• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

OC adventures with a 1100T

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

older&dirtier

New Member
Joined
May 23, 2018
First of all props to Dolk!

Too happy to be here dudes! Hello to all!

This post may seem like it has no purpose, but I really need your help to push the Thuban further and correct some mistakes.

My build is:
GIGABYTE GA-990FX-GAMING
AMD PHENOM II X6 1100T BE (Silentium PC Fortis 2 XE1226 + 2*Akasa Piranha)
G.SKILL F3-17000CL11-4GBXL (2*4GB modules)
COOLEMASTER HAF 922 (2*200mm+1*140mm IN, 1*200mm+1*120mm OUT)
Phobya Touch 6 FAN Controller
COOLERMASTER V750 PSU
AMD Sapphire Radeon R9 280X Tri-X OC
Samsung 840 EVO 250GB
Seagate Barracuda Green 1.5TB
WIN 10 x64 PRO

CPB, Cool & Quiet, C1E turned off.
So I have found that stock voltages(1.375V) get me up to 3.5-3.6 GHz stable-ish. (did not test much)
For 3.7GHz(200*18.5) I need at least 1.4V.
For 3.8GHz(200*19) 1.42-1.45V
For 3.9GHz(200*19.5) at least 1.45-1.475V.
For 4GHz and above nothing seems to work.

I am trying to find at least a stable setup for 3.9GHz in combination with 1333MHz(200*6.66) RAM, because I have read that 1600MHz puts to much stress on the Thuban NB and I have found that 1333MHz has about the same perfomance and better latency, at least in my system. I did leave RAM at the default 1600@[email protected], as I was trying to find the highest CPU clock, but after a point I tried 1333@[email protected] and got stabler. These settings are from the SPD. In the following screenshots, I also include one from AIDA64, as it shows some values that CPU-Z doesn't.

AIDA_SPD.PNG CPU-Z_SPD_Slot#2.PNG CPU-Z_SPD_Slot#1.PNG CPU-Z_Memory.PNG CPU-Z_Mainboard.PNG CPU-Z_CPU.PNG

Sorry, I don't know how to make the pictures thumbnails-like and zoom in when you click on them... I guess if I knew Javascript I could do it. Well, I hope it doesn't hurt too much...

Yesterday, as you can see I was trying with 2600 NB & HT, and got the following:

prime95_smallFFT_38MinuteRunning.PNG prime95_smallFFT_2HoursRunning.PNG prime95_smallFFT_2HoursRunningStop.PNG intelburn.PNG intelburnfail.PNG

Passed 2 hours of Small FFT Prime 95 (to find if the issues are NB related) and crashed in Intel burn test,so I thought memory related. I changed to looser RFC (my MB has 5 settings 90,110,160,300,350 ns),crash again. I upped the RAM voltage, crash. I even tried the default 1600@[email protected], crash. Then I played with different NB-HT combinations to give room to the RAM.
@cpu-NB 1.30V,
2800 both
2800-2600
2800-2400
2800-2200
2800-2000

@cpu-NB 1.35V
3000 both etc

Again only crashes in Intel Burn test and OCCT Large Data Set.
Just for laughs, then I wanted to see How High(impossible movie by the way, summarizes a lot:p) I can boot and bench by disabling cores.
So I set 1.55V core & CPU-NB 1.35V with NB 3000MHz, HT 2200MHZ and:
4.2GHz 5 cores
4.2GHz 4 cores
4.4GHz 1 core
Then the BIOS refused to change to 2 cores, got unresponsive 2 times, did a CMOS reset. Good MB by the way, it was the only time last days it locked up like that, and I have tried numerous other bad combos of settings!
Here is a review of the MB, where the guy is awed about how easily you can burn stuff: http://cxzoid.blogspot.gr/2016/06/oc-review-gigabyte-ga-990fx-gaming.html
True, there is not even a slight indication that you go above recommended voltage! My old GA-890GPA-UD3H had orange and red gradients...

So, today I had my best result so far.
3.8GHz(19*200), 1.45V core, 1.30V CPU-NB, 1.25V HT, 2800MHz NB, 2400MHz HT, 1333MHz 9-9-9-24(and the rest SPD settings)@1.55V RAM


Passed IntelBurn very high 10 Pass and I currently have Prime 95 blend running for 8+ hours, no errors.
Sorry I attached so many photos again, I want you to see temps and voltages as well, from the AIDA64 OSD I have on the right, throughout the tests. Ambient temp 27-29C.
Sorry I OCD so much, I am puzzled and want confirmation that my system is OK enough, to continue to push it to find the optimal balance...

intelburn_3.8_2800NB_2400HT_1333RAM.PNG intelburn_3.8_2800NB_2400HT_1333RAM_SUCCESS!.PNG prime95_1h15min_3.8_2800NB_2400HT_1333RAM.PNG prime95_2h30min_3.8_2800NB_2400HT_1333RAM.PNG prime95_4h13min_3.8_2800NB_2400HT_1333RAM.PNG prime95_8h7min_3.8_2800NB_2400HT_1333RAM.PNG

Yeah.
Now I can give you my address to send that hitman. Or Dolk to send a flock of ironic pigeons to feast.

So, I believe that I now have eliminated issues with NB, HT and RAM and I can play only with the CPU multiplier and Vcore.
Well, I will probably OCD some more and see how the bench results change with different HT (because I am not so sure about that 2800-2400) and lower that 1.55V RAM...
To quote Dolk, Overclocking the Phenom II is a simply complicated task of check, and re-check for hours on hours a day. And tedious, oh so tedious, if I may say!

Can I ever get to 4+ GHz, or I just have lower grade silicon?
The ultimate goal is to also find an OC with about 1.4Vcore, because summer here will hit 40+C. These days it is 25-30C.
And maybe 205 FSB, because it also takes my RAM @685MHz and the timings I am now.

Guys, this is it!

Who am I kidding? Here are some screenshots of my BIOS!!!

180524004601.png 180524004620.png 180524004629.png

What do NB Voltage, HT Link Voltage do? I raised HT a little(1.25V), thinking it would help with 2400MHz HT.
Also CPU-PLL raise only when raising FSB?

Thanks in advance for taking the time to read this unbelievable thread!!!
 
Nice! I'd like to get my hands on a Thuban and play with it. They seem to be harder to hit 4 GHz with than the Denebs. My 980 BE rocks 3.9 GHz as a plug n' play. The mobo automatically runs it there on all cores.
 
Nice! I'd like to get my hands on a Thuban and play with it. They seem to be harder to hit 4 GHz with than the Denebs. My 980 BE rocks 3.9 GHz as a plug n' play. The mobo automatically runs it there on all cores.

Cool! It is newer than mine, eh? It also unlocks to 6 cores? Or not, as there were no 6 core Denebs? Mine has 3.3 base and 3.7 CPB on max 3 cores. I love it, that's why I have it from 2011 and have no FX(meh) or Intel(moneyh) :p. Oh well, maybe I will upgrade with 3rd or 4th Ryzen gen.
 
Not sure on which is newer, but both were the fastest releases of their respective cores.

Or not, as there were no 6 core Denebs?

Yeah, just a quad core. It's in a backup/secondary rig ATM. I'm putting together a benching rig (as finances allow), so I hope to slap it around and make it cry at some point. LOL It's still a great little chip that runs XP without effort.
 
Thought there was no point in going over 2600 HT (especially with slower memory), CPU/NB is the one that gives you the performance boost with Phenom II isn’t it ?
 
Nice, would love to see the results!
runs XP without effort.
:muahaha:

Thought there was no point in going over 2600 HT (especially with slower memory), CPU/NB is the one that gives you the performance boost with Phenom II isn’t it ?

Yeah, I also believe that.
But, to quote Dolk on the 3 Step Guide "On a side note, I did mention that the HTT will be used a lot more with Thuban. For some reason, Thubans like to have the HTT just as high as the CPU-NB. It often creates a much more stable environment if you do this. But, on the other hand, it may not work at all. From what I have learned, the HTT likes to stay with the CPU-NB up to 2800 – 3000 MHz. If you go above this, you may want to lower the HTT back down between 2200 – 2400 MHz."
So, I can't find any info from AMD on this relation of NB-HT and I am still trying to find that golden analogy.
 
Last edited by a moderator:
You mean this guide ?

http://www.overclockers.com/forums/showthread.php/596023-Dolk-s-Guide-to-the-Phenom-II

http://www.overclockers.com/the-importance-of-northbridge-overclocking-with-the-phenom-ii/

"For some reason, Thuban loves to have the HTT just as high as the CPU-NB. It creates a much more stable environment if you do this. But, on another hand, it may not work at all." - CPU/NB seems to peak at 2600mhz (at least on that particular sample), so keep both at 2600mhz for stability and performance ? are you benchmarking after each change to check for improvements ?

Too bad the FX series don't respond to CPU/NB as well as Phenom II, i would have a field day with my 8370 (had him at 3000/3000 before i was told that there was no benefit) :(
 
Last edited:
In answer to your question:
Yes the silicon lottery was not good to you.

Also, your ram sucks. You should be able to do 1600 @ 9-9-9-x on almost any DDR3 set without issue.
 
I agree with the above - Any good thuban of that model should be well capable of 4 GHz with far less voltage, mine can hit that with 1.44v's with ease and even run the CPU-NB at or above 3400 no problem.
An example from way back in a comp I was in over at the OCA forum:
16m 22.313.jpg

I had to work for that one guys - If I had known then what I do now I could have done better. ;)
 
are you benchmarking after each change to check for improvements ?

Yeah, I did benches but never kept records. You inspired me to test every combination! I ran AIDA64 benchmarks Memory Read/Write/Copy/Latency, CPU Queen & Hash (To see when miscalculations will start). The results are well... Some expected, some odd... Will make them in a table and upload them! Thanks!

In answer to your question:
Yes the silicon lottery was not good to you.

Also, your ram sucks. You should be able to do 1600 @ 9-9-9-x on almost any DDR3 set without issue.

Never won anything in a lottery anyway... :p Yeah man, I don't know what G.Skill's problem is! :D I got them on a discount to replace the Corsair 1600CL9 4GB Kit I had (gave it to a friend). 2133CL11, so I thought I can do 1866 on my old Gigabyte GA-890GPA UD3H which had up to 2000 support(which Phenom does that?), little did I know then... But I guess they can still do 9-9-9-x at 1.6V. Will try and post the results on the table I mention above!


Thanks a lot for that! Some pretty interesting stuff in there! Also mentions that the max HT multiplier is 13! Wasn't sure about it!
 
Dr. McCoy, I have 100 questions for you...

Very impressive overclock there (although we have to ding you a point for cutting it down to a dual core)... what mobo and ram combination was that? Do you think you could achieve that CPU NB speed with an 8GB kit? (curious about CPU-NB scaling @ 4GB kit vs 8GB kit)... And you were still hitting a 3400Mhz CPU-NB with all cores enabled? At what voltage? And with what cooling?
 
Dr. McCoy, I have 100 questions for you...

Very impressive overclock there (although we have to ding you a point for cutting it down to a dual core)... what mobo and ram combination was that? Do you think you could achieve that CPU NB speed with an 8GB kit? (curious about CPU-NB scaling @ 4GB kit vs 8GB kit)... And you were still hitting a 3400Mhz CPU-NB with all cores enabled? At what voltage? And with what cooling?

No prob with the dinging, since it's Super PI you'll see most guys doing that because Super PI is single-threaded and doesn't affect the outcome of the bench if you cut down the number of active cores plus it's easier on the chip (Less heat) when OC'ed. I take care of my stuff and make it as easy as possible on the hardware, esp if it's a good piece like this chip is.
As for what I showed it was at the time ran in a MSI 790FX-GD70 board and if memory serves I ran it with my Ripjaws X sticks (CL6 6-6-24) but I was also using at the time a set of OCZ AMD blacks (2X2GB - CL8 8-8-24). The board died shortly after (Of course) and from then on ran it in my CHV and CHV-Z boards and the Sabo 2.0 board I use for sub-zero runs.

I've never tried it with a 8GB kit, I have a formula of sorts for how I run things and using that much RAM for that much CPU-NB speeds is probrably asking a bit much of the chip. For that run is was on H2O that was at ambient, yet ambients were pretty good for the time of year it was (AC in the house set to "Freeze") ran plus I had a rather large res I was using too, it was a 2.5 gallon res so it stayed cool enough to get it done. The rad was a triple with a trio of high CFM 120's so heat removal wasn't an issue. This also helped it to reach the CPU-NB speeds I had it at, was using about 1.52v's for that in the BIOS and the chip responded well, I was wary of the voltage I had going to it and kept a close eye on things with the chip.

As for what I've really been able to make it do, here's an HWbot entry I did with it sometime ago on DICE, the specs are listed with the entry. I was disappointed I couldn't coax it to 5.0 but with a better pot and even LN2 used it could get there. I've posted links to both WPrime runs I did with it when I had it frozen, not that great overall but good enough - Clearly room for improvement here. :D
http://hwbot.org/submission/2336290_bones_wprime___32m_phenom_ii_x6_1100t_be_5sec_359ms
And:
http://hwbot.org/submission/2336292_bones_wprime___1024m_phenom_ii_x6_1100t_be_2min_57sec_781ms
 
THUBAN TIME, what fun!!!!!
friggen power house chip!!!! I still have two around that have been run hard and put away wet many, many times.
3000 on the cpu/nb, have run 6-5-5-5 on the ram at 1700 on both of them, on big water in the summer they will both do 4.2 and in 30 degree temps 4.5 p95 all day long, on small water they have a hard time hitting 4.0, but cool them a little and off they go!!!!
they are on the shelf some where so Alaric needs to pm me about loaning him one.
 
Yep - These chips will go if you can keep them cool enough. They will scale quickly with all cores enabled as I discovered not long after getting mine.
That's the biggest reason I ran Super PI with cores disabled with the CPU-NB voltage I was using, if I hadn't I'd probrably had myself a nice and crispy chip.

My 960T can go even higher on the CPU-NB with sub-zero coolng used http://hwbot.org/submission/2272724_bones_superpi___1m_phenom_ii_x4_960t_be_13sec_187ms so it's a runner that way too. I've seen several of these topping 4000MHz on LN2 before so it's possible if you have it setup correctly.
 
Damn straight! These chips rock. I had my 970 unlocked @ 4.0Ghz @ 3.0Ghz CPU NB (on air) for close to seven years, never missed a beat. Only recently have I run into some reliability issues and I think that's more related to non satisfactory delid (I had it sent out and it didn't go to plan). However, you've inspired me to swap the 8GB kit for my 4GB Corsair XMS3 1600 7-8-7-20 kit to see if I can coax a little more CPU NB out of it.

Here's my best memory effort on air, running a single fan Hyper 212+ (max CPU_NB on air = 3121Mhz)

cache and latency ddr3-1852.JPG
 
Just watch CPU temps under load, additional CPU-NB voltage always means more heat.
I believe you'll have good results trying the 4GB set and you can always give the 8GB kit a shot if you want. :cool:

I think if it's setup right you can get about 3200 reliably, again be sure things are OK with it as you go and gradually tweak up to that speed like I did to figure out what the system needs to get there and no more. If it doesn't make it, that's how it goes sometimes.

Good luck!
 
THUBAN TIME, what fun!!!!!
friggen power house chip!!!! I still have two around that have been run hard and put away wet many, many times.
3000 on the cpu/nb, have run 6-5-5-5 on the ram at 1700 on both of them, on big water in the summer they will both do 4.2 and in 30 degree temps 4.5 p95 all day long, on small water they have a hard time hitting 4.0, but cool them a little and off they go!!!!
they are on the shelf some where so Alaric needs to pm me about loaning him one.

I'm still gathering parts. I put my Hyper 212 on the girlfriend's FX rig until I get her a water loop together. The old Zalman I had on it sounded like a J79 in heat. I have at least 3 motherboards here to compare, so I'm getting impatient! Looking at massive overkill on a PSU this coming month, along with her RGB waterblock. Yeah, seriously, RGB waterblock. LOL
 
NB-HT relation

Hey friends!
I have some test results!
tests.PNG

I have highlighted some values I think are best overall... I also have screenshots of SuperPi 32M, Wprime and IntelBurn in between, but thought not overloading the thread. Some notes, yes I tested up to 2800-3000 HT to see what happens... Well, at 2800, AIDA reported 3000 and CPU-Z changed the HT Link box to FSB, with nothing in. At 3000, AIDA reported 3200 and CPU-Z changed it back to HT Link and showed 400MHz! I don't know what these mean. I shoudn't have done it, as the max HT multiplier is 13, as said here:
https://www.amd.com/Documents/AMD_Dragon_AM3_AM2_Performance_Tuning_Guide.pdf

Also, to see if the ram can handle 1600 9-9-9-27, I tested with 1.6V(as it can do 2133 at that voltage) which is also noted in the above guide, for stability. Voltages on air of 1.35-1.50V for the NB and 1.20-1.35V for the HT, are also noted in the guide.

Am I correct to take the best combinations of NB-HT from the table I made and push to find lower timings for 1600 Ram, keeping the voltage at 1.6?
Then with these lower timings to find the max CPU clock I can go with reasonable Vcore?
Then if I hit a wall, to go back and with the same NB-HT (as I see on the table, the results are pretty consistent, just 1600 is faster) to find lower timings for 1333 Ram at 1.6V, which will maybe allow me higher CPU clock?
Then, to raise a little the FSB and the CPU-PLL voltage?
My goal is 24/7 stable.

What do you guys think?
 
Back