AMD Athlon 64 X2 Overclocking Results

[DEPHCONNR]

Would anyone know why XP Pro SP2 only seems to show on CPU Usage History window vs. the 2 it should show when running x2 core...have updated drivers, and cannot seem to get the appropriate response. XP x64 has no problems recog! Thanks

 

[Duvie]

Do you have the right bios installed from asus??

 

[jpeyton]

Looks like I finally tweaked the memory so the system is stable at 2.7GHz. Still Priming right now. Here's a screenie..

 

[ProviaFan]

Originally posted by: DEPHCONNR
Would anyone know why XP Pro SP2 only seems to show on CPU Usage History window vs. the 2 it should show when running x2 core...have updated drivers, and cannot seem to get the appropriate response. XP x64 has no problems recog! Thanks

Eliminate the obvious: Go to the Performance tab in the Task Manager, select View > CPU History > One Graph per CPU

I don't think that's it, though. Did you attempt to convert to SMP drivers on an existing install? Did you go into the Device Manager under Computer and update the driver to "ACPI Multiprocessor PC"?

 

[jpeyton]

Got another X2 to play around with. A 4200+ that I got from MonarchComputer. CPU code is CBDHE 0524RPAW.

Results so far.

That looks to be its limit on air; the same as my 4400+.

 

[AsianriceX]

Originally posted by: jpeyton
Got another X2 to play around with. A 4200+ that I got from MonarchComputer. CPU code is CBDHE 0524RPAW.

Results so far.

That looks to be its limit on air; the same as my 4400+.

That's some fantastic results!

I can't wait to get my hands on my 4200+ from Monarch so I can mate it with my XP-120 and A8N-SLi Premium and get to overclocking!

 

[imported_Vodka]

finally received my 4400+ and i'm very happy with the results i get. see bottom

 

[ScottFern]

I just last night got my 4800+ up to 2.8Ghz at 234X12 at 1.45v vcore and 2.5-4-3-7 timings. It is SuperPi 32MB and SP2004 Blend test 9 hours stable.

However, my SuperPi time was horrible at around 34 minutes. I am not sure what is the cause of this, but I think it may be the timings are to loose?

jpeyton: I saw you had your 4400+ up to 1.53v vcore? How in the hell do you maintain that? I have a Koolance 3/8" watercooling setup and at 1.5v vcore I am at 40C idle and 58C load closer to 60C. It is scary see it sit so close the thermal peak of 65C.

 

[freethrowtommy]

I think jpayton's core was 60 C at that vcore... so that sounds about right for what you have also.

 

[imported_Otter]

Originally posted by: jpeyton
Got another X2 to play around with. A 4200+ that I got from MonarchComputer. CPU code is CBDHE 0524RPAW.

Results so far.

That looks to be its limit on air; the same as my 4400+.

Jpeyton, did you use the same cooling setup for both CPU's. And what is it?

 

[jpeyton]

Yes, I used a Zalman 7700Cu for both CPUs, and temps are high but stable. I'm not afraid of pushing these CPUs because I honestly don't think 65C is a hard limit max temp. That said, I probably will never breach 70C (which was my highest recorded temp on a 95 degree day). If it gets anywhere near that, I'll drop the CPU down to 2.6GHz for a while.

Scott, that's great you got to 2.8GHz on water. Those temps seem a little high for water though; what is your ambient? SuperPi times will go down if you tighten your timings; seeing as you have some good TCCD memory, you should try 280x10 or 255x11; run the RAM 1:1 with 2.5-3-3-6 or -7 timings.

 

[Acanthus]

Originally posted by: dguy6789

Originally posted by: jpeyton

Originally posted by: g33k

Originally posted by: jpeyton
L]

So with mid-end air cooling and a safe VCore of 1.52V, we have the equivalent of a AMD Athlon 64 X2 5200+, or two AMD FX-55s in one socket. Not bad.

Wow great results! Not bad? That's awesome! But if you consider the Zalman 7700cu mid-end air cooling, what would a high-end cooler be?

According to AMDZone, high-end air is a ThermalRight XP90C with a 90mm 108CFM Vantec "Bleed From Your Ears" Tornado. I agree, that is high end, but it's also completely unrealistic for anything but benchmarking.

I really need to get into some light water cooling. I've been eyeing the ThermalTake Big Water Kit for a while, and I think I'll jump on the bandwagon.

You wouldn't want something like water cooling. If you are trying to overclock, watercooling will not give you much more than a top end air cooling setup. W/C is good for high end air cooling performance without the noise. But on the downside, there is a ton of upkeep and maintanance required, expensive, and even a slight leak can fry your pc.

I have done water cooling for about 8 months running an Athlon XP 1700+ tbred at around 2.4ghz. To be honest, all the trouble it was, it wasn't worth it. I would rather have had just some good air cooling and not worry about refilling the water every month and algae growing and such.

The zalman resorator has none of those problems, and does great on my overclocked prescott flamethrower.

 

[imported_PrinceGaz]

I got my X2 4400+ last week and am currently seeing what it is capable of. It's on a DFI nF4 SLI-DR mobo and has decent air-cooling (XP-90C, 50cfm 2400rpm 92mm fan, Arctic Silver 5) so it should overclock okay.

So far I've verified using two instances of Prime95 (one on each core), that the highest stable speeds are

1.35V (default) - 232x11 = 2.55GHz (44C core, 25C room temp)
1.40V - 237x11 = 2.60GHz (47C core, 25C room temp)

I haven't verified the following as exact maximums but they seem likely judging from how long Prime95 ran for before failing at those voltages and higher speeds

1.45V - 242x11 = 2.66GHz (51C core, 27C room temp)
1.50V - 246x11 = 2.70GHz (est 55C core, 27C room temp)

I don't plan on pushing it any higher than 1.50V as 55C is approaching the 65C meltdown point :Q

 

[seanp789]

I also have the 4400+ and have been overclocking but I'm running into some problems, I was wondering if you could chat with me about some of the problems you may have run into.
I was wondering why you decide to lower your cpu multiplier? was it just to get nice even numbers or did it actually affect your stability.

CPU: amd64 x2 4400 @2.42GHz *now 2.508*
PSU: Antec neopower 480w
Asus a8n-sil deluxe (bios 1013 final)
Nforce 6.66 drivers
Mushkin 4x 512mb XP 2-2-2

HSF thermaltake PIPE101 w/ silentcat 90 (51cfm)

My specs:
PU Clock Speed: 2420MHz
CPU Multiplier: 11X
FSB: 220MHz *now 228*
LDT Multiplier: 5X
HTT Clock: 1100MHz
Memory Ratio: 1:1 (DDR456)
Memory Timings: 2.5-3-3-10
PCIe Clock: 100MHz
PCI Clock: 33MHz

VCore: 1.487V BIOS, 1.50V Actual (Idle), 1.50V Actual (Load), 1.488V (powernow @ load), 1.2V (powernow idle)

First of all my Vcore. any setting below 1.45v in bios will give me 1.38 or less. Ive had 2 different powersupplies so its definitely the board thats producing the vcore gaps. I upgraded my bios but I havent seen any changes.

bios | actual

anything below 1.45 is less than 1.38v
1.45 | 1.40
1.465 | 1.42
1.475 | 1.50
1.485 | 1.50
1.50 | 1.52
1.525 | .....

so anyways you get the picture. there is an annoying gap in voltage between 1.42 and 1.50. @ 1.5V my load temp jumps from 57C to 63C




A side note, the new amd powernow driver works very well with the X2, one thing I noticed though is that if you are using singlethread programs with no affinity the cpu multiplier will flux between 9 and 11 meaning its either not running at full speed or with the windows load balancing it doesnt need to. I suspect thats its just not running at full speed, but ill have to run some benchmarks.


The biggest problem I'm having is that if I dont have enough voltage, CORE0 fails prime in less than 5 minutes. If i turn the voltage up, The temps are just way too high and also eventually fails. Its only core 0 that needs the higher voltage, core 1 runs fine OC'd even at stock voltage. Since the cpu isnt crashing I suspect it is the 1MBcache in core 0 that is having problems.


So right now im at 2.42Ghz I was able to get 1.48v vcore using the powernow which shaves about 2C off my load temp


Oddly enough I have no problem gaming at 2.64 Ghz for hours and hours, but if it cant pass prime 95 , then I dont want to risk it as my usual setup.


Are there any test programs that strictly memtest cpu cache?


*Update*

Tried lowering the cpu multiplier and uppign the fab and it bsoded on me.

switched it back and upped the fsb from 220 to 228 and it is rock solid. ~100Mhz increase for no apparent reason.

 

[piddlefoot]

Thats awsome, l have a 4400+ on order, and now l cant wait even more...
Do you run FRAPS , l would love to know how it performs recording a game with all that cpu power ?
My 3200+ 32 bit single core can record at 20 to 35 fps in Battlefield Vietnam, on a server.
Fraps can record at 60 fps if the pc has the grunt...

 

[seanp789]

Originally posted by: piddlefoot
Thats awsome, l have a 4400+ on order, and now l cant wait even more...
Do you run FRAPS , l would love to know how it performs recording a game with all that cpu power ?
My 3200+ 32 bit single core can record at 20 to 35 fps in Battlefield Vietnam, on a server.
Fraps can record at 60 fps if the pc has the grunt...

Yes I have done though lots of tesing with fraps. Using dual core completely removes the CPU bottleneck. That said, I discovered that fraps recording at 60fps only uses about 9% cpu on the second core.

Here is where it gets tricky, you must set the affinity of your game and fraps to run on separate cores manually or you will lose an additional 3-5 frames while recording. It was not the 50-100% improvement I was expecting, but it did teach me a little about how fraps actually works so I will try to explain...

Ok, if your dual core system is capable of 80fps fraps will record at exactly 40fps. As your fps changes, so will your recording fps. with a single core system there is that roughly 9% overhead (percentage will vary depending on your proc speed, mine is 4400 @ 2.508) so you will not get the exactly half fps.


[This is my understanding of fraps, it may or may not be accurate.
Now, with dual core you have set fraps and the game on different affinities. So you shoudl be able to record at your full fps right? NO! WRONG! Fraps appears to record by allowing a frame to be rendered and sending it to the display. Then it eats up a frame cycle either rerendering or just sending a frame to your computer back through the PCI-E bus. This effectively halves your fps. *NOTE* earlier I said "fps Capable" not fps actual. So Lets say your BF2 caps out at 100fps, you should only be able to get 50 fps right? NO! you can get 60 fps if your system is capable of rendering the same scene at 120fps. half your gpu cycles will go into recording, and half will actually be rendered in-game. the bottom line is with no other bottleneck (cpu,HD,etc..) , fraps can only record as a max, half your potential fps.



Also I have had some better like Oc'iong in the past day. These 4400+ badly need to be burnt-in i went from 2.4 to 2.5Ghz stable without actually changing any other settings. I jsut had to wait a day. How weird is that? The day I got I could only get 2.2 Stable. so im not sure what my max is but i might give it a week and then try to up it again.

 

[piddlefoot]

thats the best piece of info ive seen on fraps with x2 core, lm just drooing now.
Thanks for that , l know the investment is worth it now .
Cheers for your time and effort.


paul...

 

[imported_PrinceGaz]

Also I have had some better like Oc'iong in the past day. These 4400+ badly need to be burnt-in i went from 2.4 to 2.5Ghz stable without actually changing any other settings. I jsut had to wait a day. How weird is that? The day I got I could only get 2.2 Stable. so im not sure what my max is but i might give it a week and then try to up it again.

I don't know what thermal compound you are using, or how long you have been testing your box, but apparently the Arctic Silver 5 I'm using suggests it takes several days and the computer being powered-down for a while every so often to become fully effective, and maybe other compounds are similar. The info for AS5 reads

"Due to the unique shape and sizes of the particles in Arctic Silver 5's conductive matrix, it will take a up to 200 hours and several thermal cycles to achieve maximum particle to particle thermal conduction and for the heatsink to CPU interface to reach maximum conductivity. (This period will be longer in a system without a fan on the heatsink or with a low speed fan on the heatsink.) On systems measuring actual internal core temperatures via the CPU's internal diode, the measured temperature will often drop 2C to 5C over this "break-in" period. This break-in will occur during the normal use of the computer as long as the computer is turned off from time to time and the interface is allowed to cool to room temperature. Once the break-in is complete, the computer can be left on if desired."

I have also heard that CPUs themselves can get slightly better when pushed towards their limit for several days, but I've no idea why. Maybe it's a combination of those two factors contributing to your better overclocking success. I'm assuming of course that your room temperature hasn't fallen dramatically over the last few days

 

[Duvie]

I dont agree with the theory of burn in process and most thermal compound burn in is like 2-3c max so I am not sure how that would gain so much more...stumped there....

I can say one thing for sure though....I have bumped up 3 HTT from where I first thought my limit was to now being at 2.584ghz at 1.456v-1.488v (idle/load)...I am a slow stability checker...usually like to take 1 week of testing before I will bump it up just 1 notch (after iniital 2 weeks of fiding rough limits....

However I have NEVER seen a BSOD yet. I have had a freeze from using A64tweaker and changing my trcd...I have had some superpi "not exact round" errors and a few prime errors, but not one crash from running my system....

This thing will run benches at 2.7ghz with 1.504-1.538v (idle/load) but will give prime erros quickly and EEU's in folding at home testing....My temps in that rage are only about 57-58c...so heat could be an issue. Now that I think of it I was only though t to be stable at 2.552ghz when I had my TT venus 12 and hit max temps in the 55-56c range instead of the 50-51c now....I am only using cheapo white thermalright stuff now...maybe I should use some AS5 and see if I can get the temps down to under 50 and see if 2.6ghz is a reality at this vcore...

 

[imported_PrinceGaz]

I can take my 4400+ up to 2.68GHz and 1.5V, and it is rock-solid with two instances of Prime95. Judging from the limits I reached at 0.05V inrements, and the temperatures of them, I could probably pass 2.7GHz with this Thermalright XP-90C but I don't want to run too close to the edge (above 55C is scary for me when the meltdown temp is 65C).

All the below were verified with two instances of Prime95 for about twenty hours, and the room temp was between 25-27C.

232x11 (2.55GHz) at 1.35V (+19C over room temp)
237x11 (2.60GHz) at 1.40V (+22C)
241x11 (2.65GHz) at 1.45V (+24C)
244x11 (2.68GHz) at 1.50V (+28C)

However I booted into Windows fine at 250x11 (2.75GHz) with just 1.45V and did this and that without any problems before firing up Prime95 which immediately failed. What is clear from those figures is that I was getting diminishing returns from cranking up the voltage the higher it went, possibly because the temperature was nearing 55C.

Prime95 is really tough and long-term stability is my main concern so I've settled on 2.6GHz (260x10) and 1.475V which gives me a healthy margin above what Prime95 was stable at. The reason for the 10x multiplier is so I can run my memory at DDR433, which is what the 5:6 (DDR333) ratio actually ends up being at that speed. My 1GB sticks refused anything above DDR470 so I couldn't run 1:1 but that is irrelevant anyway with the A64 as latency tests prove, so DDR433 2.0-3-2-7-1T provided optimum bandwidth and low latency for me after I juggled the FSB, CPU multipliers, and memory dividers.

In all my testing I never saw a BSOD either, but tbh unless your memory is flaky Windows itself will have no problem with speeds well above what Prime95 will even run five minutes at.

I've no idea how good the stuff in the syringe that came with my XP-90C is, I imagine Thermalright include stuff that is fairly good (better than what you might get bundled with a mobo for instance), and within a degree or two of AS5, but I'm only guessing and have no evidence. Of course when it comes to raising the voltages for more speed without approaching meltdown, every degree counts so that AS5 might buy you a few tens of MHz extra.

 

[Duvie]

I agree but coming from my winnie where it wasn't flaky memory but a picky memory controller...I really think they have dialed the memory controller in much better...

yep I am going to take a shot at the AS5....may still look at getting a different fan...perhaps a 90 to 120mm converter and a 120mm fan....I can tell you it has done wonders for my memory...I believe I was overheating at 2.85v with my gskills but with the huge overhang of the XP90 and the downdraft of air I am running this memory much better...It may have something to do with those lack of BSOD as well....

You definitely got a nice chip...The last place I could use 1.44v was at like 2.5ghz and 1.34-1.35v was at 2.4xghz...

 

[Velk]

Originally posted by: seanp789
I also have the 4400+ and have been overclocking but I'm running into some problems, I was wondering if you could chat with me about some of the problems you may have run into.
I was wondering why you decide to lower your cpu multiplier? was it just to get nice even numbers or did it actually affect your stability.

CPU: amd64 x2 4400 @2.42GHz *now 2.508*
PSU: Antec neopower 480w
Asus a8n-sil deluxe (bios 1013 final)
Nforce 6.66 drivers
Mushkin 4x 512mb XP 2-2-2

HSF thermaltake PIPE101 w/ silentcat 90 (51cfm)

My specs:
PU Clock Speed: 2420MHz
CPU Multiplier: 11X
FSB: 220MHz *now 228*
LDT Multiplier: 5X
HTT Clock: 1100MHz
Memory Ratio: 1:1 (DDR456)
Memory Timings: 2.5-3-3-10
PCIe Clock: 100MHz
PCI Clock: 33MHz

VCore: 1.487V BIOS, 1.50V Actual (Idle), 1.50V Actual (Load), 1.488V (powernow @ load), 1.2V (powernow idle)

Never increase the HTT clock above 1000MHz. As you increase the FSB, drop the LDT multiplier so that HTT remains at or below 1000.

 

[chinkgai]

my final results for my x2 4400

 

[imported_PrinceGaz]

Originally posted by: chinkgai
my final results for my x2 4400

Jeez, what sort of cooling have you got on that 4400+ to be putting 1.68V through it and running it at almost 2.9GHz ffs, and only averaging 42C?!? Can water-cooling alone manage that or have you got a peltier in between as well?

 

[chinkgai]

water cooling alone

black ice xtreme III
danger den tdx
swiftech mcp655

cpu is only thing in the loop

i run 2.8 instead of 2.9 daily and loads at 38C