AMD Thuban X6 OCing database

[ShawnD1]

I did some toying around with my 'puter using Turbo to increase the multiplier. It's not the motherboard holding it back. The CPU really does just take a crazy amount of voltage to function.

Phenom X6 1055
3710MHz @ 1.500 volts (that's the minimum stable voltage at this speed)
CPU-Z says the stepping is "0"

 

[DrMrLordX]

Is that one of the newer 95w CPUs, or is it one of the older ones?

 

[ShawnD1]

Is that one of the newer 95w CPUs, or is it one of the older ones?

125W. Not even Newegg sells a 95W version.

 

[DrMrLordX]

The 95w 1055Ts are:

HDT55TWFK6DGR (OEM)
HDT55TWFGRBOX (Retail)

Provantage has the OEM chip for $218 + shipping:

http://www.provantage.com/amd-hdt55t...r~7AAMD2H2.htm

 

[Markfw]

The 95w 1055Ts are:

HDT55TWFK6DGR (OEM)
HDT55TWFGRBOX (Retail)

Provantage has the OEM chip for $218 + shipping:

http://www.provantage.com/amd-hdt55t...r~7AAMD2H2.htm

Why bother there...newegg has it for $200 shipped !

I see those are the 125 watt versions....I would nver pay more for a lower watt version myself. Especially since I OC, and would never see the benefit.

 

[DrMrLordX]

Why bother there...newegg has it for $200 shipped !

I see those are the 125 watt versions....I would nver pay more for a lower watt version myself. Especially since I OC, and would never see the benefit.

I'm not sure what the benefit of the 95w version actually is for an overclocker. When I saw he had a 1055T that required 1.5v just to hit 3.7 ghz, I was curious as to whether it was a newer 95w chip or one of the older 125w chips. Since ShawnD1 indicated that even Newegg didn't have them, I looked around and found some to prove that the chips are readily available.

The 95w chips represent either a minor process improvement (something that does not warrant its own stepping) or just binning for lower vid. Either case might yield chips that have better OC headroom or just hit the "voltage wall" at a lower voltage. I was hoping he had one and could provide some additional information regarding how well the chips overclock.

edit: for those of us who live in a state where we pay sales tax on purchases from the Egg, the chip from Provantage might be just as cheap.

For me, the chip from Provantage is $222 shipped. The 125w chip from the Egg would cost me $218.49 shipped. If it were my money, I'd go with the 95w chip. ZZF would sell it to me for $199 shipped but they're out of stock.

 

[jvroig]

For me, the chip from Provantage is $222 shipped. The 125w chip from the Egg would cost me $218.49 shipped. If it were my money, I'd go with the 95w chip. ZZF would sell it to me for $199 shipped but they're out of stock.

Is that a hypothetical statement for illustrative purposes, or are you actually planning on buying one to replace your Sargas Sempy?

 

[DrMrLordX]

Is that a hypothetical statement for illustrative purposes, or are you actually planning on buying one to replace your Sargas Sempy?

Both? The only reason I don't have a better chip is financial, and a 95W E0 stepping chip is tempting.

I would like to see what price the 1045T will be when it comes out, though.

 

[jvroig]

Has anyone tried disabling a few cores and see how much higher they can OC?

Currently my Phenom II X4 is OCed to a max of 3.813, but when I disable 2 cores, it can reach 4 GHz at the same voltage. CPU-NB freq also goes up, from 2.6 to 2.75.

The Thuban chips seem to just clock better than the Denebs, so I was wondering if disabling a few cores does end up getting higher clocks still (I suppose this won't always be the case if the original OC with six cores active is not thermally limited but already hit the volt wall).

If it does succeed, this makes the chip more useful or versatile. For example, with my X4, when I'm working with databases and complex report generation, all it needs is single-threaded power, so I switch the OC settings to a dual-core only config 4GHz core / 2.75 uncore. When I'm gaiming, I switch the OC settings to a quad-core @ 3.8 / 2.6. (Thank god for BIOSes that have ten "save" slots for settings; I hate my MSI board that didn't have it). If the X6 can be as versatile, I would love to see how high it can actually clock as a dual-core for when you only need single-threaded performance.

 

[Idontcare]

I'm not sure what the benefit of the 95w version actually is for an overclocker. When I saw he had a 1055T that required 1.5v just to hit 3.7 ghz, I was curious as to whether it was a newer 95w chip or one of the older 125w chips. Since ShawnD1 indicated that even Newegg didn't have them, I looked around and found some to prove that the chips are readily available.

The 95w chips represent either a minor process improvement (something that does not warrant its own stepping) or just binning for lower vid. Either case might yield chips that have better OC headroom or just hit the "voltage wall" at a lower voltage. I was hoping he had one and could provide some additional information regarding how well the chips overclock.

edit: for those of us who live in a state where we pay sales tax on purchases from the Egg, the chip from Provantage might be just as cheap.

For me, the chip from Provantage is $222 shipped. The 125w chip from the Egg would cost me $218.49 shipped. If it were my money, I'd go with the 95w chip. ZZF would sell it to me for $199 shipped but they're out of stock.

There are no process or stepping differences, it is all just binning for lower Vcc at stock clockspeed as you stated.

The thing is, and this has been verified to me offline by both AMD and Intel engineers, that if you want higher clockspeeds then you actually want the chips that have higher leakage at idle (higher static leakage, Ioff and IDDQ) as these are the chips whose transistors are all setup for quicker transition from the off state to the on state and that means higher clockspeed potential.

Of course this is only true in the case that you can handle getting the heat out of the chip so you don't go unstable from higher temps. If your setup is such that your cpu is thermal limited then of course you are going to get better clocks by going with the lower leakage chip from the beginning (it just won't clock as high as someone with better cooling and a leakier chip that consumes more power at any given clockspeed).

Currently my Phenom II X4 is OCed to a max of 3.813, but when I disable 2 cores, it can reach 4 GHz at the same voltage.

Cool. How much control do you have over your ability to select which two cores are specifically disabled or enabled? Does your logical core selection always correlate to a physical or does it float around? Can you find out the weakest core by process of elimination and likewise your best clocking cores and further optimize the overclock?

 

[lopri]

I've gotten some DDR3 sticks to play with so I mated them with an 1090T, and learned the performance reliance on memory configuration can be shockingly high at least under some workloads.

• 4.0 GHz | 2.4 GHz NB | DDR-800 -> 66.98 GFlops
• 4.0 GHz | 2.5 GHz NB | DDR-1000 -> 72.27 GFlops
• 4.0 GHz | 3.0 GHz NB | DDR-1666 -> 79.38 GFlops

Not a completely fair comparison, for different amount of memory and different boards were used. Regardless, it seems like the degree of varying performance is beyond what those (4GB v. 8GB) can explain. In any case I didn't expect to see ~80 GFlops.

 

[joe1946]

I'm not sure what the benefit of the 95w version actually is for an overclocker. When I saw he had a 1055T that required 1.5v just to hit 3.7 ghz, I was curious as to whether it was a newer 95w chip or one of the older 125w chips. Since ShawnD1 indicated that even Newegg didn't have them, I looked around and found some to prove that the chips are readily available.

The 95w chips represent either a minor process improvement (something that does not warrant its own stepping) or just binning for lower vid. Either case might yield chips that have better OC headroom or just hit the "voltage wall" at a lower voltage. I was hoping he had one and could provide some additional information regarding how well the chips overclock.

edit: for those of us who live in a state where we pay sales tax on purchases from the Egg, the chip from Provantage might be just as cheap.

For me, the chip from Provantage is $222 shipped. The 125w chip from the Egg would cost me $218.49 shipped. If it were my money, I'd go with the 95w chip. ZZF would sell it to me for $199 shipped but they're out of stock.

Provantage shows the 95w 1055T as special order and I have not seen any online vendor with the 95w version in stock.

 

[lopri]

The thing is, and this has been verified to me offline by both AMD and Intel engineers, that if you want higher clockspeeds then you actually want the chips that have higher leakage at idle (higher static leakage, Ioff and IDDQ) as these are the chips whose transistors are all setup for quicker transition from the off state to the on state and that means higher clockspeed potential.

That reminds me of the some E6600 that appeared in late 2006 ~ early 2007. Those chips had "B" instead of "A" at the end of SPEC code and clocked higher than earlier steppings E6600s, but they ran at noticeably higher temps.

 

[jvroig]

Cool. How much control do you have over your ability to select which two cores are specifically disabled or enabled?

Not much control at all. I only do it through the BIOS, through the ACC Core Control options (the same way you would unlock a dual to a quad core). It has an option to enable/disable the third and/or fourth core. The first two cores are always active, so I can't make the experiment to see which of the four cores are the best.

Not a completely fair comparison, for different amount of memory and different boards were used. Regardless, it seems like the degree of varying performance is beyond what those (4GB v. 8GB) can explain. In any case I didn't expect to see ~80 GFlops.

That's linx performance, right? What apps does that reflect on, Folding-type apps?

 

[jvroig]

The thing is, and this has been verified to me offline by both AMD and Intel engineers, that if you want higher clockspeeds then you actually want the chips that have higher leakage at idle (higher static leakage, Ioff and IDDQ) as these are the chips whose transistors are all setup for quicker transition from the off state to the on state and that means higher clockspeed potential.

I heard somewhere that this is what the TWEAKR chips from AMD were, they were too high leakage to sell at any bin, so instead they packaged them nicely (all "elite"-like) and then sent them out to reviewers/known overclockers, labeling the chips as extreme overclocking bins. Since they were expected to go under LN2 anyway, being high leakage was a pro and not a con.

 

[DrMrLordX]

Provantage shows the 95w 1055T as special order and I have not seen any online vendor with the 95w version in stock.

Availability does seem a bit poor on these chips. It'll ship, it'll just take 15-30 days to get to the buyer, unless ProVantage flubs the whole thing up. Kinda sucks but oh well.

There are no process or stepping differences, it is all just binning for lower Vcc at stock clockspeed as you stated.

Figures. But yeah, that makes sense.

The thing is, and this has been verified to me offline by both AMD and Intel engineers, that if you want higher clockspeeds then you actually want the chips that have higher leakage at idle (higher static leakage, Ioff and IDDQ) as these are the chips whose transistors are all setup for quicker transition from the off state to the on state and that means higher clockspeed potential.

I have heard that as well. At the same time, that fact does not (necessarily) seem to affect chips that have a lower vid than others within the same microarchitecture/process/stepping/etc.

In some cases, the chips with the lowest vid are the cream of the crop. AMD's last foray into binning for low vid/vcore was the Athlon XP-M 2500+, if I recall, and we all know how that worked out.

Their 'e'-designated CPUs from AMD seem to be an entirely different animal that have transistors that are deliberately set up for lower leakage; in other words, I don't think those are the product of mere binning, though I could be wrong. Regardless, the 'e' CPUs are pretty poor overclockers if I recall correctly.

Cool. How much control do you have over your ability to select which two cores are specifically disabled or enabled? Does your logical core selection always correlate to a physical or does it float around? Can you find out the weakest core by process of elimination and likewise your best clocking cores and further optimize the overclock?

I think this varies based on the board. Back when I had my x4 635 working, my 790FX-GD70 would allow me to disable any core, and allegedly would let me increase or decrease clockspeed of any core by increments of 2% over the current clockspeed (though it never seemed to work right; that feature might have relied on the chip being a BE or something). You can get finer control of individual core speeds using something like K10Stat I believe.

Figuring out which cores would clock higher than others is a bit trickier, though usually it's one core in particular that likes to fail when there are memory/NB stability problems. I never OCed on a per-core basis (never got around to it before killing my chip), but it could be done, at least with K10Stat. Detecting which core(s) were most prone to failure during moments of memory instability might be a good path to follow as well.

I've gotten some DDR3 sticks to play with so I mated them with an 1090T, and learned the performance reliance on memory configuration can be shockingly high at least under some workloads.

All I can say is: awesome. Thuban's IMC is pretty impressive from what I've heard, with DDR3-2000 being quite possible with the right memory.

I am not surprised to see notable performance differences at higher NB speeds, though the amount of difference is quite striking. I suspect that NB speed will make a bigger difference on K10.5 chips with larger core counts. It would be fun to re-run all those NB performance tests I did on the 635 using a Thuban chip. Well, maybe not fun, but informative.

 

[ShawnD1]

I've gotten some DDR3 sticks to play with so I mated them with an 1090T, and learned the performance reliance on memory configuration can be shockingly high at least under some workloads.

• 4.0 GHz | 2.4 GHz NB | DDR-800 -> 66.98 GFlops
• 4.0 GHz | 2.5 GHz NB | DDR-1000 -> 72.27 GFlops
• 4.0 GHz | 3.0 GHz NB | DDR-1666 -> 79.38 GFlops

I've had a similar experience. I tested it using Intel Burn Test with the "standard" setting and "auto" number of threads.

4gb DDR3 at 1060mhz in all tests
3710mhz Phenom II X6 1055 in all test
It might be redundant but I'll just copy and paste my northbridge.txt file

1855 mhz northbridge (IBT standard, auto threads):
24.568 sec
22.755 sec
23.368 sec
22.648 sec
22.535 sec
TOTAL - 201.58 sec

2120 mhz northbridge:
26.154 sec
24.926 sec
21.830 sec
22.730 sec
24.331 sec
TOTAL - 205.53 sec

2385 mhz northbridge:
22.906 sec
21.475 sec
21.964 sec
22.022 sec
21.411 sec
TOTAL - 195.20 sec

2650 mhz northbridge
15.916 sec
16.021 sec
15.960 sec
15.955 sec
15.922 sec
TOTAL - 165.06 sec

"auto" northbridge (auto should set itself to 2650)
16.827 sec
16.722 sec
16.718 sec
16.713 sec
16.600 sec
TOTAL - 169.01 sec

2650 mhz high priority
16.320 sec
16.178 sec
16.158 sec
16.149 sec
16.116 sec
TOTAL - 166.26 sec

2650 mhz low priority
17.392 sec
17.386 sec
17.369 sec
17.328 sec
17.366 sec
TOTAL - 172.32 sec


Conclusions:
-2650mhz northbridge is 42.8% higher clockspeed than 1855mhz northbridge, and the actual measured performance increase is 22.1% higher. That's surprisingly well correlated.
-setting IBT to high CPU priority does nothing
-setting IBT to low CPU priority makes it roughly 4% slower (no other programs were running).

 

[lopri]

That's linx performance, right? What apps does that reflect on, Folding-type apps?

I suppose memory-sensitive apps (Duh), such as encoding or encrypting (WinRAR/7-Zip/TrueCrypt) would show the biggest gains. I'm not a folder so I have no idea about Folding performance..

All I can say is: awesome. Thuban's IMC is pretty impressive from what I've heard, with DDR3-2000 being quite possible with the right memory.

It indeed does DDR3-2000/8-9-8-1T. With all 4 sticks, which was a bit of surprise to me. But so far performance doesn't seem to improve from DDR-1666/7-8-7, and that's not a surprise because I haven't raised NB frequency beyond 3.0 GHz yet. I suspect 3.2 GHz+ NB will show meaningful gains again with DDR-2000+.

Another thing that's changed from DDR2->DDR3 transition is that the CPU gets a lot hotter with DDR3 than with DDR2. With DDR2 I felt 1090T system was cooler than i7-920 system (both overclocked, C0 stepping 920), but with DDR3 and 3.0 GHz NB 1090T system feels much hotter now. I suppose that's why AMD can't ship its CPUs with higher NB frequencies.

 

[DrMrLordX]

It indeed does DDR3-2000/8-9-8-1T. With all 4 sticks, which was a bit of surprise to me. But so far performance doesn't seem to improve from DDR-1666/7-8-7, and that's not a surprise because I haven't raised NB frequency beyond 3.0 GHz yet. I suspect 3.2 GHz+ NB will show meaningful gains again with DDR-2000+.

If you manage NB speeds of 3 ghz or higher, do let us know. My C2 chips just haven't wanted to go there. They make a show of it, but . . .

Another thing that's changed from DDR2->DDR3 transition is that the CPU gets a lot hotter with DDR3 than with DDR2. With DDR2 I felt 1090T system was cooler than i7-920 system (both overclocked, C0 stepping 920), but with DDR3 and 3.0 GHz NB 1090T system feels much hotter now. I suppose that's why AMD can't ship its CPUs with higher NB frequencies.

Interesting. I'm not sure why that would be, unless you're hitting higher NB speeds/voltages overall when using DDR3.

 

[busydude]

It indeed does DDR3-2000/8-9-8-1T. With all 4 sticks, which was a bit of surprise to me. But so far performance doesn't seem to improve from DDR-1666/7-8-7, and that's not a surprise because I haven't raised NB frequency beyond 3.0 GHz yet. I suspect 3.2 GHz+ NB will show meaningful gains again with DDR-2000+.

What is your NB voltage, I am interested to know. I might try OC'ing my NB just for the fun of it.

 

[Sokar]

I was lucky enough to get a 95w 1055t when I built my new rig. Complete luck because I never even knew they existed. I noticed that something was different when the max voltage it ever went to(outside turbo) was 1.375v. I tested it and it's stable a 3.8GHz at stock voltage(at least for 10 runs of linpack) but my motherboard makes a scary hissing noise running linpack at those speeds(I guess power draw is becoming too much) so I tuned it down to 3.5GHz.

Mine still runs at 1.475v in turbo. Looks like AMD is REALLY playing it safe with these chips because the voltages are super high compared to what is really needed by the chip. On thing that annoyed me is that my cpu NB voltage was 1.15v and every monitoring program was reporting it as "red" but not because its high but because it's low. I bumped it up a notch but it was stock stable a 2.6GHz at least.

One thing I find really annoying is the cpu temps. Right now my motherboard is reporting 26c temps even though my ambient is like 25c. Core temps are even lower in the 20c range and these are obviously wrong.

Anyways I can tell all you guys wondering if they should get the 1090 or 1055 that for 100$ less I had a lot more fun tweaking this thing than I would have had with the more expensive cpu. Overclocking my friends 965 was boring compared to this lol.

 

[DrMrLordX]

I was lucky enough to get a 95w 1055t when I built my new rig. Complete luck because I never even knew they existed. I noticed that something was different when the max voltage it ever went to(outside turbo) was 1.375v. I tested it and it's stable a 3.8GHz at stock voltage(at least for 10 runs of linpack) but my motherboard makes a scary hissing noise running linpack at those speeds(I guess power draw is becoming too much) so I tuned it down to 3.5GHz.

Hmm, interesting. What board are you using? That's kind of a shame since it sounds like your chip has a lot more headroom . . . but really, who could predict that something like that would happen when running a Thuban? People with my board (790FX-GD70) have reported numerous board deaths when running a Thuban, so I'm going to have to be cautious as well. That's part of the reason why I'd prefer a 95w chip since the power draw at any given clockspeed should be lower.

 

[Sokar]

Hmm, interesting. What board are you using? That's kind of a shame since it sounds like your chip has a lot more headroom . . . but really, who could predict that something like that would happen when running a Thuban? People with my board (790FX-GD70) have reported numerous board deaths when running a Thuban, so I'm going to have to be cautious as well. That's part of the reason why I'd prefer a 95w chip since the power draw at any given clockspeed should be lower.

Yeah I have the MSI 890fx board. Lots of bizarre things going on with this board. AMD overdrive reporting weird temps for TMPIN when I was at 3.8GHz(like in the -280c range....) when right now at idle im getting 28,35,35,0. It never reports anything but 0 for the 4th reading. I assume those are the power regulation chip temps but really im not sure. Also the voltage setting options confuse me. As far as I can tell cpu voltage does nothing but cpu vdd is what you have to use. Also cpu-z always reports my voltage in the over 1.4v range(usually 1.41v) even though its set to 1.375v in the bios and overdrive as well as msi control center are telling me 1.375.

 

[thelvman]

Hi guys, I'm trying to overclock my 1055t and have some questions.
I have a gigabyte 890 ga-890gpa motherboard and 8Gb ram.
I notice by default the bios reports a cpu voltage of 1.475V. This seems high but I'm not as concerned because it operates at a range of 1.22V to 1.47V depending on load.
I overclocked to 3.5Ghz with a turbo of 3.9Ghz. If I run prime 95 with 6 cores, my Vcore goes up to 1.44V. If I set affinity to use only 1 core, my Vcore goes up to 1.47V but my core speed stays at 3.5Ghz. I will see it spike to 3.9Ghz maybe once every minute, but it's only for an instant.
Is there a better way of testing turbo boost? Are my Vcore's too high?
P.S. I'm using CPUID Hardware Monitor to check Vcore and I'm using Coretemp to check my CPU frequencies.

 

[richierich1212]

Yes vcore is too high. Can you manually set your vcore instead? What mobo are you using?