AMD "Thuban" Core (Phenom II X6) XS Overclocking Charts

[PetNorth]

My Updated OC

• CPU Model: AMD Phenom II X6 1090T BE
• CPU Stepping: CCBBE CB 1011APMW
• CPU Frequency: 4000 MHz
• CPU vCore: 1.393v (BIOS) -- ~1.415v (Prime95 full load)
• CPU Multiplier: 20x
• CPU Turbo: Disabled
• CPU NB Speed: 2000 MHz (1.125v)
• HT Ref Speed: 200 MHz
• RAM Speed: DDR3-1333
• RAM Timings: 9-9-9-24-2T
• RAM Configuration: 4 x 2GB
• RAM vDIMM: 1.55V
• Motherboard: ASUS Crosshair IV Formula
• Chipset/Socket: 890FX + SB850, AM3
• Cooling: Air (Prolimatek Megahalems); 322W Peak Power Consumption Prime95 Full Load (~300W average)
• Temps: Everest: Idle (CPU 38C; Cores 20C) -- Load (CPU 69C; Cores 49C)
• Operating System: Windows 7
• 32/64-Bit: 64
• Stable/Suicide/Untested: Stable (Prime95 x64: Blend; 9h 39m)

SS

[Polizei]

My stable entry. RAM is a bit weak... I'll work on it later when I try for 3900mhz stable.

CPU Model: AMD Phenom II X6 1055T
CPU Stepping: CCBBE CB 1013 CPDW
CPU Frequency: 3808 MHz
CPU vCore: 1.387V
CPU Multiplier: 14x
CPU Turbo: Disabled
CPU NB Speed: 2448MHz
HT Ref Speed: 272 MHz
RAM Speed: DDR3-1451
RAM Timings: 8-8-8-24-1T
RAM Configuration: 2 x 2GB
RAM vDIMM: 1.7V
Motherboard: MSI 890GXM-G65
Chipset/Socket: 890GX + SB850, AM3
Cooling: Xigmatek HDT-S1283 + Yate Loon Medium
Temps: 30c idle, 52c load
Operating System: Windows 7
32/64-Bit: 64
Stable/Suicide/Untested: Stable

[SquattingDog]

Updated. No, I haven't tested wall power, but I can do so if you wish. I do have a watt meter. What kind of configuration are you interested in? I'll have to use a discrete GPU as this isn't a GX board, which would probably be what you'd use in an actual HTPC. I can test it with any combination of the following you'd like:

- (2 Max) Radeon HD 5850 OC
- Radeon 4850
- Radeon 3870
- GeForce 8400 GS

Any configuration would be fine - just interested to see stock clock/voltage vs undervolted on stock clock. I know ramping voltage upwards exponentially increases power usage...Only do it if you feel like it/have time, no pressure

[demonkevy666]

Then lemme rephrase:

Of the ones I've used: (namely: prime95, linpack, y-cruncher, and OCCT)
y-cruncher runs hottest on Core i7 when HT is on.


This does sound like a very interesting program. I'll try it later (when my i7 rig finishes its current task - which will take a few more days).

From my experience, pure CPU/execution unit programs don't produce the most heat. That's because the execution units are only a small portion of the processor.
To achieve maximum heat, you need the right balance of computation, memory access, and branching to be able to stress every single portion of the processor simultaneously. (execution units, branch prediction, instruction reorderer, cache, memory controller, etc...)

So far, I have yet to see a single program that can do this. I'll see if core damage can do it.

a few days... you disappoint me...

[richierich]

Today is so much cooler, so close to suicide 4.4GHz Cinebench 11.5

[poke349]

a few days... you disappoint me...

Why? Is it wrong to do something that takes more than a week?
I do it all the time. (And I usually can't pause in the middle of it.)


It finished early this morning. So here are my results:

Core i7 (HT on):

Core Damage: 73 - 74C
y-cruncher: 71 - 72C
prime95: 70 - 71C
IBT (Linpack): 69 - 71C


So yes, Core Damage does what it says.

[Particle]

My hope would be that he was simply being sarcastic. I think we all appreciate your work. After all, he could test them all himself if his need were that pertinent.

--

Did that HTPC simulation run that was requested. At 2800 MHz using a VID of 1.15V, full system power from the wall was 253W during OCCT. 28.5C load temp. Full system power from the wall during idle was 174W at the same frequency and VID. Idle power could be reduced by having CnQ running, but it doesn't work properly for Thuban on this motherboard yet. Also note that I had a Radeon HD 5850 OC installed as the graphics adapter which will pull as much power idle as a traditional HTPC's GPU probably would at full load: 25W or so. I ran it with one SSD for booting and one HDD for storage as might be found in a mainstream HTPC. The SSD's contribution to power draw would be nearly insignificant, so just discard that if you don't think an SSD is commonplace for an HTPC. The point was primarily to get a mechanical mass storage drive in there.

[poke349]

meh... I'm a lot better at spotting sarcasm face-to-face than online.

[Daveburt714]

My first X6...

CPU Model: AMD Phenom II X6 1055T
CPU Stepping: CCBBE CB 1012APMW
CPU Frequency: 4130 MHz
CPU vCore: 1.4625v (BIOS)
CPU Multiplier: 14x
CPU Turbo: Disabled
CPU NB Speed: 2655 MHz (1.20v)
HT Ref Speed: 295 MHz
RAM Speed: DDR3-1575
RAM Timings: 7-8-7-18-30 1T
RAM Configuration: 2 x 2GB
RAM vDIMM: 1.475V
Motherboard: ASUS Crosshair IV Formula
Chipset/Socket: 890FX + SB850, AM3
Cooling: Custom Water
Temps: PC Probe / Tcase Idle 28c, End of run 48c
Operating System: Windows 7
32/64-Bit: 64
Stable/Suicide/Untested: Stable (Prime Blend/33 Minutes, your call )




I'll leave it up to you as to whether you wan't to call it Stable or not Particle... I've killed a few boards in the past with extended stress runs and I don't really want to fry my new gear...

[charged3800z24]

My first X6...

I'll leave it up to you as to whether you wan't to call it Stable or not Particle... I've killed a few boards in the past with extended stress runs and I don't really want to fry my new gear...

Is that all you got Dave lol.....

[scbl]

A very sweet upgrade from my E6600.

I'm having trouble going over 3.9GHz, but will keep trying.

• CPU Model: AMD Phenom II X6 1055T
• CPU Stepping: CCBBE CB 1013CPMW
• CPU Frequency: 3900 MHz
• CPU vCore: 1.456V
• CPU Multiplier: 13x
• CPU Turbo: Disabled
• CPU NB Speed: 2400 MHz
• HT Ref Speed: 300 MHz
• RAM Speed: DDR3-1600
• RAM Timings: 9-9-9-24-1T
• RAM Configuration: 4 GB (2 x 2 GB)
• RAM vDIMM: 1.65V
• Motherboard: Gigabyte GA-890GPA-UD3H
• Chipset/Socket: AM3, 890GX + SB850
• Cooling: Air (Noctua NH-D14)
• Temps: 28C Idle / 45C Load
• Operating System: Windows 7
• 32/64-Bit: 64
• Stable/Suicide/Untested: Stable

[Polizei]

scbl - I also upgraded from an E6600... some of the best $400 I've ever spent.

[Particle]

That chip is looking pretty hot, Daveburt. You've got yourself a real gem there if it's stable.

I've color-coded the overclock charts. Green = stable, yellow = untested, and red = unstable. Let me know what you guys think. I believe it helps make the charts more easily readable.

[demonkevy666]

Why? Is it wrong to do something that takes more than a week?
I do it all the time. (And I usually can't pause in the middle of it.)


It finished early this morning. So here are my results:

Core i7 (HT on):

Core Damage: 73 - 74C
y-cruncher: 71 - 72C
prime95: 70 - 71C
IBT (Linpack): 69 - 71C


So yes, Core Damage does what it says.

thanks.

My hope would be that he was simply being sarcastic. I think we all appreciate your work. After all, he could test them all himself if his need were that pertinent.

--

Did that HTPC simulation run that was requested. At 2800 MHz using a VID of 1.15V, full system power from the wall was 253W during OCCT. 28.5C load temp. Full system power from the wall during idle was 174W at the same frequency and VID. Idle power could be reduced by having CnQ running, but it doesn't work properly for Thuban on this motherboard yet. Also note that I had a Radeon HD 5850 OC installed as the graphics adapter which will pull as much power idle as a traditional HTPC's GPU probably would at full load: 25W or so. I ran it with one SSD for booting and one HDD for storage as might be found in a mainstream HTPC. The SSD's contribution to power draw would be nearly insignificant, so just discard that if you don't think an SSD is commonplace for an HTPC. The point was primarily to get a mechanical mass storage drive in there.

I would have tested my self but I don't have my Phenom II 940 in my board or a six core 1090T/1055T to test, I'm on my valcano Phenom X4 9850 BE.
I do appreciate seeing the I7 tested on that thanks, seems to make the most heat in a cpu, once I get a 1090T will be using it too.

[charged3800z24]

I've color-coded the overclock charts. Green = stable, yellow = untested, and red = unstable. Let me know what you guys think. I believe it helps make the charts more easily readable.

I like, one thing though... Can we get the user name on the left side? I like to see everyones results and I have to keep scrolling to see the same and back to view the score

Not a big deal, just thought I'd mentioned it, great job on the charts

[Andi64]

I have a request.

Can someone please try OCCT:linpack at 4.0Ghz or more? I see a lot of Prime Blend, but in my experience OCCT stability is harder to achieve. I've had CPUs that prime for hours and coundn't resist 15 minutes of OCCT.

[MrRuckus]

Nice to see the X6 runs with the DDR2 crowd if you dont want to upgrade your entire box. I dont think I need the current volts for the current clock. I'll work on that and NB speeds later. All update once they change. (I didn't take a screen of my ram timings, let me know if thats a requirement and I can make another prime run.)

CPU Model: AMD Phenom II X6 1090T BE
CPU Stepping: CCBBE CB 1012APMW
CPU Frequency: 4027 MHz
CPU vCore: 1.500V
CPU Multiplier: 20x
CPU Turbo: Disabled
CPU NB Speed: 2215 MHz
HT Ref Speed: 201 MHz
RAM Speed: DDR2-1066
RAM Timings: 5-6-6-20-2T
RAM Configuration: 4 GB (2 x 2 GB)
RAM vDIMM: 2.1V
Motherboard: Gigabyte GA-MA790X-UD4P
Chipset/Socket: AM3, 790X + SB750 (EDIT: 790X, not FX, my bad)
Cooling: TRUE 120
Temps: 25C Idle / 53C Load
Operating System: Windows 7
32/64-Bit: 64
Stable/Suicide/Untested: Stable

[Particle]

The 4004 number I have on the chart actually was tested with OCCT for eight hours, but not using the linpack method. Linkpack wasn't achieving 100% load.

--

As for names on the left, I agree. I have to scroll back and forth quite a bit myself. I'm not sure if I made column order modular or not in the software--I'll have to try it and see what happens.

Edit: Looks like I did. Let's see how people like the left-handed names.

[MrRuckus]

I have a request.

Can someone please try OCCT:linpack at 4.0Ghz or more? I see a lot of Prime Blend, but in my experience OCCT stability is harder to achieve. I've had CPUs that prime for hours and coundn't resist 15 minutes of OCCT.

Here ya go. Did it for a little over the time requested. Seems it goes in a loop of sorts where it ramps up to 100% load and then restarts the loop of whatever it does.

(ambient temps are a little cooler today, so all the temps are a tad bit lower then before.)

[Andi64]

Thanks Ruckus!

Can you test for 1 hour, please?

[PetNorth]

The 4004 number I have on the chart actually was tested with OCCT for eight hours, but not using the linpack method. Linkpack wasn't achieving 100% load.

--

As for names on the left, I agree. I have to scroll back and forth quite a bit myself. I'm not sure if I made column order modular or not in the software--I'll have to try it and see what happens.

Edit: Looks like I did. Let's see how people like the left-handed names.

Now it's better

I have a suggestion too, to obtain a shorter table... how about to supress some columns (I'm thinking of "HT Mult" -almost blank-; "+Mhz/+mV"; or including "Bitness" with "OS": x64, x32... and supressing "Professional" hehe).

[Particle]

Anyone notice the trends for 1055 vs 1090 on +MHz/+mV? The 1090s seem to be providing better results in terms of frequency gains per extra millivolt applied. The statistic takes base frequency into account. Of the stable results, 1090s seem to be offering about 10MHz per extra millivolt. 1055s seem to average around 8MHz per extra millivolt.

@PetNorth: +MHz/+mV is pretty important, but I like the other ideas.

[charged3800z24]

As for names on the left, I agree. I have to scroll back and forth quite a bit myself. I'm not sure if I made column order modular or not in the software--I'll have to try it and see what happens.

Edit: Looks like I did. Let's see how people like the left-handed names.

Thanks, much appreciated.

Anyone notice the trends for 1055 vs 1090 on +MHz/+mV? The 1090s seem to be providing better results in terms of frequency gains per extra millivolt applied. The statistic takes base frequency into account. Of the stable results, 1090s seem to be offering about 10MHz per extra millivolt. 1055s seem to average around 8MHz per extra millivolt.

@PetNorth: +MHz/+mV is pretty important, but I like the other ideas.

I hadn't paid attention to this portion of the chart yet. Interesting though. I will throw in the 1090T and see if I can up the stable chart.

[slaveondope]

Anyone notice the trends for 1055 vs 1090 on +MHz/+mV? The 1090s seem to be providing better results in terms of frequency gains per extra millivolt applied. The statistic takes base frequency into account. Of the stable results, 1090s seem to be offering about 10MHz per extra millivolt. 1055s seem to average around 8MHz per extra millivolt.

@PetNorth: +MHz/+mV is pretty important, but I like the other ideas.

Keep in mind the mobo and cooling too.
My 1090T is eating up to 1.62 at 4.2ghz and 2.8ghz NB @ 1.265v on prime95 on a H50 at ~53c on the mobo in sig.
As reported by cpuz and hwmonitor which I consider the standard for ocing.

[richierich]

Funny, I've used OCCT Linpack for testing stability before on overclocks and it would show stable. I then tried Prime95 on the same OCs and they would fail. Is it because it's not @ 100% load all the time?