Printable View
Which benchmark, the cpu-z latency test? Oh and its starting to seem like different northbridge multis are better at different htt settings cause of when each of them change internal latencies, but I dunno. I'll let you guys pick through it when I'm done, which should be sometime tomorrow or tuesday at the latest. School got canceled for me:D !
Hows doing you Lightman!? :DQuote:
Originally Posted by Lightman
Any new better BIOSes for you guys?
You and I have done these tests a while ago with 9500s so when you get the BE which will oc on NB with multi's because it's made to and all tools read the values in-sync properly, then try and run some comparisons between 9x multi and say, 12x multi, or 1.8G NB and 2.6G NB. Like 32M, Sandra and WinRAR ;)
BTW, I flashed a BIOS yesterday and it all seemed to go fine but... since then I've had no boot. :(
Every time I boot it gives me a checksum error and puts me in BIOS -> reboot same -> reboot same. Doesn't let me go ahead so can't flash anything either. :down:
Oh and with K10h... HT multi is only locked by NB multi BTW. :p:
CPUZ latency is no good here. It uses CPUZ values to compute, so if CPUZ shows high NB, then it'll give NB low latencies. Run Sandra Latency and Bandwidth comparisons instead. ;)
That sucks about the bios flash KTE hope you get it worked out.
Sounds about as bad as me breaking the pin on my Phenom :sQuote:
Originally Posted by KTE
Last time I tried to install sandra it froze my computer every time I tried to open the installer, but I can try again.
Edit: 10x nb data is done, screenies will be up soon for what was requested.
Yeah booted up now, had to mess about a lot. A lot of hassle just to get this system to boot.
NOW, all the perf looks real bad. The OS is totally gone, my profile is missing and some other new one loads up with my old profile name with none of my apps/docs on there anymore!! :mad:
Just because of flashing a bad BIOS. I didn't even try loading Windows with it until I had successfully flashed another because it wouldn't even get to that stage before, so although I'm confused, it's exactly what happened.
Somethings just not right here. Registry corruption for sure but how to fix it, I can't make out yet as it destroyed all the System Restores too although I had saved them. AV/FW doesn't work anymore neither. So far this has took me hours and hours to fix already. :(
Also one odd happening: the chip is idling at 42C stock all settings now with the same everything used and loads at 203W AC! That's like 35W more than before. :confused:
Oldguy932: Sandra doesn't have to be installed, you can run the zip file. ;)
Dang man... best of luck fixing all that.
http://www.geocities.com/ranger_x3/P...mparison9x.jpg
http://www.geocities.com/ranger_x3/P...parison10x.jpg
There are the ss I promised. Right now I'm debating whether I should even bother doing the table for 11x just cause of how time consuming the benchmarking is. The longest anything goes is 10 minutes, so its not like I can really go out, do something, and come back, or even just watch a tv show.
Well that doesn't show anything that we don't know long before. Just 260x11.5, x9 NB, x9 HT and CPUZ giving scores based on what it's reading (synthetic). Which I've said a long time back. Can see the actual multi with your HT very clearly.
Guys who're showing 5.xGHz NB in CPUZ will get sub 20ns in CPUZ Latency. But not in Sandra or anything other.
Actually Super-Pi 1m was faster with the 10x nb setting, only settings it wasn't was at 220 and 230 htt and it tied at 240 htt. Not by a massive margin, but around .05sec better. I'm taking a break from benching for that massive table for right now, I think I'm gonna try and get my 3850 ranked on hwbot, which will be a challenge.
Oh, and memory reads are overall faster, writes and copies are about even.
Well when you beat 24.5s 1M at 260x11.5 2.99G, I'll then know you had a faster NB than 9x for sure. NB there is 2340MHz in both scenarios.
As I said: http://img225.imageshack.us/my.php?image=latnecylk1.png
False.
All that would take is some faster memory, if I could keep my memory at 667 divider I'd probably get at least 25.0, but the memory controller seems to be a voltage grubber like the nice clocking opty.
Edit: I tried moving back up to the 667 divider and setting htt to 240. It reset all my bios settings... So I went in and I thought I changed them all back, 3DMark06 just dropped major fps...11.2k before it reset, now its 10.2k. I need a friggin break...
Do you still have a 9600BE?
C3 DDR2-610 = 5ns
C4 DDR2-800 = 5ns
C5 DDR2-1000 = 5ns
C6 DDR2-1200 = 5ns
C3 DDR2-692 = 4.34ns
C4 DDR2-692 = 5.78ns
C4 800 is not much better than C4 692, although it would make a difference, but only in very memory intensive benches (like 32M).
I do still have the 9600 BE, but my regular just runs faster overall so I don't use it now. And with the 667 divider it would be doing around 880 or so. Maybe I should play with the timings some, haven't done much of that yet with this memory.
I just got the vapo mount in the mail. I probably won't have results till tomorrow though, stupid social plans...;)
How much did it set you back?
$120 without tax. Hopefully it was worth it...
I will get mine tomorrow hopefully!Quote:
Originally Posted by KTE
I've waited for new delivery from AMD factory to show up in U.K. hoping it will be from this year production. I count for small manufacturing improvements, but this is lottery so I need to wait and then test new CPU ;)
I will report back tomorrow :)
I think I need to get a thermal probe and mount it, cause there is no possible way its idling at -13C in windows...cause idling on air at the same speed was 42C max according to the board. These temp monitors really are screwed up... But yeah it boots perfectly, no coldbugs noticed at all.
Edit: It's starting to annoy the crap out of me. Since my mobo turns the computer completely off everytime I go into the bios, I have to turn off the psu as soon as I hit save settings or else it'll turn off the vapo and gives me an overclock that doesn't start. The temperature thing seems like its not getting good enough contact, so I'm gonna have to check on that. On the plus side stock voltage got changed to 1.2 somehow from 1.25. Here we go again...
Edit 2: Stupid heater thing, it was set to 100% for some reason last time I used it. Turned it off, now its down to -19C, but who knows how accurate that actually is...
Alright i'm new in the whole overclocking arena so please be patient with me. I have a phenom 9500 with a MSI K9A2 platinum with bios 113. I also have 2x2GB G.Skill DDR2 1000 ram. Just have a quick question and i'm hoping someone could answer. If I overclock my processor to 2.4ghz using the FSB vista takes forever to start where as if I boot in x64 pro it starts without a problem. That overclock was stable in xp but for some reason vista doesn't care for it much. Any idea why?
vista = crap is alot of the reason IMHO but are you using sp1 on vista?
On a side not KTE what bios was it that craped the system for you? it wasn't one you where fudging around with was it :D
Also I added fans to SB, NB, and the power area(or whatever its called)...
Think I may try to play with it again soon not 100% sure as I stated I'm happy at 2.55 ATM.
I figured I was going to get something to that extent. I personally haven't had any major problems with vista and I actually like it as opposed to xp...i'm strange I suppose...but no i'm not using SP1...it doesn't come out until the end of march? right now it's in beta?Quote:
Originally Posted by jonspd
Oldguy, I've been doing alot of testing with your theory about the higher NB multi on a standard 9600. To be honest I was surprised it would even boot with a NB FID higher than 5 (9x)....
Here are my results. Numbers are "HT Ref x CPU Multi x NB Multi x HT Multi"
This is a ss from my everyday stable clock:
SuperPi may be a little low, I didn't delete the data files before this run, the rest of the runs are clean....
250x10x8x8
http://i30.photobucket.com/albums/c3...250x10x8x8.jpg
223x11x10x9
http://i30.photobucket.com/albums/c3...11x10x9_PC.jpg
229x11x11x9
http://i30.photobucket.com/albums/c3...11x11x9_PC.jpg
232x11x11x9
http://i30.photobucket.com/albums/c3...x9_PC_play.jpg
These are just my observations, but it seems like a NB Multi higher than 9x on a standard 9600 only causes the monitoring software to get screwy, especially CPU-Z and Latency.exe. Although AOD gets kinda wierd too, NB "Current Speed" seems to be right, while "Target Speed" is way out of line... It's just the opposite for CPU speed, "Target" seems to be right while "Current" is way out of line...
I say that because my chip always gets squirley around 2.55Ghz, just as it did at 232x11... Maybe you did hit the CPU Lottery twice Bro...;)
Wanna sell that BE? I live in Ohio too! :D
Bob...
As long as it runs stable I wouldn't sweat the slow Vista boot, I think it's just the nature of the beast...
Bloated beast that it is... ;) I still miss 2000Pro, I'd be willing to argue that was the best OS Microsoft ever released .........
Some Vista users might find the below useful.
http://www.pcstats.com/articleview.cfm?articleID=2238
Oldguy: Stock settings, 40-45C on the stock AMD cooler at 20-26C ambient is exactly correct core temp readings. ;)
They're not external but internal readings. How are you measuring evap temp?
-12C internal looks like probably -30C evap. Maybe you need to check your contact and vapo again and check the load calibration is good.
Jon: it was a tester BIOS. ;) ;)
Note: XS keeps timing out for me so I can't check/reply much yet.