaded windows and benched everest. Then I opened memset to show the full settings (I didn't change anything there) and took the screenshot.

[Cable_Tie_Guy]

Ok I got it now, you are saying that I should change Refresh Period from 4K to 8K or 16K for better latency and maybe a little better bandwidth.
Well, now it is like this. Lets wait for Grnfinger benchs..

[FELIX]

Where do you know it: tREF == ( tRFC[ns] * rows * banks ) / IO frequency !??!

[CryptiK]

The Maximus II Formula is not so great, but it is acceptable. Perhaps your average 'gamer' would not even realize it's shortcomings. The fact that I can do better on my P5Q-Deluxe is not a good thing though. P45 is not the platform to be on for efficiency at a given FSB/memory speed. The P35 and X38/48 beat it by a substantial amount from what I have seen. I'm interested to see the results of the head to head X38 vs P45.

This is the best bandwidth & lowest latency the M2F can do at 434 FSB & ram at 651 MHz. The other result with the high 50ns latency was at tRD (PL) 10. If i had 520 FSB+ it would look a lot better, but even at those clocks, the P35/X38 would destroy it.

Everest @ 434 FSB, 651 MHz, tRD (PL) 6



Everest @ 570 FSB, 570 MHz, tRD (PL) 10

[Cable_Tie_Guy]

CryptiK,

I don't wanna fight or anything. Don't get me wrong, my english is bad and sometimes I may sound bad, but its not my intention.

There is something very wrong with your results with your Maximus II Formula. Your results are very bad. I don't know much about memory o/c, but I can say for sure, that your result for a DDR2-1300 5-5-5-15 AT PL6 is VERY WRONG!

Maybe you have a bad mobo, or a bad ram, I don't know, but this is not what I see here.
You keep saying this board is bad, but maybe is the other way arround?

Check this out:
2.15v for DDR2-1200 5-5-4-10 PL8 on Maximus II Formula:
PS: Remember "write / copy" speeds are very related to the cpu clock. I am at 3.5GHz only

[CryptiK]

It's not my ram, I have run those sticks on P35, X38 and P965 boards, and the results were much better. Also when you say the write/copy speed is related to the cpu clock, that is not exactly correct, it is much more related to the cpu FSB, overall clock makes a comparatively small difference.

This is my board @ 500 FSB, 600 MHz ram PL 10, and with the exception of the read bandwidth and latency (lower and higher respectively, due to higher PL of 10) I get a better result than you.

Also my result at 651 MHz is not 'very wrong', it is the result I got and is completely in line with other results from P45 boards. You said yourself you do not know a lot about memory overclocking, so perhaps you should consider stopping the arrogant comments.

[Cable_Tie_Guy]

You can't say that because it is not true.
Post a 7x500 = 3.5GHz DDR2-1200 and then lets compare oranges to oranges ok? I sure can't bench @ 4GHz with a 65nm B3 on air... or you can just wait for my 45nm to arrive, and they we will see too

You are showing 10% more latency @ 4GHz then me on 3.5GHz... your result is WAY worst, we can't even compare.

[CryptiK]

How old are you, 16? You are annoying a few people already on this forum. Stop with the attitude, it does not suit xtremesystems at all; here we are more mature and reserved. I have seen you post WTF this and WTF that, continually capitalize and repeat yourself ("HEAT MONSTER' is a good example and you wrote it numerous times consecutively), and you overuse exclamation marks like they're going out of style. You also arrogantly question people and results in such a way as to be insulting. Please start to conduct yourself in a more appropriate manner, befitting the way we do things here at XS.

Concerning the results, look back one page, at the everest screen I posted of 570 FSB with 10 MHz lower overall cpu clock than the one I just posted above. It has higher read/write/copy bandwidth and has lower latency despite having a lower overall cpu clock, but a 70 MHz higher FSB. That should demonstrate FSB has a greater effect, as I just said above. If you don't want to believe me I couldn't care less, but those two screens prove it.