Another stubborn X4 945 (125W) C2 Overclocker

[Anonaru]

The "results.txt" log says the other workers were on the 160k self test. I've long since closed that Prime95 screen. I tried last night to add CPU-NB voltage and going to 1.325V caused 2 to fail immediately (rounding error) so I thought maybe I was over volting, and going to 1.275V caused 2 to fail immediately (rounding error) as well. I've knocked the memory multiplier down so the frequency is 1020 and I'll re-run the tests tonight with the CPU-NB voltage back at 1.3V.

I've also knocked the DRAM voltage down from 1.650V to 1.6V (default). Six hours isn't too bad though, I flew from Laguardia (averaging about 20 frames per second in the cockpit and 25-30 on the exterior) to Orlando International last night (averaged 25-30 in cockpit and 40 on the exterior) and the new clock speed has made it at least playable. I'm going to continue to tweak the game as well as the clock speed. I didn't think I would get this far and even be 6 hour stable in Prime95 so I can't really complain!

Sounds like a IMC trip up in my opinion. Voltage becomes touchy quick on these Deneb parts. I'm thinking the NB frequency and RAM frequency was fine-- I'm thinking the RAM should be at something like 1.625 and test with the timings loosened one click, just to isolate for sure if it was the RAM or not.

 

[Whiskey11]

Sounds like a IMC trip up in my opinion. Voltage becomes touchy quick on these Deneb parts. I'm thinking the NB frequency and RAM frequency was fine-- I'm thinking the RAM should be at something like 1.625 and test with the timings loosened one click, just to isolate for sure if it was the RAM or not.

I'm going to run this test first to make sure it isn't frequency related first. This memory is stupid sensitive to timing changes. I'm not sure if I'm just not picking the correct loosened up timings or what, but they don't like 9-9-9-24 at all (wont boot due to "boot failures caused by overclocking") and RAM timing is extremely foreign to me so that makes it worse. I'm going to find a RAM "overclocking" guide to see if I can't understand a little clearer what all those timings can cause in terms of stability problems and Prime95 failures

 

[Whiskey11]

Sounds like a IMC trip up in my opinion. Voltage becomes touchy quick on these Deneb parts. I'm thinking the NB frequency and RAM frequency was fine-- I'm thinking the RAM should be at something like 1.625 and test with the timings loosened one click, just to isolate for sure if it was the RAM or not.

Double posting... sorry...

I tried my test to see if it was the ram frequency and after 6 hours and 23 minutes, worker 3 got a rounding error again. Time to go play with RAM timings and uncover the mystical ancient Egyptian hieroglyphs that are RAM timings! Wish me luck! It's a shame that Prime95 has to run for 6 hours and 23 minutes in order to get to that point...

I also changed my fan configuration a little bit. I added my "spare" 120mm to the radiator so one side is push/pull and I installed two of my 140mm's in the drive space between my DVD drive and the start of the HD trays, and another sitting on the ledge around the PSU area by the HD trays. My rear exhaust fan lost a fan blade somehow so I took it out. Results, well Prime95 ran for 6hrs and 23 minutes and recorded a max temp of 50ºC so something is working right! I also forced my GPU fan to run at 100% because it was idling around 60ºC and getting to almost 80ºC in gaming... at 100% fan speed it idles around 48ºC and is lucky to get to 60ºC in gaming. Seems like it should help keep that from burning up. Fan noise has increased a little bit but it is all white noise, no whine so I'm over it already.

EDIT: Well. Prime95 failed last night at 4hrs and some odd minutes of running. Same rounding error different core. I'm currently testing the JDEC #3 timings per trents recommendation on the first page (8-8-8-22-30) with the DRAM volts to 1.65V (this is the first step up from the 1.6V default) and the frequency up at 1359MHz again. I also upped the HT-Link up one multiplier so it is at 2295MHz. Prime95 is running. I'm hoping that I can get it stable at this speed as that would be awesome!

 

[Anonaru]

Double posting... sorry...

I tried my test to see if it was the ram frequency and after 6 hours and 23 minutes, worker 3 got a rounding error again. Time to go play with RAM timings and uncover the mystical ancient Egyptian hieroglyphs that are RAM timings! Wish me luck! It's a shame that Prime95 has to run for 6 hours and 23 minutes in order to get to that point...

I also changed my fan configuration a little bit. I added my "spare" 120mm to the radiator so one side is push/pull and I installed two of my 140mm's in the drive space between my DVD drive and the start of the HD trays, and another sitting on the ledge around the PSU area by the HD trays. My rear exhaust fan lost a fan blade somehow so I took it out. Results, well Prime95 ran for 6hrs and 23 minutes and recorded a max temp of 50ºC so something is working right! I also forced my GPU fan to run at 100% because it was idling around 60ºC and getting to almost 80ºC in gaming... at 100% fan speed it idles around 48ºC and is lucky to get to 60ºC in gaming. Seems like it should help keep that from burning up. Fan noise has increased a little bit but it is all white noise, no whine so I'm over it already.

EDIT: Well. Prime95 failed last night at 4hrs and some odd minutes of running. Same rounding error different core. I'm currently testing the JDEC #3 timings per trents recommendation on the first page (8-8-8-22-30) with the DRAM volts to 1.65V (this is the first step up from the 1.6V default) and the frequency up at 1359MHz again. I also upped the HT-Link up one multiplier so it is at 2295MHz. Prime95 is running. I'm hoping that I can get it stable at this speed as that would be awesome!

If it still drops on you around the 4-6 hour mark, then try once more with a more substantial RAM bump, like 1.675. The Deneb parts are designed to be compatible with 1.8v DDR2 RAM, so don't concern yourself too much-- the RAM can take it too (Just don't do something crazy like push it to 1.9v)

I still think it is the IMC rather than the RAM-- Will be awaiting your results

 

[Whiskey11]

If it still drops on you around the 4-6 hour mark, then try once more with a more substantial RAM bump, like 1.675. The Deneb parts are designed to be compatible with 1.8v DDR2 RAM, so don't concern yourself too much-- the RAM can take it too (Just don't do something crazy like push it to 1.9v)

I still think it is the IMC rather than the RAM-- Will be awaiting your results

Unfortunately the DRAM voltage goes up in .050V increments so the next step up is to 1.7V's. If it's the IMC rather than the RAM would this require more vcore voltage or more CPU-NB or less and a declock or? Sorry I'm a bit dense in this sense. I'm kind of guessing what needs to be changed at this point! Obviously there is plenty of heat removal with the current fan setup. I'm 2 hours into prime and max recorded core temp is 50ºC but averaging around 45º-46ºC. I would say that I don't think this little C2 can go much higher then where it is at! I could be wrong, it would be worth playing with once I get this OC stable because I'm crazy like that!

 

[Anonaru]

Unfortunately the DRAM voltage goes up in .050V increments so the next step up is to 1.7V's. If it's the IMC rather than the RAM would this require more vcore voltage or more CPU-NB or less and a declock or? Sorry I'm a bit dense in this sense. I'm kind of guessing what needs to be changed at this point! Obviously there is plenty of heat removal with the current fan setup. I'm 2 hours into prime and max recorded core temp is 50ºC but averaging around 45º-46ºC. I would say that I don't think this little C2 can go much higher then where it is at! I could be wrong, it would be worth playing with once I get this OC stable because I'm crazy like that!

Alright, well set your RAM to what it was at when you got your first 6 hour stable run..

Then tell me, what is your NB set at again? CPU-NB Frequency and voltage.

 

[Whiskey11]

Alright, well set your RAM to what it was at when you got your first 6 hour stable run..

Then tell me, what is your NB set at again? CPU-NB Frequency and voltage.

Alright, the above test failed again at 4 hours. I've bumped the timings back to the 6 hour timings (8-8-8-21-27 timings) and upped the RAM voltage to 1.7V. I'm determined to make it stable demmit!

Anyway here are the stats in text form:
CPU Frequency: 3.825GHz
RAM Frequency and Timings: 1359 MHz @ 8-8-8-21-27
FSB: 255 (15x Multiplier)
CPU-NB Frequency: 2550MHz @ 10x Multiplier (max)
HT-Link Frequency: 2295MHz @ 9x Multiplier (one below max)
CPU Voltage: 1.500V
CPU NB VID Control: 1.300V
DDR3 Volt: 1.650V (now at 1.700V)
HT-Link Voltage: Auto
CPU Core Temp never got about 52ºC, no idea what the socket temp was at but it has been about 4º-5ºC cooler than the core temp.

Did I get all the info you needed??

Prime95 Blend running again with the 1.700V DDR3 voltage.

 

[Anonaru]

Alright, the above test failed again at 4 hours. I've bumped the timings back to the 6 hour timings (8-8-8-21-27 timings) and upped the RAM voltage to 1.7V. I'm determined to make it stable demmit!

Anyway here are the stats in text form:
CPU Frequency: 3.825GHz
RAM Frequency and Timings: 1359 MHz @ 8-8-8-21-27
FSB: 255 (15x Multiplier)
CPU-NB Frequency: 2550MHz @ 10x Multiplier (max)
HT-Link Frequency: 2295MHz @ 9x Multiplier (one below max)
CPU Voltage: 1.500V
CPU NB VID Control: 1.300V
DDR3 Volt: 1.650V (now at 1.700V)
HT-Link Voltage: Auto
CPU Core Temp never got about 52ºC, no idea what the socket temp was at but it has been about 4º-5ºC cooler than the core temp.

Did I get all the info you needed??

Prime95 Blend running again with the 1.700V DDR3 voltage.

You may need to set your NB to 2295 as well.. Hopefully you don't need to do this. If it is the IMC vs the RAM then this will most likely be your only recourse It may end up being something a little more simple, like you have to back off your FSB 1MHz or something

 

[Bobnova]

With a BIOS update that board'll do the X6 chips too. It's a pretty good board.

Don't be afraid to go to 1.55v CPU if the core temps are good.

 

[Anonaru]

With a BIOS update that board'll do the X6 chips too. It's a pretty good board.

Don't be afraid to go to 1.55v CPU if the core temps are good.

Haha, I'm always a whimp suggesting over 1.5v.. I can't imagine it being the vCore at this point, but an additional bump is always worth a shot

 

[Whiskey11]

You may need to set your NB to 2295 as well.. Hopefully you don't need to do this. If it is the IMC vs the RAM then this will most likely be your only recourse It may end up being something a little more simple, like you have to back off your FSB 1MHz or something

Well, I'll keep on tweaking and hopefully I wont have to drop the NB Frequency at all. As for backing the FSB off 1 MHz... well... I'd like it to go UP not down! Tweaking will continue until I'm left with ZERO other options!

With a BIOS update that board'll do the X6 chips too. It's a pretty good board.

Don't be afraid to go to 1.55v CPU if the core temps are good.

Good to know about supporting the X6 chips. I will keep this in mind should I decide not to go with an Intel chip (it's hard to leave AMD!).

As for the CPU voltage, is this something that could help remove my rounding error issue either through supporting the IMC better or making the CPU itself more stable? Obviously I've got some room temperature wise to up the voltage but I don't really want to do it unless I can confirm that upping the voltage will result in a more stable overclock rather then added heat that I have to find a way to remove from the water and case!

 

[Anonaru]

Well, I'll keep on tweaking and hopefully I wont have to drop the NB Frequency at all. As for backing the FSB off 1 MHz... well... I'd like it to go UP not down! Tweaking will continue until I'm left with ZERO other options!




Good to know about supporting the X6 chips. I will keep this in mind should I decide not to go with an Intel chip (it's hard to leave AMD!).

As for the CPU voltage, is this something that could help remove my rounding error issue either through supporting the IMC better or making the CPU itself more stable? Obviously I've got some room temperature wise to up the voltage but I don't really want to do it unless I can confirm that upping the voltage will result in a more stable overclock rather then added heat that I have to find a way to remove from the water and case!

For giggles: Without changing anything else, change your CPU-PLL voltage to 2.545 and run it. I haven't heard about any relationship tests done showing what effect PLL has on stability on these Deneb parts.. It is a safe jump to voltage so hey, why not?

 

[Whiskey11]

For giggles: Without changing anything else, change your CPU-PLL voltage to 2.545 and run it. I haven't heard about any relationship tests done showing what effect PLL has on stability on these Deneb parts.. It is a safe jump to voltage so hey, why not?

It is? I had to go look at what that even was! I'll keep it in mind for if/when this test fails!

EDIT: Didn't need attached photo as I figured out which one you were talking about!

 

[Anonaru]

It is? I had to go look at what that even was! I'll keep it in mind for if/when this test fails!

EDIT: Didn't need attached photo as I figured out which one you were talking about!

Changing PLL voltage was very relevant in the older days, and INCREASING it on AMD parts back then was a good deal. Lowering it can decrease temps, but this can potentially hurt stability. I've been able to repeat positive results when increasing PLL voltage on the newer AMD parts (Namely Vishera chips) in order to acheive a higher, stable overclock with less vCore.

LOWERING PLL voltage on intel rigs gives you more thermal headroom, and, in my case, a stable OC at 4.9~GHz without an increase to vCore from 4.8.

I'd give it a shot next, it will essentially micro-bump your VTT / VDDA voltages (Which most AMD boards don't even allow you to touch) and give you an insignificant bump to vCore and vNB. It is a safe voltage to toy around with in the 2.385-2.650 range. Safe to go above 2.695 on vishera parts, but I never do anyways.

 

[Whiskey11]

Changing PLL voltage was very relevant in the older days, and INCREASING it on AMD parts back then was a good deal. Lowering it can decrease temps, but this can potentially hurt stability. I've been able to repeat positive results when increasing PLL voltage on the newer AMD parts (Namely Vishera chips) in order to acheive a higher, stable overclock with less vCore.

LOWERING PLL voltage on intel rigs gives you more thermal headroom, and, in my case, a stable OC at 4.9~GHz without an increase to vCore from 4.8.

I'd give it a shot next, it will essentially micro-bump your VTT / VDDA voltages (Which most AMD boards don't even allow you to touch) and give you an insignificant bump to vCore and vNB. It is a safe voltage to toy around with in the 2.385-2.650 range. Safe to go above 2.695 on vishera parts, but I never do anyways.

My computer has a SICK sense of humor. With DDR3 voltage up to 1.7V and Prime95 goes 7 hours and 6 minutes then pops up with a "Rounding was 0.5, expected less than 0.4 error. It also had a "Hardware failure detected, consult stress.txt file." attached to it but it has said that plenty of times before but I don't think anything is wrong with any of my hardware. Surely a fault in the memory or CPU would not make it 7 hours running full bore and then fail randomly right?

I will give your CPU-PLL voltage increase a shot tonight. I'm crashing for the evening. I hope that when I wake up in the morning this issue will go away! Either way, it appears that increasing memory voltage had SOME impact in the length of time the test runs before failing. In this case, 0.050V's resulted in a half hour of increased run time in Prime95. At that rate it will only take .25 more V in the DDR3 to get 12 hour stability right?

Load temperature max was 51ºC on the cores and 47º/48ºC on the socket (TMPIN1 or TMPIN2, both change more than a few degrees, neither got above 48ºC).

EDIT: I ran the increased CPU-PLL last night at and it failed at about 3:30 this morning according to the logs. I unfortunately DON'T have the Prime95 window open since Windows Update restarted my computer last night... I then restarted the computer, increased RAM voltage to 1.75V (getting nervous about this really) and the CPU Vcore to 1.55V and re-running the tests. If I get home tonight and I still have the rounding error, I may give up on trying for the long Prime95 stability testing. Seven hours is pretty stable. The only game I play that comes close to maxing out the CPU is FSX and it only ever maxes out one core fully, two at 50%-70% and the fourth almost none.

 

[Anonaru]

My computer has a SICK sense of humor. With DDR3 voltage up to 1.7V and Prime95 goes 7 hours and 6 minutes then pops up with a "Rounding was 0.5, expected less than 0.4 error. It also had a "Hardware failure detected, consult stress.txt file." attached to it but it has said that plenty of times before but I don't think anything is wrong with any of my hardware. Surely a fault in the memory or CPU would not make it 7 hours running full bore and then fail randomly right?

I will give your CPU-PLL voltage increase a shot tonight. I'm crashing for the evening. I hope that when I wake up in the morning this issue will go away! Either way, it appears that increasing memory voltage had SOME impact in the length of time the test runs before failing. In this case, 0.050V's resulted in a half hour of increased run time in Prime95. At that rate it will only take .25 more V in the DDR3 to get 12 hour stability right?

Load temperature max was 51ºC on the cores and 47º/48ºC on the socket (TMPIN1 or TMPIN2, both change more than a few degrees, neither got above 48ºC).


EDIT: I ran the increased CPU-PLL last night at and it failed at about 3:30 this morning according to the logs. I unfortunately DON'T have the Prime95 window open since Windows Update restarted my computer last night... I then restarted the computer, increased RAM voltage to 1.75V (getting nervous about this really) and the CPU Vcore to 1.55V and re-running the tests. If I get home tonight and I still have the rounding error, I may give up on trying for the long Prime95 stability testing. Seven hours is pretty stable. The only game I play that comes close to maxing out the CPU is FSX and it only ever maxes out one core fully, two at 50%-70% and the fourth almost none.

7 hours is probably "Good enough" You're at the very top for your voltages, if this doesn't work then I'd say either chance it or set it back a FSB tick.

Really 1.525 for 24/7 running would be as high as I'd go personally. Your RAM is probably at its absolute maximum too (Again, in my opinion)..

 

[Whiskey11]

I definitely agree that any higher on voltages is really being on that edge. A seven our even the six and a half hour prime settings were good enough to play FSX without crashing the computer and that is really the goal for right now. Maybe if I end up with a new motherboard and CPU I will crank it up and really abuse this poor C2 to the point of failure. Again I'm just happy that it is as stable as it is for as long as it is at the speed that it is. It really has been a good little chip.

EDIT: Yeah, looks like that didn't work either. 6hrs and 17 minutes and rounding error, but it was much closer, instead of being over .5 it was .4921875! It sounds to me like it could be related to memory, so either the issue is with a stick of memory (Maybe I should go get Memtest and run it? ) or maybe something related to how the processor responds to memory input. Maybe dropping the CPU-NB Frequency will help?

 

[Whiskey11]

Decided to do some Cinebenching and of course an official CPU-Z Validation

Here is the CPU-Z Validation, for no other reason than because I can:
http://valid.canardpc.com/2614786

Here is the Cinebench result is attached. Score is a 4.47. No idea if that is good or not for this ole C2!

 

[Anonaru]

Decided to do some Cinebenching and of course an official CPU-Z Validation

Here is the CPU-Z Validation, for no other reason than because I can:
http://valid.canardpc.com/2614786

Here is the Cinebench result is attached. Score is a 4.47. No idea if that is good or not for this ole C2!

I got 4.49 on my 955 C2 at 3.8, but my NB was set at nearly 2800. You're right about where you should be, the highest I've ever gotten a C2 to finish cine at was 4.57, and that was above 4GHz (Couldn't hold stable though). Good show

 

[Whiskey11]

I got 4.49 on my 955 C2 at 3.8, but my NB was set at nearly 2800. You're right about where you should be, the highest I've ever gotten a C2 to finish cine at was 4.57, and that was above 4GHz (Couldn't hold stable though). Good show

Good to know! Glad I am in the ballpark for other similar chips! I shall dub my CPU "The Little C2 that Could".