yeah you can actually feel the difference between 10.5k vs 11-12k as the needed raw cpu clock to achieve 11-12k for DDR2 is >4500-4800mhz dual core versus 10.5k which can do be around 4000-4400mhz.
Printable View
yeah you can actually feel the difference between 10.5k vs 11-12k as the needed raw cpu clock to achieve 11-12k for DDR2 is >4500-4800mhz dual core versus 10.5k which can do be around 4000-4400mhz.
11K & 12K Everest are milestones of performance achievement, and yes, you can feel the difference between 10.5 and 11 - 12K, but feeling it isn't the aim we dont use those settings 24/7. As eva said, it also has to do with the CPU speed used to attain that bandwidth.
A lot of people here find it fun to overclock and find the limits of their hardware. They also enjoy tweaking things for max stable 1M, 32M, 3D benches etc at various voltages. Some people like yourself just want to find max stable clocks for a sensible voltage and leave it at that.
Hey zsams_ would you mind re-running that Everest bench (using D9's or G.Skill doesnt worry me) at 556 FSB & 667 MHz except use 7x multi? I can't do 4.7 with my chip, so if you could run it at 7x I can directly compare to your P35 board. I'm also interested to see how much read bandwidth is gained from the CPU @ 4.7GHz compared to at ~3.85 - 3.9 GHz.
Also, what vNB & PL did you use to do that run?
I think this is the closest I'm going to get to 650MHz stability. It passed one loop of #5 in memtest but had some errors on the second. I figured that was good enough to boot into Windows and it's dual SuperPi 1M stable. This was with the DFI P35 and, oddly enough, my Team 1200 D9GKX kit. Supposedly P35 and GKX don't get along, but I tried just about every D9GMH kit I have and the GKX beat them all out.
Nice one, is that 2.52v?
Hey since you have an E0 chip, what kind of bandwidth can you pull at 540 - 560 FSB with ram at ~650 - 680?
soon i willQuote:
Originally Posted by CryptiK
pl was 9
you cant compare mfII bandwith to ut-p35 whatever setting you will be -500mbs
1 error is too manyQuote:
Originally Posted by Shiranui Gen-An
Shiranui Gen-An run 510x8 you will like it
from 4 sets of d9gkx i cant do 1200 on dfi:rofl:
but they do 1100 cas4
:yepp: eva2000 it feels like i pushed the nitro buttonQuote:
Originally Posted by eva2000
It was more like 2.55V set in the BIOS. I'm seeing if my Reapers can do 650MHz, if they can I'll try the FSB you recommended.Quote:
Originally Posted by CryptiK
I like that, feels like i hit the nitro button :rofl:
Ok I thought so :).Quote:
Originally Posted by CryptiK
It looks like the best combination was right under my nose. My 24/7 main system houses my Maximus Formula with the Rampage BIOS, and it had been paired with my CellShock 8000 C4 kit at 600MHz C5 2.05V real. I'd since taken them out to test on other systems, but I thought, "What the heck, why not try my main setup with these?" It took 2.45V real on the memory and 1.60V on the NB, but I got it. Passed about 4 or 5 loops of #5 in memtest before I booted to Windows and did these benches. Kind of peeved I bought the DFI P35 T2R and the Maximus II when the board I had already can apparently outdo them both for RAM clocking...
Very nicely done Shiranui, 32M at 651 MHz is very hard, even 620 - 640 is really tough. 2.45v for that is also quite good.
I just can't figure out why I couldn't do the same on my Maximus II. I'd give those sticks more than 2.2V and Memtest would just freeze up. I don't know if I needed less vMCH or more, or what. I might try again.
Well i now know why i have bandwidth problems, i read a article on x-bits, about gigabyte boards, it claims gigabyte boards up till the x48 dq6 have had peformance problems when overclocked, that they are below average peformance when overclocked, zsasm, your right about gigabyte boards, but i guess the new boards dont have this issue, my board and memory are now on the auction block :down:
thats why gigabut boards are easy to overclock cause performance suksQuote:
Originally Posted by neo mike
i tried the 965-dq6 the p5b vanila was better
p35c-ds3r 2 died less than a month 1 died in a month @ stock:rofl:
x38-dq6 thank god it died in 3 weeks
finaly 6 sets of ram rip due to gigabut boards
took em 2 years to make better bioses but i happy i didnt fall in that trap again
Link? I'd like to read that article myself.Quote:
Originally Posted by neo mike
Can i have a general concensus on what the best mobo for clocking 4*1GB Cellshock 1066 C5 with a q9550 might be?
I will get the link, i dont think its the board itself, they dont know how to make good bios, and board no matter how good the componets are, need a very good bios to make it a great board, gigabyte guys cant make bios.
May I ask what problems/limitations you're having with your P5Q-Deluxe?Quote:
Originally Posted by cliffy
Had some free time to mess with another set of Crucial Ballistix non-tracer 16FD5 set @650mhz 5-5-5-15 at 2.425v bios set
http://fileshosts.com/intel/DFI/DFI_...0m53s656ms.png
OMG you changed your avatar!
Really nice run, those are pretty tight timings for 650 MHz and only 2.425v. That board is going really nice too.
Yeah the memory and board handle tight timings pretty well i.e. trfc 28 etc. 2:3 divider seems to max out 32M pi runs at 669mhz 5-5-5-15 though which is lower than what DFI UT P35-T2R/LT X38-T2R could handle (683mhz 5-5-5-15). But i'm loving fsb on DFI DK X48-T2RSB Plus at 563FSB max 32M Pi stable... best P35/X38/X48 board i have had to date :D
Yeah changed my avatar after having same one for nearly 7yrs :)
Seems to fit your nick better, as it is an Eva unit ;)Quote:
Originally Posted by eva2000
i upped the juice also on my GKX and at 2,44v (2,46v in bios) they became stable at DDR2 1333. Passed memtest, memtest86+ and goldmemory.
Not a 24/7 setup imo with that voltage but nice for gaming and stuff cuz its stable now.
http://www.abload.de/img/capture0x4d.jpg
You have absolutely useless setup. Where is MemTest 3.8 screenshot?
Look mine:
Only 1,99 vDimm.
http://img227.imageshack.us/img227/644/32979893kb7.jpg
I'll have to try that setup when I get home. I've been trying to push them on my Maximus II with the FSB at 541 on the 5:6 divider (650MHz). I actually had to play with the DRAM Clock skews to even get memtest to run :eek:Quote:
Originally Posted by Elektrosadist
waterflex, for 1200 i also only need around 2 - 2,04v. But everything above that needs much more voltage. at least my sticks. so for me FSB 500 with ram at 1200 is also possible with what u posted but i wanted to have the ram on 1333 MHz.
Ok. But do you get any advantages? Or you can feel 100mb Read and 1 nanosec?
But you give +0,4 volts to poor sticks.
i no. lol but i don't care. It's called xtremesystems isn't it?
i don't mind simply cuz i bought 6 sticks with micron chips. And in several months i will go to nehalem anyways. And as i mentioned it isn't a 24/7 setup. Whenever i feel like to i load this setting from bios and have my fun with the performance. Most of the time the system runs at 2,5GHz (0,998v) and DDR2 1066 (1,92v @ 5-5-4-12).
;-) More juice more bandwith.
Yeah! it`s XS!
But you said "Not a 24/7 setup imo with that voltage but nice for gaming and stuff cuz its stable now". So then I asked you, can u feel 100 mbps READ and 1 nanosecond :D
When I get my NB block, I`ll try to make Xtreme clock :)
nah man i can't feel any difference at all. i'm just an everest whore.
from FSB 400 on (3,4 GHz with DDR2 1200) i can't feel any difference in my gaming or whateva im doing. I just like to push my hardware a little. i just used your setting it's nearly the same only that im at 52,1 ns access time. but it just gave me a memerror with 2,04v. maybe my sticks or my board needs just a little more at that fsb cause at fsb 400 i can do ddr2 1200 with 2,04v.
You want a NB block ? M8 it's about 35° all the day here with water in air like 99% and I'm benching with no AC just with copper & fan...
But what I want is a real bios to get more BW so !! Please ASUS help us !! THX
Right I'm bored there's nothing to do for getting more than 666 ...were damned lol
my nb is under water too. i put a temp sensor on the nb and mine reaches also 35 max after hours of benching when water reaches 27 degrees. most of the time my nb is at 25 i think.
ha ha, i was trying to figure out what you guys were talking about on the last page (avatar changing) but i couldnt tell what you were talking about,until i turned on show avatar in my edit options. i was blind, but now i can see!
I know at first glance this doesn't look very impressive (500FSB, 600MHz C5 on the memory, PL6) but this was with my DFI LT P35 and my Cellshock 8000 C4 kit. Supposedly P35 and D9GKX don't get along, but I only needed 2.05V set in the BIOS (don't believe Smart Guardian's reading). I had to tweak the Clock Delay settings, but I don't really see much difference between using these and D9GMH.
Can u pass memtest 3.8?
memtest86 2.10 is just as good, if you can pass either one your fine.
Nope. Windows MemTest 3.8 is much more stressful for RAM and NB.Quote:
Originally Posted by neo mike
I used to be unstable even after success memtest86+ pass.
But Memtest 3.8 will tell you, if your RAM and NB stable or not.
I confirm that too, it's true ;)Quote:
Originally Posted by WaterFlex
My Kingston Hyper-X 1066MHz D9GKX kit (I'm lucky - not so many Hyper-X GKXs out there ;)), can bench SPi 1M @ 660MHz 5-5-5-5 with 2.65V ... It kind of sucks ... :(
My Ballistix PC8500 GMH can bench SPi 1M @ 680MHz 5-4-4-4 with tight sub-timings @ 2.8V ;)
So my Kingston kit sucks ... :(
Cant find that article on gigabyte boards peformance, will keep looking :shrug:. I have decided to try the dfi ut 45 trs2 :up: looks like a sweet board. I now boycott gigabyte :down:
Very impressive well done. I'm trying to get my M2F to do some high ram clock SPI32M and so far it won't do it, I know the ram can, but the board is so hard to work with.Quote:
Originally Posted by Elektrosadist
Please try and be more supportive. The guy said it's not a 24/7 clock, but its 32M stable, and that's a real achievement.Quote:
Originally Posted by WaterFlex
Good luck, my M2F is also incredibly hard to bench 32M with, anything over 500 FSB 600 MHz ram is next to impossible to make pass 32M. Even using lower FSB and higher ram speed, I can't get 640 MHz to pass.Quote:
Originally Posted by Shiranui Gen-An
Ok look on one hand you are undermining peoples overclocking achievements, and then later you say you will also try to make an xtreme clock.Quote:
Originally Posted by WaterFlex
When & if you do post an 'xtreme clock', would you expect support or do you want people saying its no good, and asking you if it can pass memtest? Be more supportive please, the aim here is not 24/7 stability with every screen shot we post. They are going to be stable enough for the bench we are doing, there's no point proving 24/7 stability and putting unnecessary strain on the ram when we do not intend to use it 24/7.
well it looks like the mushkin pc-8000 2x1 kit i got from memoryc has one bad stick. i was finally able to get the good one to boot up on my gigabyte UD3P. so i thought id try to see if it would play nice with my cellshock 8000's and it does.so far so good.now im debating on whether to rma them,cause they will probally send something different. but then again i dont even know if these mushkin are even good(d9)?at least now i have 3gigs of ram installed.http://i33.tinypic.com/2a7v7lh.jpg
finally got a little peformance out this stinking gigabyte
Just did 3 runs with my M2F and Cellshock PC-8000 GKX sticks. This board has strange behaviour, it is most stable on the 5:6 divider. The times are bad as I had to run PL8 (PL9 on first bench), Ai Clock Twister on moderate, and have static read and dram read training disabled. The GKX did ok, max SPI32M stable on 2.20v was 630 MHz.
510 FSB, 612 MHz, PL 9
http://i266.photobucket.com/albums/i...PL922vdimm.jpg
517 FSB, 620 MHz, PL 8
http://i266.photobucket.com/albums/i...PL822vdimm.jpg
525 FSB, 630 MHz, PL 8
http://i266.photobucket.com/albums/i...PL822vdimm.jpg
I cant get a trd lower than 8 at this speed, i wish i could run a 7, would really boost the performance, but i think with high fsb and memory speed 8 is not bad i guess.
Quote:
Originally Posted by neo mike
Not bad results. It's impossible to run lower than PL8 on a P45 on the 4:5 divider. All you can do is tweak the other chipset enhancements to get better bandwidth & latency.Quote:
Originally Posted by neo mike
Ok didnt know that, still kinda learing about all these dividers, seeing what i can do on the 400 strap.
I have a question, whats the deal with the straps? Mine always on auto, if i change it do i get better performace?, more stable?
Straps change the chipset timings in relation to the FSB of the CPU being used. The lower the strap the tighter the timings and the better the performance. The strap also defines what memory speeds you have available to choose from.
When you leave the strap on auto, you just choose whatever memory speed you want from the list, and it auto sets the appropriate strap for the speed you chose.
So 400 strap being the worst peformance, and 200 being the best ?. I was on the 266 strap, i was now playing with the 400, my everst score is lower with about the same memory speed, but to me my computer seems to respond faster on the 400 strap, or that is maybe just in my mind. :shrug:
Yes that's correct. If you want to see whats faster, run a SPI32M at 266 strap, then 400 strap, but the difference in mem speed is going to impact that too. I found that on the 266 strap on my M2F, I could not run a certain chipset enhancement on max, I had to back it off a bit. However with the 333 strap, I could run it on the most aggressive setting and even though I had 48MHz less ram speed, I was almost (within 100 mb/s) as good as on the 266 strap. I did not run SPI32M benches to test though.
I'm about to go insane with my board right now, what was stable is no longer stable, increasing/decreasing vcore & Vtt, backing off chipset enhancements, increasing PL, raising vdimm, nothing works. It's really annoying me. Couple that with the lack of ram overclocking ability this board has and I'm looking at getting rid of it. there is two boards I really like, and have to choose between, that's providing I can get rid of my M2F and P5Q-Deluxe (brand new in box) without loosing too much money.
Ok i will try that test, that a very good idea, cause when i was on the 266 strap, at 510 fsb, i was getting 10,350 everest, memory was at 1270 i think, now im the 400 strap at 485 fsb, and 1280 memory, and its about 200 points lower, but im able to run a pl 7 on the 400 strap, and even though my score went down, my latency improved quite a bit, but i will try what you said.
Hey i have that happen you get a overclock, and it sticks for days, then you mess around, and try to go back and now it wont stick, i have saved ones in my bios, that works somedays, and others is trouble:shrug:. Im getting a new board to, im going to lose money, unless i can find someone to buy it for cheap and get something back. Its tough to pick new board, there are alot of choices, but have been doing some serious reading of reviews, and checking threads here, its dfi ut 45 t2rs , or asus p5e3, seem to be 2 very good boards, if i can figure out how to shrink my photos in my digital camera, im going to put them in the for sale section, although it dont look like it gets alot of action.
Well, according to my tests, 266MHz strap is the best when it comes to benchmarking in general ;)
I was playing with P5K Deluxe like one and a half month ago, and did some SPi 32M tests on 266MHz and 333MHz straps ... First one was slightly faster :)
But when I say "according to my tests", I don't mean only SPi 32M benching, 'cause it's just one bencmark and you can't come to a conclusion when your only fact is that one ... Thus, not only I benched SPi 1M & 32M and PiFast (ram timings/sub-timings do not have a great impact on wPrime, that's why I didn't even bother trying it, besides if you run it @ Vista or Server 2008, your score will improve in comparison with the XP one, so :p:), but also 3D benchmarks, and mostly 3DMark01 ... ;)
But I didn't run any Everest tests, or use a specific strap in order to see its impact on 24/7 usage, therefore, you should not take my findings in consideration as far as 24/7 pc usage is concerned ;)
yeah for sure running a whole suite of tests is the best way to ascertain which settings have greatest performance. i just suggested 32M as it's a better indicator than everest.
Yeah i havent run any test really, i know that i really cant get much on the 333 strap, my memory, or computer, whatever dont like it, 266 and 400 works best for me, im going to run some benches on 266, and 400, and then stick with the fastest and best performance, im sure its the 266.
@ waterflex: I can't get any stability on my p45 with trd 8 @ FSB 500 with DDR2 1200. Even at a higher tRD. I can only run 1250,1333 or 1000 @ FSB 500.
Strange though cuz with my OCZ with Infinion chips i could run 1200 at FSB 500.
My bad, mate. Sorry :(Quote:
Originally Posted by CryptiK
You don`t have 5:6 devider? 1200 is not available or what?Quote:
Originally Posted by Elektrosadist
Anyway, try to fix secondary timings: 9,4,5,6,6,5,6.
:rolleyes: It`ll help, i`m sure.
Not a problem mate, just trying to keep this thread nice & friendly :)
Some boards just don't like certain dividers, for example I have found my M2F doesn't really like anything except the 5:6 divider when aiming for 32M or better stability.
Thanks i'll do my own test some time :).Quote:
Originally Posted by CryptiK
This thread just won't die will it? :p: I never thought so many people still had Micron D9s, the fact people do pleases me :D easily the best ICs DDR2 will see. Shame D9s were not developed more otherwise we could have some real badass DDR2 right now.
1200 is available but like i said i couldn't get it stable to pass at least super pi, memtest not to mention.Quote:
Originally Posted by WaterFlex
i may try later to change the sec timings.
but i get for me the best performance at FSB 471 with DDR2 1256 (Performance lv 7, 5-5-4-12 and refresh cycle of 30 @ 2,26v.
Seems its the P45 Chipset that drives us insane. My board does the same thing, I'll run OCCT/Prime/memtest 3.8 and have a completely stable OC, however if I go back and rerun the tests they will fail immediately or close to that.Quote:
Originally Posted by CryptiK
Almost seems like it is Heat related, but that is not the case here. I take temps with my Laser while testing and after to see how hot the NB/Mosfets/SB etc are. 32C is the max I get on the NB for instance,(very good cooling), with about 28C idle.
Funny thing is that if I wait an hour or so, reboot and try the exact same tests again, they pass? Not looking for help just a statement. Just a quirk in the chipset I believe. Again, seems heat related but does not show as such.
Larry
I had the same problem running memtest86. Would pass with flying colors 3 full passes, try it a hour later get errors, i think at least my problem is heat related, need better cooling on my ram, though it dont seem really hot, only warm.
Actually Memtest still passes if I retest others do not, very strange. Like I stated, seems it would be a NB/Mosfet overheating, but the temps don't show it. Most deffinately not the cpu as even running Linpack my temps never exceed 41c.
Anyhow, back on topic, I have some HZ that I have been testing here and have them stable at 583mhz with 2.1v bios set, ( memtest 3.8 and 1 hour OCCT Mix). Bout as high as I can get my cpu stable. Nice ram, got some Cellshock PC8000 on the way to test also.
Larry
Heh, think I sorted it out, it's my cpu, further degrading. Damn 45nm's.Quote:
Originally Posted by Big Lar
I jacked up the Vtt to 1.25v (was at 1.20v) and bumped up the vcore one notch and now it's perfect. I just backed the vcore down one notch to where it was leaving the Vtt at 1.25v and I'm retesting. It's weird as my cpu hates any Vtt setting between 1.20v and 1.24v, but at 1.25 it seems to be fine.
I don't think it's the P45 chipset, although I had very odd behaviour from my P5Q-D, which is one of the reasons why I RMA'd it. I just wish my board could do higher than 666 MHz on the ram, it's pretty weak in that department.
Guys. I`m sure that all owners of MIIF with watercooling, MUST BLOW the CPU socket area with fan even if your NB, SB, POWER, MB temps are ok. It works for me.
I don't have the MIIF but the P5Q-D. However, I made a fan holder if you will quite a while ago out of Aluminum and attached it to the Bottom of the top of my case,( Built my case out of Oak about 5 years ago ) :D Anyhow, I have an 80mm fan on it pointed at my NB which keeps it cool. I also have that ugly Asus fan on the Rear Mosfet sink, as well as a 92mm and an 80mm attached to the mobo tray, 1 pushing air across my SCSI Raid card, the other across my ram. I have for input on the Mobo side of the case 2 x 120mm medium speed fans in the lower/middle front. I have 2 x 120mm fans on the upper rear of the case exhausting out.Quote:
Originally Posted by WaterFlex
On the right side where I have my drives/dvd roms etc, I have 1 120mm on the lower front for intake and 1 120mm exhausting directly behing my drives. PSU is above the drives, about 10"s above exhausting out of course. I have great temps on everything, which is why I point a finger at the chipset/shrug.
Larry
Still just like you Cryptik...waiting for some bios proof ^^ ...are we far from Xmas so :D
So much ramkits to test, so I can't stand waiting like this....
I know Eeky Nox, i can't get any kits over 666 MHz. I'm sure it's the board too. It needs more drive strength to do it, the UD3P has drive strength settings based on target ram speed, M2F has OC Mem charger, enabled. Not good enough for 666+.
can anyone offer up some advice on what a good safe voltage for 4*1GB Cellshock 1066 C5 DIMMS (d9 GMH) would be and a clue as to how much northbridge voltage i might need to run on me P5Q deluxe, starting off aiming for 1133mhz @ 3.6Ghz
I think the EP45 Extreme has those same drive strength settings too. Considering dumping my MIIF for one of those (I know zsamz, Gigabut).Quote:
Originally Posted by CryptiK
I'm not supposed to change mobo before 3month, so it's a bit enoying I feel :D Took this board to be sure to keep it more than 1 week :D
I guess this issue to be fixed up really...or it'll make me cry men ^^ (MIIF is nice in game so I would be able to cell it but I don't want so)
Just pray my friends, just pray....
:rofl::rofl: mfII needs some bios updates to make it better i doubt if the gigubut boards any betterQuote:
Originally Posted by Shiranui Gen-An
hey i'll trade ya x38 to it lol
Quote:
Originally Posted by Elektrosadist
I wonder why you can't get yours to run SuperPi, as of my experience with Gigabyte boards the 5:6 FSB : DRAM setting (or memory 2.4x in Gigabyte P35/P45/X38) and rest of 333 straps are the most stable for 1333MHz CPU's and with all RAM kits I've been using. Would you mind to show me your BIOS settings in Easytune6 and your memory settings in BIOS, maybe I can think of something that you can try?
http://img296.imageshack.us/img296/1...3013905qp2.png
http://img296.imageshack.us/img296/5...3003721uy2.png
Wow! That`s really nice. Well done pangingIII, love yout Everest results!
Here, i can't even run everest benchmark at 1200 with fsb 500. so i booted up with 1250 and 266 strap which works. i would use the same settings for 1200 mhz. I adjusted the cpu and mch ref in easy tune to show you where its at (0,861v in bios). The Dram refs are left on auto. drive strength is DDR2 1333 Mhz. The timings should be seen in memset and in everest. pi time was 11min49.68Quote:
Originally Posted by pangingIII
Ps: yeah nice scores i think in february i buy a dualcore just for the FSB heck of it.
http://www.abload.de/img/capture7ldz.jpg
Those drive strength options really seem to make a difference. I wish Asus would add some to the MIIF; as Cryptik said, it seems that "Memory OC Charger" may control drive strength but Gigabyte's different profiles are better.
yes but from what i've understood not only vdimm also high drive strength kill ram fast.
;-) so happy pumpin'
first go with 2*1gb Cellshock 1066 c5, 1298ish@2.24v
http://i38.photobucket.com/albums/e1...fy2006/415.jpg
dfi took them out of bios options cause it will kill ram if set too highQuote:
Originally Posted by Shiranui Gen-An
Without those options how can one hope to clock memory over 1250MHz without shoving ridiculous amounts of VDIMM into them?Quote:
Originally Posted by zsamz_
So is that possible why my P5E has a lot of trouble getting 1400MHz?Quote:
Originally Posted by Shiranui Gen-An
Edit: YAY I just ordered a UD3P :D I hope i'll be able 2 overclock my RAM better.
when you bench you need to pump in the juice its all dependant of how much juice you ram needsQuote:
Originally Posted by Shiranui Gen-An
for 24/7 you dont need 1250+ just find the sweetspot me i'm content with 11k in everest for 24/7
all my ballistix need 2.06v for 1250mhz n my d9gkx need 2.4v
Thats nice for you having 11k for 24/7, i cant break 11k pumping juice or anything, and its not my memory holding me back, its the gigabyte:down:
Wow pang hits some nice everest on the ud3p gigabyte, hmmmm
The X48's have them afaik, the T2RS does anyway.Quote:
Originally Posted by zsamz_
There is no way to, drive strength = current, vdimm = voltage, without being able to control current, the only way to increase clocks is to pump the voltage. You can play with dram Vtt & Vrefs, and also MCH Vref, but they have minimal effect, just fine tuning.Quote:
Originally Posted by Shiranui Gen-An
Possibly, could also be the chipset, it's hard to say without owning one. I'm interested to see how your UD3P goes, so be sure to post some everest and SPI32M when it comes in :)Quote:
Originally Posted by fatguy1992
From PangingIII's results, it looks like my board is faster in SPI32M, but some more comparisons would be great.
finally able to get ddr1300 stable.
http://i38.tinypic.com/nmjhht.jpg
http://i37.tinypic.com/e5puvk.jpg
Sure, if i'm lucky i'll have it on Friday, but probably Monday.Quote:
Originally Posted by CryptiK
@ radaja nice work there :up:
thanks, fatguy1992.
Very nice, what MCH core voltage did that require? Are you on 1333 drive strength profile? That's pretty low vdimm for 650 MHz stable, especially considering you timings are quite tight.Quote:
Originally Posted by radaja
@cryptik, 1.38nbv.heres a new quik test.i tried with 1333 oc profile and my everest scores were worse.
these are with profile on auto.also memset is reporting the wrong mhz values.
http://i34.tinypic.com/s4m520.jpg
http://i34.tinypic.com/330ryw8.jpg
OMG yes I want them, plz. I pm'ed u.Quote:
Originally Posted by VOID WARRANTY
Have you tried more MCH core voltage when going for 500 FSB on the 5:6 divider?Quote:
Originally Posted by Elektrosadist
For my run I used 1.45v, but I'm not sure what the minimum I'd need for that is, I just guessed and it ram fine. We have different boards but it may be worth trying.
Also from what I've seen of eva2000's benching on the EP45-Extreme (found here http://i4memory.com/f90/gigabyte-ga-...0-review-8339/), it's pretty damn fast for a P45 board. At what appear to be looser ram & chipset settings than my M2F it appears to be about 15 seconds faster. I'd like to see a side by side comparison of the EP45-Exteme and the UD3P, as I have a feeling the UD3P is going to be slower clock for clock, as it appears to be a more FSB oriented board. Perhaps though with the right chipset tweaks, the A34 rev. chipset may even pull ahead.
What is also odd about the gigabyte boards (EP45-Extreme & UD3P) is that they can run PL9 at 1:1. With an Asus board you can only run PL10 minimum. It may just be different tRD nomenclature though, who knows.
cryptik, heres with oc 1333 set. also i just noticed
pangingIII 545fsb run on the previous page and his scores smash mine.
http://i35.tinypic.com/qq39rd.jpg
At 545 FSB and 654 MHz, you are actually pretty close to his scores. Your write is a little lower and your copy is lower, it must just be certain chipset settings you have different.
I have no idea why setting your Driving Strength Profile to 1333 would alter your everest scores, AFAIK it only affected ram driving strength??
I can just break 11K (11004MB/s) at 540 FSB, 648 MHz @ 5-5-5-15, and 3.78GHz cpu speed. The read would be higher if my cpu would do ~4.32GHz. With some tweaking, your Gigabyte boards are right about where they should be for P45 performance in Everest.
However I'd like to see some more SPI32M runs for a more real world comparison against the Asus platform. Can someone with a UD3P please do a SPI32M at 500 FSB & 600 MHz ram, PL8? (That's if your boards will play nice on 5:6).
@ Radaja - can you please run a SPI32M at your 540 FSB - 648 MHz setting? But can you please use the 7x muliplier for the cpu, I cant do 4050 or higher SPI32M stable.
Quote:
Originally Posted by Elektrosadist
as far as i understand right now, it seems that the different in some of the voltage values that you set in BIOS and the values that show on Easytune6 can cause confusion.
even though some of the voltage values are not exactly the same, but as far as i know, actually, both the BIOS values and the ones on easytune are the same.
sorry for my English, i'll try to explain this later, maybe in the UD3's thread.
anyway, here below an example of a UD3R/E8400 500 FSB 2x1GB 5:6 DDR2-1200 system, i just ran a two instances of SuperPi 32M and took this screenshots.
drive strength is DDR2-1200MHz, memory timing is 5-5-5-15 rest of timing settings are auto's,
this RAM kit is quite old i've been using it for like 2 years now, vdimm required by the modules is now high, so disregard the 2.3v vdimm.
http://img208.imageshack.us/img208/6...3151233er3.png
try your BIOS setting as follow... i'll try and explain it to you, at least the way that i understand it.
what i would like you to check is the "MCH Reference" and "CPU Reference"
CPU Reference... since you're using Quad maybe try to calculate this value so that it's equal to around 67-68% of the normal/standard CPU Termination or VTT (which is 1.20v on this P45 board)...
normal VTT = 1.20v
normal CPU Reference = 0.76v, which is 63.33% of VTT... this value seems to work okay with dual-core, however i sometime find that CPU Reference at 0.785v, which's around 65.5% of VTT, can work better with my E8400s system running on high FSB overclock.
MCH Reference... i find 0.74v is seem to always work with my 500FSB CPUs and 5:6 DDR2-1200. 0.74v is the Normal or standard value, and this voltage value is link to VTT.
for an easy way, once you get into BIOS, 1st thing you do is to set the "CPU Termination" to 1.20v (then you can change this from 1.20v to say 1.36v at the last before you save/exit the CMOS setting. then 2nd setting is the MCH Reference, set it to 0.74v, and the 3rd one is the CPU Reference...
====================
Load-Line Calibration [Enabled]
CPU Vcore [depends on your CPU]
CPU Termination [1.36]
CPU PLL [1.59]
CPU Reference [?????]
MCH Core [1.4]
MCH Reference [?????]
MCH/DRAM Refernce [try setting this to Normal (not Auto)]
ICH I/O [Normal]
ICH Core [Normal]
DRAM Voltage [2.2v. more or less depends on your RAM]
DRAM Termination [Normal]
Channel A Ref [Normal]
Channel B Ref [Normal]
===========================
Quote:
Originally Posted by WaterFlex
Thanks, WaterFlex.
Very nice Radaja, those boards are really something else for ram clocking, 648 MHz 5-5-4-12 @ 2.26v is great. What was tRFC set to? You have a nice cpu too, wish mine would do that.
i know that with the mch and cpu refs. i don't know what it is i can't get FSB 500 stable with 1200MHz. even with loose timings 5-5-5-18 and adjusting the refs.
anyways i don't mind so much FSB 500 and DDR2 1250 works so im good.