Bulldozer Overclocking Guide (Performance Scaling Charts, max OCs)LN2 Results coming!

[crazydiamond]

wow tell us more dave! 4.9ghz @ 1.4v stable?! not too shabby!

[charged3800z24]

sin: Have you had problems with P95? My machine seems very stable ~4.9 and will run every benchmark I've thrown at it including 20 passes of IBT.
For some reason though, P95 fails on a couple cores almost instantly unless I really crank the v's (>1.4v) and then heat becomes a problem when loaded (~56c)....

Maybe I'm just asking a little too much, but I wondered if you've noticed any incompatabilty with Prime95.

Thanks, Dave

I had that problem with one of the Phenom II chips I tested a long time ago. I was using a older version of Prime95 and didn't realize it. I'm curious if this is a Prime issue?
And I see you have your chip before me, Bah.... lol

[trans am]

sin I gotta hand it to you. Its been a year since I did anything with AMD and its been a year since Ive used a container. I totally skipped sandy because it just defeated the purpose of cold benching. I just got my 8120 and 990FX Ud7 tonight and im messing around with an air cooler and just trying to get back into the swing of things. I would like to personally thank you for putting the time into this guide for xs. you know the prerequisites and basic common sense but its always nice to have a foundation to work from. You laid the foundation and now its up to everyone else to elaborate. thanks man. outstanding job! finally a thread about bulldozer that isnt full of noobs talking trash.

[Dumo]

Finally we can focus on overclocking BD

Still a lot of surprise waiting to be discovered with this platform

[redrumy3]

Hey,

Having a problem was wondering if you might know whats causing it.

First specs:

GIGABYTE GA-990FXA-UD5 F5 Bios
FX-8120
16GB G.Skill DDR3 1600 (1.5v)
Crucial M4 128GB SSD [OS]
Samsung 1TB F3 32MB Raid 0 (x2) [STORAGE]
GIGABYTE 6870 1GB
AuzenTech Prelude
Prolimatech Megahalems
Corsair HX620


This is my friends build and we go into bios and at first it shows Stock voltage for cpu @ 1.41 but when we reboot it shows it as 1.368v. But when we go to overclock and under load his Vcore is all over the place 1.45v and will bounce to 1.31, 1.200v and obviously fails linx right away. Was wondering if you might know what is causing this problem. His temps are fine as well no problems on temps.

sorry for semi hijack if anything mod can delete this and I will make a thread

[sin0822]

sin: Have you had problems with P95? My machine seems very stable ~4.9 and will run every benchmark I've thrown at it including 20 passes of IBT.
For some reason though, P95 fails on a couple cores almost instantly unless I really crank the v's (>1.4v) and then heat becomes a problem when loaded (~56c)....

Maybe I'm just asking a little too much, but I wondered if you've noticed any incompatabilty with Prime95.

Thanks, Dave

I used wprime and ran 1024M a few times to quickly test consistency, but not prime95. I use IBT for stress testing as i like being able to set a limit on the amount of time run, but i will give p95 a shot.

I decided to pick up 10L of LN2 today, and I reapplied some LET, so I am going to do an LN2 run tomorrow morning or somthing, and then the rest of the weekend doing tests for users like you. i will toss P95 into the mix, if not only to test if this platform like SB is better stressed on prime.

Finally we can focus on overclocking BD

Still a lot of surprise waiting to be discovered with this platform

Agreed man, a lot of surprises with cold as well, I have to be honest, i have never benched AMD in the cold with LN2, and i gotta say i love being able to just fill up the POT. Now to find a paste that will hold lol. If i do see any improvements, or any tips provided by other users, and they don't mind then id add them to the first post, and give them credit. That is just what i was able to come up with in a week or so, so I am sure with months there will be much more info.

sin I gotta hand it to you. Its been a year since I did anything with AMD and its been a year since Ive used a container. I totally skipped sandy because it just defeated the purpose of cold benching. I just got my 8120 and 990FX Ud7 tonight and im messing around with an air cooler and just trying to get back into the swing of things. I would like to personally thank you for putting the time into this guide for xs. you know the prerequisites and basic common sense but its always nice to have a foundation to work from. You laid the foundation and now its up to everyone else to elaborate. thanks man. outstanding job! finally a thread about bulldozer that isnt full of noobs talking trash.

Thanks I appreciate it Yea i wanted to keep benchmarks out just for that reason, but i still get guys telling me that their 1100T scores better in cinebench, lol. I am just glad some people find a use for this, because i didn't know how many ppl would actually buy BD after the reviews. There is one thing tho, its just really fun to OC this CPU, if AMD got one thing right it was that BD would be fun to OC>

Hey,

Having a problem was wondering if you might know whats causing it.

First specs:

GIGABYTE GA-990FXA-UD5 F5 Bios
FX-8120
16GB G.Skill DDR3 1600 (1.5v)
Crucial M4 128GB SSD [OS]
Samsung 1TB F3 32MB Raid 0 (x2) [STORAGE]
GIGABYTE 6870 1GB
AuzenTech Prelude
Prolimatech Megahalems
Corsair HX620


This is my friends build and we go into bios and at first it shows Stock voltage for cpu @ 1.41 but when we reboot it shows it as 1.368v. But when we go to overclock and under load his Vcore is all over the place 1.45v and will bounce to 1.31, 1.200v and obviously fails linx right away. Was wondering if you might know what is causing this problem. His temps are fine as well no problems on temps.

sorry for semi hijack if anything mod can delete this and I will make a thread

nah no problems man, is core boost enabled?

[PatRaceTin]

Hey,

Having a problem was wondering if you might know whats causing it.

First specs:

GIGABYTE GA-990FXA-UD5 F5 Bios
FX-8120
16GB G.Skill DDR3 1600 (1.5v)
Crucial M4 128GB SSD [OS]
Samsung 1TB F3 32MB Raid 0 (x2) [STORAGE]
GIGABYTE 6870 1GB
AuzenTech Prelude
Prolimatech Megahalems
Corsair HX620


This is my friends build and we go into bios and at first it shows Stock voltage for cpu @ 1.41 but when we reboot it shows it as 1.368v. But when we go to overclock and under load his Vcore is all over the place 1.45v and will bounce to 1.31, 1.200v and obviously fails linx right away. Was wondering if you might know what is causing this problem. His temps are fine as well no problems on temps.

sorry for semi hijack if anything mod can delete this and I will make a thread

giga UD5&UD7 has a vcore swing problem

[redrumy3]

nah no problems man, is core boost enabled?

Yeah, I followed pretty much what you changed except only 4.5ghz but vcore seems to be all over the place while stress testing.

giga UD5&UD7 has a vcore swing problem

Really? If so that is really horrible cause 1.42 in bios while under stress it goes to 1.36 to 1.2 and back its crazy :X

[Daveburt714]

Sin: You laid this out so well, hope you don't mind if other folks post their observations...

I've only had a few hours to play with this so far (gear in sig), but I thought I'd post some results.
Keep in mind I'm not looking for any records yet, basically just trying to find stability and efficiency first.

Maybe I'm just stuck on the 53c temp limit I saw on all my PHII's, but I did notice this chip heats up pretty fast
with voltage. Does anyone know where temps actually come into play on the arch? It'd save me alot of testing...
Anyway, to try and limit this effect I've actually set the voltage a little higher (on the CPU) and allowed LLC to let the voltage
droop under full load. This may not be the best option, but it does seem to help a little. Time will tell I guess...
From the v's shown in the SS, it resulted in 1.344v under load.

I also noticed that the Gflops shown in IBT seem pretty low compared to what I saw with Thuban. Either I have some efficiency
problems, or it's just one of those things FX just isn't that good at.

Here's what I've gotten bench stable so far (except P95, still not sure what's going on there ):

[muziqaz]

Hm looks good Dave. 1.38v for 4.9ghz. Waiting for them to arrive in our land here, pick one up as well. Will definitely need this guide to get started

[sin0822]

well the coreboost enabled takes your VID up a lot, a lot more than it should, and then the turbo might be enabled too. It depends how much you have set enabled for all the default settings, but you should know that

dave it looks very nice!

[Daveburt714]

Thanks man. All the power conservation features are turned off as well as Turbo Mode.
I'm sure there are differences with our MB's, but did you notice a particular temp were things got squirley?

So far this chip is ALOT of fun to play with....

[sin0822]

nothing odd with the temps, other than the CPU heats up pretty hot scaled with voltage. But in general i tried to keep it under 55C if i deemed it stable, there were some clocks that were good to 65C but i didn't think that was good for 24/7. I used easytune for my temps tho, and AMD OD.

[Postmodum]

Great thread people! Congrats Sin, once again, for the terrific work

I'm just w8 for the BD to arrive to the stores in Portugal ... i'm going green baby ( not very ECO though )

Really don't care what 99% of people are saying about the BD ... the remaining 1% are those who love to have fun with hardware... and not just "playing" benchmarks ...

[charged3800z24]

Thanks man. All the power conservation features are turned off as well as Turbo Mode.
I'm sure there are differences with our MB's, but did you notice a particular temp were things got squirley?

So far this chip is ALOT of fun to play with....

I'll have mine Monday afternoon and will give it a go. I am also making a chiller for my water setup. Won't have it done by Monday though....
I saw a video on Youttube that looked like with temps and voltage the CPU was getting slower even though the Frequency was going up. I do not know the settings in bios though...

[M.Beier]

Maybe it differs per region you know?

Thanks for the post, i appreciate it. I did a lot of testing, and i am trying to make OC guides for platform i have my hands on, so that i can help others. I like doing them with my reviews, and i am always trying to make them better. I thought overclockers like yourself would like those charts, as they show a few benches and not just one, and they show trends, wether good or bad. Many times the results don't match up across all benchmarks, and that kind of got to me, but nevertheless i posted them.

Thank tho

Scandinavia, but also/most frequent asked directly at HQ in Taiwan.
(I thought it would be cool for showcase - something like this, but also for events)
I am definantly jealouse on the case, I think it is freakin' cool... Asked Tim directly for it, I have asked regional marketing manager, and the person in charge of samples last year for my region as well, making a total of 3 persons.

But deffo think it is good to see the tests of scaling on each setting, please dont get me wrong, the guide part is not against you or anything, but I'd never go for "suggested" voltage settings etc.... I have to play around myself, as most others - getting the touch and understanding for the platform myself.
But I think it is very useful for alot of people, and after all, this is not xtreme-overclocking section (which I most frequently read on this forum)

But really, you have put alot of effort into this, and I personally like this alot more than most other stuff you have done, you use a Gigabyte product, very potent product, can handle what ever you throw at it, and you use it to prove a point on how something works, the AMD chip which is in focus.

Summed up; Cudos for the work, I prefer this approach!

[sin0822]

hey guys so, i am told that the ES CPU i used had some issues with some of the scaling, and that i should retest with retail. So i Will try and do that in the following weeks, if I find any difference i will let you guys know. So test for yourselves, and if you find different let me know.

[Daveburt714]

I've actually had time to sit down and play with this chip today...
FWIW, heres a few observations I've made using custom water cooling:

1) It scales with voltage REALLY well, in fact I haven't found a point where scaling stops!
Should be fun for short benchmarks where temps don't really come into play.

2) I haven't seen a temp barrier similar to the PHII's, but I just don't feel comfortable with temps >60c, so I've held it below that (~ 1.4v on CPU under load).

3) Prime95 seems to be the best stress test. It would run every other bench (including IBT) without problems at my last settings.
At first I thought it was just something funky with P95, but when it failed to encode using RipBot I went back and got it Prime stable.
After a good 30min P95 run (socket temp hit 59c ), I succesfully encoded an HD movie that works perfectly!

4) The IMC gets flakey ~2600mhz. I only pushed the v's to 1.25 but scaling actually slowed drasitcally ~1.20. I quit pushing there due to bullet point #2.
I figured higher CPU clocks were more important and I didn't need the added heat.
OTOT, it clocks memory frequency VERY well! 2040mhz wasn't possible on Thuban and I get the feeling I may be able to tighten timings even more (7-10-7?).

Oh well, after 10 hrs of fiddling I've come up with what would seem to be my new 24/7 clocks :

[sin0822]

nice yea i found the same things, except you only need like tiny voltage increases to hit higher clocks, and it becomes a bit exponential rather than linear.

i wasn't able to do over 2300mhz, it might be because of my memory, but are you talking about CPUNB at 2600?

[Blacklash]

So I'm guessing swinging @ 4.6 with something like an FX 6100 would require water cooling too.

[Daveburt714]

nice yea i found the same things, except you only need like tiny voltage increases to hit higher clocks, and it becomes a bit exponential rather than linear.

i wasn't able to do over 2300mhz, it might be because of my memory, but are you talking about CPUNB at 2600?

I agree 100%! You definetly have to be gentle with v's, they offer alot, but temps get out of control quickly!

Yeah, when I say IMC (Integrate Memory Controller), I'm talking about the CPU-NB. Old school PHI talk I guess.

I'm using a Mushkin (lovo) PSC mem kit and it's doing pretty well. 2550 on the IMC looks like the limit on this kit with reasonable v's though.
I was getting >3ghz on Thuban, but it doesn't look like FX is going to run that high...
Still, it clocks mem frequency much better!

[amdsempron_xs]

@Daveburt714: Something went wrong with the GFlop number? Even an Athlon II X4 @ 4GHz can produce the same number.

[Daveburt714]

@Daveburt714: Something went wrong with the GFlop number? Even an Athlon II X4 @ 4GHz can produce the same number.

Yep, I agree with you man! It's really low... That being said, it doesn't seem like a fluke...
I guess it's just one of those things where FX doesn't shine. Now that I've gotten it stable it's even lower (~42 GFlops).

All I can say is, overall the system feels really snappy even compared to a 4.1Ghz Thuban, reguardless of those numbers...
Maybe it's because it's a whole new uArch and software just hasn't taken advantage of it's features yet.

In all honesty I'm not sure this is a good upgrade from an X6. IMHO though, it's ALOT of fun to play with, and for everyday use it just feels better.

People who only look at benchmarks probably won't be happy with FX, but it's not a POS reguardless of some folks opinions.
Like any brand new hardware it's going to take some time to mature.

[redrumy3]

Sin: You laid this out so well, hope you don't mind if other folks post their observations...

I've only had a few hours to play with this so far (gear in sig), but I thought I'd post some results.
Keep in mind I'm not looking for any records yet, basically just trying to find stability and efficiency first.

Maybe I'm just stuck on the 53c temp limit I saw on all my PHII's, but I did notice this chip heats up pretty fast
with voltage. Does anyone know where temps actually come into play on the arch? It'd save me alot of testing...
Anyway, to try and limit this effect I've actually set the voltage a little higher (on the CPU) and allowed LLC to let the voltage
droop under full load. This may not be the best option, but it does seem to help a little. Time will tell I guess...
From the v's shown in the SS, it resulted in 1.344v under load.

I also noticed that the Gflops shown in IBT seem pretty low compared to what I saw with Thuban. Either I have some efficiency
problems, or it's just one of those things FX just isn't that good at.

Here's what I've gotten bench stable so far (except P95, still not sure what's going on there ):

Awesome. Are you willing to share all your bios settings please. I returned UD5 and grabbed a sabertooth for friends build. So if you would be so kind/awesome to post bios settings please (:

[Manicdan]

@Daveburt

for your 24/7 settings, why not use the turbo feature to give you around 5ghz for a few cores? you should be able to use AOD to set up the higher voltage for the turbo state, and it wont be anywhere near hot with only a few cores using it.