Overclock.net › Forums › AMD › AMD CPUs › Phenom II x6 1090t BE overclocked at 3.8Ghz
New Posts  All Forums:Forum Nav:

Phenom II x6 1090t BE overclocked at 3.8Ghz - Page 3

post #21 of 61
Ok finally got to my rig so I can post. (Hate trying to post from my phone).

This is kind of a quick and dirty explanation:

Anyway, quarter breed, you've tried one method of overclocking --raising the multiplier (the noob way IMO). While sure raising the multi is easy, it also doesn't give as good overclocks.

Here's the first things you need to do. First and foremost, run memtest. Let it run for at least two FULL passes (all tests). If you get no errors, then you're ready to move on the next step. Obviously, if you get errors, locate which stick is the one causing it and replace (RMA) as necessary.

The FIRST thing you should do when overclocking, IMO, is find the FSB limitation of your board. Raise the FSB 10MHz at a time and run Prime95 for ~10minutes to check for stability.
Now keep in mind when you raise the FSB it also raises the frequency (speed) of the CPU-NB, HT Link, and the RAM. So adjust those 3 settings either slightly below or to stock speeds. (CPU-NB 2000MHz, HT Link 2000MHz, RAM 1333MHz).

Repeat those steps until you find the max your board can run at. If you BSOD add a little voltage to the NB (NOT CPU-NB).

Once you find the max, then move on to the multi (since you've already done this you can skip it).

Now at this point is where you'll combine the two. You'll notice that if you increase the FSB and the mutli you can get a higher overclock with less volts. I'll use my old x3 720BE as an example. When OCing it to 3.6, if I simply raise the multi to 18x and up the volts, I'll be hard pressed getting it Prime stable even @ 1.55v. However, if I up my FSB to 225, and raise the multi to 16x it's 24hrs prime stable @ 1.46v.

At any rate, once you find what's a reasonable limit (for you), let it prime for several hours. You want to be sure that it's actually stable. At this stage you don't need to run it for 24hours. (Be sure to write down all of your settings)

So now lets say it's 6 hours stable. That's pretty good overall and you really shouldn't expect any errors at this point. Now you can move on to the RAM.

Personally, I like to find the limits of the RAM first, then move on to the CPU-NB. Same as with the CPU, you can use the dividers or use a combination of dividers and FSB. As with the CPU, upping the FSB will net you better overall results.

NOTE: Making sure your memory and CPU-NB is 100% stable is crucial! These are the ONLY things you OC that can, and will, corrupt your OS.

Couple of RAM tips: 1. never exceed 1.65v. That's all the integrated memory controller is rated for, if you exceed this spec IT WILL degrade. been there, done that.
2. When adjusting your timings follow this formula, (it's one that I've devised but it seems to work pretty well).
CL + tRCD +tRP = tRAS ; tRP + tRAS = tRC.

I personally haven't noticed a bit of stability difference between 1T and 2T command rates with DDR3.

Run memtest to check for errors (2 passes), then run P95 blend.

Once you've found the limit of your RAM, move on to the CPU-NB (be sure you set your RAM back to stock settings)

Pretty much as before, you can up the multi or up the FSB. When finding the initial speed, doing quick tests (30minutes) with Prime95 blend is sufficient.
If you BSOD, try raising your voltage to the CPU-NB a little at a time. When you think you've found the limit, let P95 Blend run for a few hours just to be sure.

Once you find your max CPU-NB, try coupling this with your RAM overclock and see if you can get it stable. This is a balancing act between CPU-NB frequency, RAM frequency and timings. Typically the tighter the timings the lower the RAM frequency. This simply takes trial and error to find the best settings. Most likely you'll need to drop your CPU-NB frequency ~50MHz, your RAM 1 divider and you'll be able to keep some pretty tight timings.
NOTE:AMD CPU's gain A LOT of performance from overclocking the memory settings. It's better to drop your CPU OC 100Mhz and increase all of the memory frequencies as the overall performance will be better.

Now you're ready to put all of this together. This may seem like a daunting task, but it's really not, however it does require patience. The last AMD I OC'd took me 3 full days to get it stable. Obviously, most of this was stress testing. Although, on the up side, I did get a lot of house work done that I had been neglecting. biggrin.gif

Hope this helps.

If you have any questions, don't hesitate to ask as I'm sure I probably have omitted something. wth.gif
Edited by BlackOmega - 2/11/12 at 2:31pm
Black Box
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770k Asus P8Z77-Vpro Evga 780 Classified Crucial Ballistix 
Hard DriveHard DriveOptical DriveCooling
A-Data 128GB SSD Samsung Spinpoint F3 1TB Pioneer BDR-207DBK Corsair H80 
OSMonitorKeyboardPower
Win 7 pro 64 Hannspree 25" 1080p LCD 2ms Razer Lycosa PC Power & Cooling 760w 
CaseMouseMouse PadAudio
Corsair 550D MX 518 Allsop Creative XFi Xtreme music 
  hide details  
Reply
Black Box
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770k Asus P8Z77-Vpro Evga 780 Classified Crucial Ballistix 
Hard DriveHard DriveOptical DriveCooling
A-Data 128GB SSD Samsung Spinpoint F3 1TB Pioneer BDR-207DBK Corsair H80 
OSMonitorKeyboardPower
Win 7 pro 64 Hannspree 25" 1080p LCD 2ms Razer Lycosa PC Power & Cooling 760w 
CaseMouseMouse PadAudio
Corsair 550D MX 518 Allsop Creative XFi Xtreme music 
  hide details  
Reply
post #22 of 61
I find the fsb method to be a little dated, and it causes a lot more undue stress on the user trying to fiddle with all those different clock changes at once. Its easier to do each multi/divider independently. I support the OP's method, with some modifications. First, find the multi limit of your system, until your happy with voltage/temps for your rig. (working smarter, not harder, is not "noob" for using the multi first) Then, once you are satisfied, you can either return cpu multi to stock, or leave it where you found it to be stable, then commence with the northbridge multi increases, raising voltage to where you are satisfied. Then, once that is done, you can up your ram divider one notch to start. Nearly ALL kits will go up one divider with about .1 to .12v increase. Once you are there, do the typical stress tests (make sure you memtest before booting into OS when messing with memory of course) and your golden. Now, once you find those good levels, you can add a few FSB notches to juice out those last little ounces of speed. This is way easier to attain the SAME SPEEDS over the "non noob FSB method" I suppose once you find your limitations on each component via the multi/divider method, you can then go back and try different permutations of FSB and divider increases to attain those same speeds you just found, but whats the point?

so instead of doing some ridiculous 247x 17 multi or whatever you might end up doing, you can do 200x 20 on the cpu, nb divider to 2800, and ram divider to 1600 (if using 1333) or 1866 (if using 1600). Simple right? thumb.gif
post #23 of 61
Damn i got luck with my 1055t 3.92@1.42v lol, been that way for 6 months or more. Never crashes
post #24 of 61
Quote:
Originally Posted by guitarmageddon88 View Post

I find the fsb method to be a little dated, and it causes a lot more undue stress on the user trying to fiddle with all those different clock changes at once. Its easier to do each multi/divider independently. I support the OP's method, with some modifications. First, find the multi limit of your system, until your happy with voltage/temps for your rig. (working smarter, not harder, is not "noob" for using the multi first) Then, once you are satisfied, you can either return cpu multi to stock, or leave it where you found it to be stable, then commence with the northbridge multi increases, raising voltage to where you are satisfied. Then, once that is done, you can up your ram divider one notch to start. Nearly ALL kits will go up one divider with about .1 to .12v increase. Once you are there, do the typical stress tests (make sure you memtest before booting into OS when messing with memory of course) and your golden. Now, once you find those good levels, you can add a few FSB notches to juice out those last little ounces of speed. This is way easier to attain the SAME SPEEDS over the "non noob FSB method" I suppose once you find your limitations on each component via the multi/divider method, you can then go back and try different permutations of FSB and divider increases to attain those same speeds you just found, but whats the point?
so instead of doing some ridiculous 247x 17 multi or whatever you might end up doing, you can do 200x 20 on the cpu, nb divider to 2800, and ram divider to 1600 (if using 1333) or 1866 (if using 1600). Simple right? thumb.gif

What's the point? Well would you rather have a system that's stable @ 1.46v or one that's not @ 1.55v? That's the difference. You WILL NOT attain the same speeds, period. Well not one that any overclocker that has some time under their belts would deem stable. It's the same reason why I don't use Intels burn test on AMD's. It's absolutely worthless. A member of another club on here (socket 939, he's an admin now) ran 2 instances of IBT on his overclocked 1100T, 500 passes (that's 18 hours). Passed with flying colors. He fired up Folding at Home, BSOD within 5 minutes. He's never had that happen with a rig that was 24hours Prime 95 stable.

Also, keep in mind that this is an AMD, the Intel method of finding your max multiplier is fairly worthless here. With AMD's it's always better to keep the multiplier as low as possible and raise the FSB. Not only will it keep your CPU cooler, you OC will require less voltage and be more stable.
Also, using odd dividers or halves is not recommended, as it's not in sync with the rest of the hardware. It's better to keep the clock cycles even.

And you mention that the days of FSB OCing are the "old" way, well you must remember that the K10 architecture is based off of "old" tech, so that method still applies here.
Black Box
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770k Asus P8Z77-Vpro Evga 780 Classified Crucial Ballistix 
Hard DriveHard DriveOptical DriveCooling
A-Data 128GB SSD Samsung Spinpoint F3 1TB Pioneer BDR-207DBK Corsair H80 
OSMonitorKeyboardPower
Win 7 pro 64 Hannspree 25" 1080p LCD 2ms Razer Lycosa PC Power & Cooling 760w 
CaseMouseMouse PadAudio
Corsair 550D MX 518 Allsop Creative XFi Xtreme music 
  hide details  
Reply
Black Box
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770k Asus P8Z77-Vpro Evga 780 Classified Crucial Ballistix 
Hard DriveHard DriveOptical DriveCooling
A-Data 128GB SSD Samsung Spinpoint F3 1TB Pioneer BDR-207DBK Corsair H80 
OSMonitorKeyboardPower
Win 7 pro 64 Hannspree 25" 1080p LCD 2ms Razer Lycosa PC Power & Cooling 760w 
CaseMouseMouse PadAudio
Corsair 550D MX 518 Allsop Creative XFi Xtreme music 
  hide details  
Reply
post #25 of 61
Dangit,knew I forgot something. Be sure to disable cool n' quiet and turbo when OCing. Turbo is whats giving you the weird voltage fluctuations to 1.5.
Black Box
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770k Asus P8Z77-Vpro Evga 780 Classified Crucial Ballistix 
Hard DriveHard DriveOptical DriveCooling
A-Data 128GB SSD Samsung Spinpoint F3 1TB Pioneer BDR-207DBK Corsair H80 
OSMonitorKeyboardPower
Win 7 pro 64 Hannspree 25" 1080p LCD 2ms Razer Lycosa PC Power & Cooling 760w 
CaseMouseMouse PadAudio
Corsair 550D MX 518 Allsop Creative XFi Xtreme music 
  hide details  
Reply
Black Box
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770k Asus P8Z77-Vpro Evga 780 Classified Crucial Ballistix 
Hard DriveHard DriveOptical DriveCooling
A-Data 128GB SSD Samsung Spinpoint F3 1TB Pioneer BDR-207DBK Corsair H80 
OSMonitorKeyboardPower
Win 7 pro 64 Hannspree 25" 1080p LCD 2ms Razer Lycosa PC Power & Cooling 760w 
CaseMouseMouse PadAudio
Corsair 550D MX 518 Allsop Creative XFi Xtreme music 
  hide details  
Reply
post #26 of 61
Thread Starter 
Quote:
Originally Posted by BlackOmega View Post

Dangit,knew I forgot something. Be sure to disable cool n' quiet and turbo when OCing. Turbo is whats giving you the weird voltage fluctuations to 1.5.

yupp turbo core is off, does it really matter to have cool n quiet off, i read in alot of places it didnt make much of a diff
post #27 of 61
Frequency is frequency is frequency. That simple. I have tried many times when I owned my 965, my 1090t, my 1100t, to see what was better 4ghz via fsb, or 4ghz via multi, no difference whatsoever, so I say that from personally experiencing this. Same speeds were reached, same stability, same instabilities, same temps. I have only recently moved to intel so Im not coming here saying "do it like they do sandy bridge" because sandy bridge is different. Now on sandy bridge, FSB will mess you up, but on amd, at least the phenoms, and you have unlocked multis, why spin your wheels for nothing?

edit: let me elaborate more. 243 fsb x 16.5 multi is 4009mhz. This results in a cpu no more stable than 200x 20 multi. All you have done now is introduce 4 other variables to confuse the user and cause instability. With the multi and divider method, you isolate each increase. Omega, you are saying increase the fsb til you get instability....then what? Is it in the cpu? is it in the ht? is it in the ram? is it in the northbridge? What then? increase the voltage over all? not very efficient. This is why science utilizes controlled and experimental variables. It would kind of be like treating a cancer patient with medicine that also treats chlamydia, aids, and the common cold just to cover your bases.
Edited by guitarmageddon88 - 2/11/12 at 10:53am
post #28 of 61
Quote:
Originally Posted by guitarmageddon88 View Post

Frequency is frequency is frequency. That simple. I have tried many times when I owned my 965, my 1090t, my 1100t, to see what was better 4ghz via fsb, or 4ghz via multi, no difference whatsoever, so I say that from personally experiencing this. Same speeds were reached, same stability, same instabilities, same temps. I have only recently moved to intel so Im not coming here saying "do it like they do sandy bridge" because sandy bridge is different. Now on sandy bridge, FSB will mess you up, but on amd, at least the phenoms, and you have unlocked multis, why spin your wheels for nothing?
edit: let me elaborate more. 243 fsb x 16.5 multi is 4009mhz. This results in a cpu no more stable than 200x 20 multi. All you have done now is introduce 4 other variables to confuse the user and cause instability. With the multi and divider method, you isolate each increase. Omega, you are saying increase the fsb til you get instability....then what? Is it in the cpu? is it in the ht? is it in the ram? is it in the northbridge? What then? increase the voltage over all? not very efficient. This is why science utilizes controlled and experimental variables. It would kind of be like treating a cancer patient with medicine that also treats chlamydia, aids, and the common cold just to cover your bases.

doh.gif

Frequency is relative to the sum of it's parts. So it's NOT that simple. And I'm very curious, what did you use for stability testing and for how long?

Regardless, you used a half in your multiplier, as I mentioned before, AMD's tend not to like that too much. So once again what did you use for stability testing? I'm positive that it wasn't P95, well at least not for any real length of time. Because if it was, with the higher FSB, it should require less voltage as the motherboard is sharing the load. Hence why my 720 wont stabilize @ 3.6 via multiplier only, even with 1.55v (and that's not even OCing the CPU-NB or the RAM), yet @ 225FSB it's 24hours stable @ 1.46v with the CPU-NB @ 2650 and the RAM @ 1600 @ 7-7-7-21-28-1T.

And you sort of proved my point. You don't know very much about why you test the way I suggested. The point of raising your FSB as high as it will go, within reason, is to find the MOTHERBOARDS limit not the CPUs. You go until it wont POST anymore.This does not bring about instability. It either works or it doesn't. You do this to give you more options in overclocking the rest of your components.

And when you're combining the variables, you add one at a time and test. You don't go in like an idiot and haphazardly mash it all together and hope it works. Why? When adding one at a time, you know EXACTLY what component is causing instability. It's not a guessing game, it's trial and error.

And how is it not efficient? You're removing voltage from one area and adding it to another. So most likely, the voltage difference would be a wash.

And another point, the reason why you want to keep your CPU voltage down is to reduce electromigration. The higher the voltage, the faster this occurs.

Look, you OC the way you want. I do it the way I know works well and have done for a long time specifically with AMD CPU's. I'm REALLY particular about stability, most people's OC's wouldn't pass my stringent standards.
Black Box
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770k Asus P8Z77-Vpro Evga 780 Classified Crucial Ballistix 
Hard DriveHard DriveOptical DriveCooling
A-Data 128GB SSD Samsung Spinpoint F3 1TB Pioneer BDR-207DBK Corsair H80 
OSMonitorKeyboardPower
Win 7 pro 64 Hannspree 25" 1080p LCD 2ms Razer Lycosa PC Power & Cooling 760w 
CaseMouseMouse PadAudio
Corsair 550D MX 518 Allsop Creative XFi Xtreme music 
  hide details  
Reply
Black Box
(16 items)
 
  
CPUMotherboardGraphicsRAM
i7 3770k Asus P8Z77-Vpro Evga 780 Classified Crucial Ballistix 
Hard DriveHard DriveOptical DriveCooling
A-Data 128GB SSD Samsung Spinpoint F3 1TB Pioneer BDR-207DBK Corsair H80 
OSMonitorKeyboardPower
Win 7 pro 64 Hannspree 25" 1080p LCD 2ms Razer Lycosa PC Power & Cooling 760w 
CaseMouseMouse PadAudio
Corsair 550D MX 518 Allsop Creative XFi Xtreme music 
  hide details  
Reply
post #29 of 61
i just got my 1090t three days ago and i cant believe i got 3.8 without any voltage increase and its stable 6 hours prime 95,thats stable enough for me,i'll take it further when i get more time but im very impressed by the overclocking ability of this chip,
strix Z270E
(18 items)
 
HAF 922
(14 items)
 
902
(11 items)
 
CPUMotherboardGraphicsRAM
I7 7700K strix Z270E asus gtx970 dcII 4gb 16gb gskill ripjaws 2400mhz 
Hard DriveHard DriveOptical DriveCooling
seagate sata 1 tb samsung 850 evo 500gb ssd lg blueray corsair H105 
OSMonitorMonitorMonitor
windows 10 home  asus 27inch 1080p  acer 24 inch 1080p acer 24 inch 1080p 
KeyboardPowerCaseMouse
logitech G110 evga supernova G2 750watt fractual design define XL R2 mx518 
Audio
sony wireless,,,headphones 
CPUMotherboardGraphicsRAM
fx 6300 asus m5a99X evo asus gtx 670 2gb  8gb 1600mhz gskill ripjaws 
Hard DriveHard DriveOptical DriveCooling
corsair 120gb ssd WD 1 tb 7200rpm lg blueray H100 
OSMonitorKeyboardPower
windows 7 home premium 50 inch 3d plasma 1080p,,smart tv logitech wireless  corsair tx 750 watt 
CaseAudio
coolermaster 922 denon 7.1,home theatre 
CPUMotherboardGraphicsRAM
amd 965x4 asus,m4n72-e asus gtx 570 1.25gb 8gb 800 gskill ddr2 
Hard DriveOptical DriveCoolingMonitor
samsung 840 evo lg cheap zalman something cpu cooler??,,,works pretty good lg 22 inch 
PowerCase
coolermaster 850 antec 902 
  hide details  
Reply
strix Z270E
(18 items)
 
HAF 922
(14 items)
 
902
(11 items)
 
CPUMotherboardGraphicsRAM
I7 7700K strix Z270E asus gtx970 dcII 4gb 16gb gskill ripjaws 2400mhz 
Hard DriveHard DriveOptical DriveCooling
seagate sata 1 tb samsung 850 evo 500gb ssd lg blueray corsair H105 
OSMonitorMonitorMonitor
windows 10 home  asus 27inch 1080p  acer 24 inch 1080p acer 24 inch 1080p 
KeyboardPowerCaseMouse
logitech G110 evga supernova G2 750watt fractual design define XL R2 mx518 
Audio
sony wireless,,,headphones 
CPUMotherboardGraphicsRAM
fx 6300 asus m5a99X evo asus gtx 670 2gb  8gb 1600mhz gskill ripjaws 
Hard DriveHard DriveOptical DriveCooling
corsair 120gb ssd WD 1 tb 7200rpm lg blueray H100 
OSMonitorKeyboardPower
windows 7 home premium 50 inch 3d plasma 1080p,,smart tv logitech wireless  corsair tx 750 watt 
CaseAudio
coolermaster 922 denon 7.1,home theatre 
CPUMotherboardGraphicsRAM
amd 965x4 asus,m4n72-e asus gtx 570 1.25gb 8gb 800 gskill ddr2 
Hard DriveOptical DriveCoolingMonitor
samsung 840 evo lg cheap zalman something cpu cooler??,,,works pretty good lg 22 inch 
PowerCase
coolermaster 850 antec 902 
  hide details  
Reply
post #30 of 61
Quote:
Originally Posted by BlackOmega View Post

doh.gif
Frequency is relative to the sum of it's parts. So it's NOT that simple. And I'm very curious, what did you use for stability testing and for how long?
Regardless, you used a half in your multiplier, as I mentioned before, AMD's tend not to like that too much. So once again what did you use for stability testing? I'm positive that it wasn't P95, well at least not for any real length of time. Because if it was, with the higher FSB, it should require less voltage as the motherboard is sharing the load. Hence why my 720 wont stabilize @ 3.6 via multiplier only, even with 1.55v (and that's not even OCing the CPU-NB or the RAM), yet @ 225FSB it's 24hours stable @ 1.46v with the CPU-NB @ 2650 and the RAM @ 1600 @ 7-7-7-21-28-1T.
And you sort of proved my point. You don't know very much about why you test the way I suggested. The point of raising your FSB as high as it will go, within reason, is to find the MOTHERBOARDS limit not the CPUs. You go until it wont POST anymore.This does not bring about instability. It either works or it doesn't. You do this to give you more options in overclocking the rest of your components.

And when you're combining the variables, you add one at a time and test. You don't go in like an idiot and haphazardly mash it all together and hope it works. Why? When adding one at a time, you know EXACTLY what component is causing instability. It's not a guessing game, it's trial and error.
And how is it not efficient? You're removing voltage from one area and adding it to another. So most likely, the voltage difference would be a wash.
And another point, the reason why you want to keep your CPU voltage down is to reduce electromigration. The higher the voltage, the faster this occurs.
Look, you OC the way you want. I do it the way I know works well and have done for a long time specifically with AMD CPU's. I'm REALLY particular about stability, most people's OC's wouldn't pass my stringent standards.

I love it when people condescendingly put a doh.gif in their post and that somehow means that theyre just so smart and everyone else is just so dumb.

Ill tell you what you want to hear. No, I didn't test prime 95 at 24 hours with 90 percent memory....no no no..I just ran a pass of unigine heaven and called it good, that seems to work most of the time. Thats what you're supposed to do right? thumb.gif
Quote:
Originally Posted by BlackOmega View Post

And when you're combining the variables, you add one at a time and test. You don't go in like an idiot and haphazardly mash it all together and hope it works. Why? When adding one at a time, you know EXACTLY what component is causing instability. It's not a guessing game, it's trial and error.

thats kind of an oxymoron. Your method does entail going in like an idiot haphazardly combing crap. And are you actually arguing my point for me there? When you ADD ONE VARIABLE AT A TIME (multi/divider) you can find out, to quote you "exactly what component is causing instability"

Tell me this my friend. You begin at 200 fsb, 17 multi. You commence increasing the fsb one notch at a time. Finally you get to 215 and you crash....According to you, you know exactly what component causes that, even though your 215 fsb increase (that up until now was stable) also increase ht speed, nb speed, cpu speed, as well as memory speed. Again, if you have a locked cpu, you would have to do it this way.

So, you are assuming that my overclock wouldnt meet your condescendingly strict criteria.

Quote:
Originally Posted by BlackOmega View Post

And another point, the reason why you want to keep your CPU voltage down is to reduce electromigration. The higher the voltage, the faster this occurs.

When was it ever said in my post that more voltage is better? I have seen no evidence to prove that exactly FSB overclocking to the same frequency as multi overclocking demands less voltage. I mean, you say it does, but in all the testing Ive done to specifically find that out, I have no evidence to prove it. Just like the "half multiplier isnt as stable" claims.

Also, are your strict overclocking standards also applicable to your 1090t on a 790fx board, or is that only until the board fries?
Edited by guitarmageddon88 - 2/11/12 at 1:11pm
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD CPUs
Overclock.net › Forums › AMD › AMD CPUs › Phenom II x6 1090t BE overclocked at 3.8Ghz