Trying to OC my 1100T

[Jekerdud]

Well, stock for the HT link and CPUNB are auto on this board, is there a specific number I should put in?

yeah, the memory is stable

 

[trents]

Well, stock for the HT link and CPUNB are auto on this board, is there a specific number I should put in?

yeah, the memory is stable

If you're going to be overclocking with the FSB then let those two frequencies rise as they will up to about 2800 mhz. You will likely need to supplement the CPUNB voltage to the range of about 1.225 to 1.3 volts.

 

[Jekerdud]

Ok, but I don't know what a good starting frequency is for the CPU-NB and HT. As I said, "stock" for my motherboard/bios is having it all set to auto, and I don't know what a good starting number is for setting it manually. I tried googling a good starting number, but my luck wasn't good finding a good starting point.

It is a drop down list for both the HT and CPU-NB, not manual input, so I'd have to go with the closest number.

 

[trents]

A good starting frequency would be 2000 mhz or 10x.

 

[trents]

A good starting frequency would be 2000 mhz or 10x. That's stock.

 

[Jekerdud]

I had to set the CPU-NB and HT to 1800. They both stay the same rate on this board (don't know if that's true for all boards). The whole system wouldn't start up when I set the CPU-NB and HT to 2000. I had to set the FSB to 250 to get the RAM back to 1666, and doing that put the CPU-NB and HT at 2500. So, we now know my system doesn't like one of those being above 2500 (CPU-NB was at 2460 when the FSB was at 205).

If you're going to be overclocking with the FSB then let those two frequencies rise as they will up to about 2800 mhz. You will likely need to supplement the CPUNB voltage to the range of about 1.225 to 1.3 volts.

As for the CPU-NB voltages, auto had it set on 1.35, I left it alone for now since I am doing my 4 gig test, but should I lock it into 1.35 when I increase the multiplier again?

 

[RGone]

but should I lock it into 1.35 when I increase the multiplier again? Test it and see if it needs that voltage.

 

[Jekerdud]

I locked it at 1.275 for now. So far, it hasn't needed an increase.

I have noticed that the tests in Prime95 are going faster! The iterations are finishing quicker and moving on to the next. I LIKE IT! Before, it was going maybe 1 minute, 30 seconds to up to 3 minutes between iterations, now it's down to a minute or less per.

 

[Jekerdud]

Well, I think I am done! The settings are below. I tried turning down the Vcore to 1.4225, and it a worker did fail two times, so I set it back up to 1.45 and decided to stick with it since two separate two hour tests ran just fine with the FSB, multiplier, and Vcore at the current settings.

If there is anything you guys think I need to adjust still (besides my fans ), let me know. Otherwise, thanks for the help!

 

[RGone]

Would have been nice to have HWMonitor running while doing P95 Blend tests so the Min/Max Temps and volts were actually logged. Then the capture of HWMonitor would give more detailed information. That capture is only showing idle information.

Otherwise it looks pretty good.

 

[walldow]

everything looks great from what i see . only thing i would do is set you HT and NB divider to 10 instead of 9 for 2500 mhz HT. the 990fx runs 2600mhz HT without even adjusting the voltages. if you realy wana see what it will do change divider too 11 for 2750mhz (might have to bumps the ht link volage and cpu-nb voltage just a tad) i run 2700mhz fsb 24/7 and can run 2800mhz for benches. i run all air zalmann cpu cooler and antec 900 case and temps on the cpu bearly ever top 65+ unless im just all out benching but runing normal games and apps 24/7 2700mhz is fine on the 990fx! nice job on the OC grats

 

[keny]

There are no real gains to be had from overclocking the HT and I'm sure the HT only runs at 2600 stock when a FX CPU is in the board, either way its not going to gain you nothing from overclocking it, where the CPU NB will bring gains in mem performance.

 

[walldow]

There are no real gains to be had from overclocking the HT and I'm sure the HT only runs at 2600 stock when a FX CPU is in the board, either way its not going to gain you nothing from overclocking it, where the CPU NB will bring gains in mem performance.

2000mhz (4000mhz) vrs 2600mhz (5200mhz) =1200mhz effective bandwith. a 1200mhz increase of effective bandwith to the cpu is going to improve any app game bench realworld ect...

 

[keny]

Trust me there's NO gains from increasing the HT over stock

 

[walldow]

Trust me there's NO gains from increasing the HT over stock

ok guess my cpu is differant then evryone elses i just ran sandra on processor arithmatic and only thing i changed was the HT from 2000mhz to 2600mhz. on both test the NB was kept at 2600mhz

my 1100T @3700mhz HT 2600mhz cinebench cpu 6.78 single core 1.14
Aggregate Architecture Performance 72 GOPS
Drystone ALU 79 GIPS
whetstone iSSE3 65 GFLOPS

my 1100T @3700mhz HT 2000mhz cinebench cpu 6.62 single core 1.10
Aggregate Architecture Performance 71.4 GOPS
Drystone ALU 77.82 GIPS
whetstone iSSE3 65 GFLOPS


needless to say they are not threw the roof numbers but a GIP is a GIP and a GFLOP is a GFLOP and over a matter of hours the little fraction adds up pretty nice. ill take the extra edge of running the higher HT even if is just a tid bit

 

[keny]

I bet if you ran it again you would get a different result, as those numbers are so close and that performance difference if it stuck would be no real world gain in performance, but I would hazard a guess as they would fluctuate if run again. I don't want to argue this but if you go to the link and look at the top results on hwbot for you're CPU you will see that most if not all of them are running around stock HT or below, for some reason it just doesn't seem to help, maybe it does in Sandra but nothing else that I know of

http://hwbot.org/hardware/processor/phenom_ii_x6_1100t_be/

 

[walldow]

I bet if you ran it again you would get a different result, as those numbers are so close and that performance difference if it stuck would be no real world gain in performance, but I would hazard a guess as they would fluctuate if run again. I don't want to argue this but if you go to the link and look at the top results on hwbot for you're CPU you will see that most if not all of them are running around stock HT or below, for some reason it just doesn't seem to help, maybe it does in Sandra but nothing else that I know of

http://hwbot.org/hardware/processor/phenom_ii_x6_1100t_be/

ya those guys are runing 4500mhz nb too. guess evry chip has its own sweet spot i know ive benched mine evry witch way but inside out. i have 2100mhz ramm and have run 200-225-250ish with higher nb freqencies and lower ht but for some reason even with the ram runing at only 900 mhz like in sig it runs faster than runing 1000mhz on lower ht/fsb. could be just the combo of ram and chip? all i know is i get beter performance out of mine runing higher ht than lower ?

 

[keny]

Yeah every chip is different, if it works for you then that's all good, if you are running that ram at 2100 you are doing well as my old thuban got naughty at anything over 20000 on the ram and then it was not 24/7 at 2000 as prime would have crashed it within a breath at 2000, but the NB on these thubans is pretty good and makes a good jump on performance when that's clocked up

 

[Jekerdud]

Well, when I increased the HT, I noticed that iterations in Prime95 seemed to be completing faster, so there was an increase for me anyways.

 

[keny]

P95 runs at different speeds throughout the tests, I've said what I know and provided a link to figures but if you want to run it above stock it's you're call.