[Thundergod989]

So far i've only been able to get it to 2.83 stable at 1.5Vcore and 1:1 memory speed running at cl2, i can get a bit higher at cl2.5 but i get less game performance due to increased latency.

Anybody else have relatively good success with their 3700's?

(i'm running on air only btw, but i'd like to see what ppl were able to do with water...)

 

[GatorLCA]

I know you can get to 3.0, depending on the setup with air, I've been able to get mine to about 2.75 with everything stable...i really dont see the need to get that much faster, you won't see that much of a gain if you go higher. I would just keep it where you have it, unless you have a WC setup.

 

[ThaWaxShop]

http://www.overclock.net/amd-cpus/67...ng-thread.html
heres my progress

 

[Thundergod989]

ya i could probably touch 3 with cl2.5 seeing as i was at 265x11 already. I'll see whats the max fsb i can get with the ram tuned down to 3:4 and CL3 with the multi at 9 or 10. Should be able to get around 3.1 or 3.2 (310/320 fsb) i think realistically.

My new personal challenge is devising a way of cooling my northbridge and trying to hit 3.0 with CL2 without water.

I'll keep you guys posted.

 

[tytlyf]

Quote:

Originally Posted by Thundergod989 trying to hit 3.0 with CL2 without water.

that wont happen....
you shouldn't be trying to find the maximum of your cpu with your memory the way you are running it anyway. Its better to have higher mhz vs tighter timings.
the bandwidth of high mhz cl 2.5 is greater than cl2 at lower speeds....
i want to see some superpi runs on this cl2 memory you claim is running 1:1, or even a screenshot of your memory tab in cpuz....

 

[Melcar]

2.6-2.7 is about average for a 3700+. Really good ones do 2.8-2.9. I have mine at 2.88 with 1.65V (won't go any higher even if I increase voltages). I'm currently experimenting with lower voltages and see how it goes.

 

[Thundergod989]

Quote:

Originally Posted by tytlyf that wont happen.... you shouldn't be trying to find the maximum of your cpu with your memory the way you are running it anyway. Its better to have higher mhz vs tighter timings. the bandwidth of high mhz cl 2.5 is greater than cl2 at lower speeds.... i want to see some superpi runs on this cl2 memory you claim is running 1:1, or even a screenshot of your memory tab in cpuz....

Bandwidth isn't everything on memory, access time and queue time is quite important when doing quick iterations. Either way when i'm done work for the week i'll have some time to delve into it a bit more and give you some test results (and screenshots for you skeptics).

 

[ThaWaxShop]

Quote:

Originally Posted by Melcar 2.6-2.7 is about average for a 3700+. Really good ones do 2.8-2.9. I have mine at 2.88 with 1.65V (won't go any higher even if I increase voltages). I'm currently experimenting with lower voltages and see how it goes.

not with the new steppings. the kab2e and newer avg is 2.9-3.0 on air look around the forum the new 3700 steppings are nice

 

[Thundergod989]

Quote:

Originally Posted by ThaWaxShop not with the new steppings. the kab2e and newer avg is 2.9-3.0 on air look around the forum the new 3700 steppings are nice

Haha i stirred up some sh** with the retailler i bought my proc from, I had them open the box and check the identifier on the cpu before they shipped it. I did teh same thing back in the day when hunting around for my 1700+ thoroughbred (lots of palmino's running around back then). I've heard some pretty dissapointing figures coming from the 3500+ (venice core w/512k L2), like caps at 2.6 and less.

As far as what i've seen, they clocked a 3700+ to 2.94 on air and ran it stable at 2.5-3-3-7 and 1.45V on one review i read ( on a dfi tho ;0 ).

Concerning the Cl2 @ 3.0, i don't know if it'll happen and if it happens it'll be the first i've seen on air. I'm gonna test my ram sticks individually with lower multipliers and higher fsb to see where exactly they give out each, then label them so I know which is the weaker. Also I have a friend with a very nice pair of mushkins that he's tested to 272 at cl2.0, so if my ocz platinum el's wont do it then i'll try it with them (thats only 2992mhz tho, lol) .

read up ;0
http://www.phoronix.com/scan.php?pag...item=219&num=1

 

[ThaWaxShop]

Quote:

Originally Posted by Thundergod989 Haha i stirred up some sh** with the retailler i bought my proc from, I had them open the box and check the identifier on the cpu before they shipped it. I did teh same thing back in the day when hunting around for my 1700+ thoroughbred (lots of palmino's running around back then). As far as what i've seen, they clocked a 3700+ to 2.94 on air and ran it stable at 2.5-3-3-7 and 1.45V on one review i read ( on a dfi tho ;0 ).

what stepping did you get??

 

[Thundergod989]

Quote:

Originally Posted by ThaWaxShop what stepping did you get??

E4 i think, but i'll check it when i get home, i'm gonna get some ac5 and change the 80MM stock fan for my 80mm xaser11 with the rheobus.

edit: its a DKA5CF so whichever stepping that is., cpu-z reports it as SH-E4

 

[Thundergod989]

Quote:

Originally Posted by tytlyf that wont happen.... you shouldn't be trying to find the maximum of your cpu with your memory the way you are running it anyway. Its better to have higher mhz vs tighter timings. the bandwidth of high mhz cl 2.5 is greater than cl2 at lower speeds.... i want to see some superpi runs on this cl2 memory you claim is running 1:1, or even a screenshot of your memory tab in cpuz....

this was as far as i could get it and I had to boot 3 times into windows before it stopped giving me blue screens (had to inch the chipset voltage up .1V to get it to stay stable for a pi run). Gonna change my hsf for the xaser and see how she runs then, hopefully i'll be able to push 3, but i'm sure my ram will crap out as it is already hardly usable at 272, i only got into windows once and not long enough to run anything.

 

[Zardokk]

I can hit 2.8 stable with stock vcore. I can hit 3.1 not-quite-stable with a 1.6vcore.

 

[Thundergod989]

Quote:

Originally Posted by Zardokk I can hit 2.8 stable with stock vcore. I can hit 3.1 not-quite-stable with a 1.6vcore.

ya the sandy's do very well on stock vcore. What is your memory timing at 2.8? and 3.1?

 

[Zardokk]

My mem timings (this is with Value Select, remember) are 2-3-3-5 at 2.8, mem running at 195MHz. At 3.1, I have to up mem vcore from 2.6 to 2.8 and lower timings to 3-3-3-8, and even then it's not stable. But mem runnings pretty high freq. when CPU is at 3.1. Can't remember exactly how high, but this value RAM will actually go up to 225MHz with loose timings. With 2-3-3-5 timings, the best I can do is about 205MHz before things crash.

 

[Thundergod989]

Quote:

Originally Posted by Zardokk My mem timings (this is with Value Select, remember) are 2-3-3-5 at 2.8, mem running at 195MHz. At 3.1, I have to up mem vcore from 2.6 to 2.8 and lower timings to 3-3-3-8, and even then it's not stable. But mem runnings pretty high freq. when CPU is at 3.1. Can't remember exactly how high, but this value RAM will actually go up to 225MHz with loose timings. With 2-3-3-5 timings, the best I can do is about 205MHz before things crash.

oh man i read 195 as 295 i was like holy crap on valueselect?!?!? ***

 

[Thundergod989]

oh man my htt was at 800 instead of 1000, so i set it manually and it gives a noticable difference. System now runx identically to an fx-57 in both arithmatic and multimedia benchmarks in sandrasoft bench and the memory kills the ddr2 pc5300 on the int and float bandwidth. Shaved about 2 seconds off the superpi.

 

[Tobuk]

i hold mine at 3Ghz all the time. it held 3.1Ghz for a 1Mb superPI run, but that's about all i could get on my Zalman 9500...

 

[tytlyf]

those times are really slow thunder, something is up, you should be well into the 28 second range with the clock and the memory timings....your system is struggling to perform, turn off any unnecessary programs while benching..
I've never seen those OCZ el's run cl2 on any system at that speed. 2T doesn't count

, that's probably your problem and why your times are slow...
never run 2T, 1T at all costs..

 

[Thundergod989]

Quote:

Originally Posted by tytlyf those times are really slow thunder, something is up, you should be well into the 28 second range with the clock and the memory timings....your system is struggling to perform, turn off any unnecessary programs while benching.. I've never seen those OCZ el's run cl2 on any system at that speed. 2T doesn't count , that's probably your problem and why your times are slow... never run 2T, 1T at all costs..

its at 1T in the bios, but it reports as 2T in most programs except sandrasoft.

I remeber reading something about my board having issues with 1T and tccd etc. Also when i put the memory on 4:5 it reports as higher in the bios than it does in cpu-z, so something seems to be up with that. Also check out my cpu vcore, .032 volts... either my board is running mythical efficiencies or the reporting is off a bit.

Btw i was running bitcomet at the same time I ran the superpi.

There is also something fishy with my scsi320 card it has a monitoring program in windows and it almost always uses 2% cpu.