there is more to testing a bios besides running an Everest bench.
tref value is the same so bandwidth should be the same, the bios is far from :banana::banana::banana::banana:
Printable View
there is more to testing a bios besides running an Everest bench.
tref value is the same so bandwidth should be the same, the bios is far from :banana::banana::banana::banana:
Awaiting test results :)
Especially as it mentions increasing memory compatibility.
I agree. I am on bios 1901, and so far, loving it.
I am trying a more extreme 24/7 overclock, but for that, I will need to leave vNB at 1.45v. You think it is safe for 24/7? Temps are not a problem, since I seen to have an insane good mobo, that will never ever pass 39C NB during prime... It stays on 29C-30C idle. And it is aircooled. Do you guys think it is safe to go for 1.45v NB 24/7, if temps will never pass, lets say 42C under prime at that volt?
Here some pics of my system...
http://img388.imageshack.us/img388/6536/full3mj8.jpg
http://img388.imageshack.us/img388/987/full4is8.jpg
http://img504.imageshack.us/img504/2604/temps1gi4.jpg
[]'s
Simps
HAHA Two 38mm Deltas for push/pull on the Ultra120. Thats awesome! :up: :up:
I run one of the pull side, 25mm on the push and I thought that was nuts.
HEHEH!
But those are no deltas... Those are 80-85CMF 38mm Panaflo's @ 30-32dB... And those are the loudest fans I have... Others are all <28dB's, so the system is actually quite, since I use it with closed doors, and it stays inside a cabinet I have...
What about the 1.45vNB for 24/7... You guys think it is safe?
If you can keep the board stable and temps manageable then why not. Do you need that much for starters, I've never needed more than 1.36v Vnb on a P45 and that was with over 470fsb, q6600 and 2gb ram sticks.
Ahh that would explain it! The 38mm Delta i'm using is a ~130cfm 2800rpm unit, got a 210cfm 3700rpm 38mm as well but it sounds like an aircraft taking off its so loud, so it just sits in a box.
I am going for a 65nm quad at 7x500 (3.5GHz) and DDR2-1200 CL5 PL8.
I am doing some tests at that speed, and I need 1.45vNB... I am a bit worried about using 1.45vNB 24/7, but I guess since temps are ok, I will give it a try....
OMG man, those 220CMF 38mm deltas are insane! I believe they are 65dB... But they are really nice for short benchs... Thanks for your repply...
Edit: I just found a youtube link of those deltas's 220CMF... They are 4600RPM. That is really insane! I don't think these are the ones you have... You probably have a quiter one... This one is just insane LOL!
http://www.youtube.com/watch?v=KmDGPFLoPwc
Now, these is the loudest computer fan ever tested... Thats 350CMF 80dB+, and can kill your ears :/
http://au.youtube.com/watch?v=Ws385t...eature=related
So far, I find it more stable at 500FSB on my 65nm quad. I used 1307 before. It is stable actually, I just think I can drop some Vcore and Vdimm, before I call it done. The only voltage I can't drop anymore, is vNB, witch I will have to leave at 1.45v :(
[]'s
Simps
Try messing with cpu and nb clock skews, and see if they help any.
I gave up trying to get a P45 to 500 on a quad at about 1.40v Vnb so maybe you do need it after all. What Performance Level are you trying to use?
I use PL7 12:10 on an X48 and need 1.49v Vnb for 481fsb, so maybe your ball park isn't too far after all since i need at least 1.65v for tighter performance level at this point.
I am running 7x500 = 3.5GHz, DDR2-1200 5-5-5-15 (rest on auto), PL8, stronger. Gives about 10500 MB/s and 49 latency on everest. 1.43vcore on my 65nm quad, 1.38VTT, 2.1 Vdimm, 1.59PLL and 1.45vNB. IBT 20 loops and prime blend 8h stable. But i am still on intense testing, trying to drop vcore and vdimm.
Don't know if you'll get Vcc much lower than that but you can sure try I guess, 1.43v is pretty good for 3.5ghz, and just remember a little extra voltage will be necessary at such high FSB to ensure both dies are kept stable under load! 12:10 the divider you are using is the most stable I've found with quads.
Nice latency by the way :D Impressive for a P45 as they don't usually take much fancy to latencies in the low 50s let alone high 40s!
Yeah... This is what I am getting for my 24/7 settings.
http://img296.imageshack.us/img296/2472/mf2qqn8.jpg
Still a bit worried about keeping 1.45vNB for 24/7...
Anyone here is running 1.45vNB for 24/7?
I do 1.448 24/7 So far so good and I have no objections getting it higher if it will be needed (also Grnfinger is quite high on vNB) - but vNB is a not a big deal. ppl use up to 1.55 :) and they are fine.
Little Update on my random reboots - either CPU or NB clock skews (I set them to 100) did the job.
No more boots. I will test which option exactly but I can bet it is NB Skew as my MOBO is badly warped in NB area (I still wonder if my NB has any contact at all with HS ;) ).
So in next few days when I got more time I will try hitting higher FSBs and will see how far SKEW train will take me.
Cheers
EDIT: Your vNB temps are nice - I can see all those fans - but can you chack what rev u got of NB? Maybe A3 finally?
@ Seban: i'm wondering if exists a method to calculate the skews like the gtls ... can you give me a tip or something?
KURTZ,
Get $50-100K USD together and get yourself a logic analyzer ;)
Thats the method of calculating skews. We just guess them throughn trial and error.
i just plugged in my e7300 for a 4ghz quick run and tested it on orthos. after i tested in orthos, core 1 fails after just 1 sec at 1.35v. i bumped up the vcore all the way up to 1.45v just so it passes the stress test for sure but to my surprise, core 1 still fails after only 1 sec. this means i have some other settings wrong. what should i set the vnb, vtt, GTL, and memory divider to be? here is my current settings:
Code:Processor: e7300
Ai Overclock Tuner: Manual
Ratio CMOS Setting: 10
FSB Frequency: 400
CPU Clock Skew: Auto
NB Clock Skew: Auto
FSB Strap to North Bridge: 266 MHz
DRAM Frequency: 1003 MHz
DRAM Timing Control: Manual
1st Information: 5-5-5-15
DRAM Static Read Control: Disabled
DRAM Read Training: Disabled
MEM. OC Charger: Enabled
Ai Clock Twister: Auto
Ai Transaction Booster: Auto
PCIE Frequency: 101
CPU Voltage: 1.3750
CPU PLL Voltage: 1.53975
FSB Termination Voltage: 1.35175
DRAM Voltage: 2.01200
North Bridge Voltage: 1.35175
South Bridge 1.5 Voltage: 1.50000
South Bridge 1.1 Voltage: 1.10000:
CPU GTL Reference (0): +30mV
CPU GTL Reference (1): Auto
CPU GTL Reference (2): +30
CPU GTL Reference (3): Auto
NB GTL Reference: Auto
DDR2 ChA Reference Voltage: Auto
DDR2 ChB Reference Voltage: Auto
North Bridge DDR Reference: Auto
Load-Line Calibration: Enabled
CPU Spread Spectrum: Disabled
PCIE Spread Spectrum: Disabled
Guys
Any working link for bios 1901 ?
Sergio
Just like mikey said there is no real calculation to be done the method is set and pray ;).
In my case because rest was already tuned and almost stable It was enough to bump Skews to first non default option - 100 (still have to test which in fact gave me stability).
Is your ram stable?
You could try strap 333 to NB as it works nicely for most of us.
Also run small ftt to be sure that you CPU can handle your OC.
but i don't understand one thing: if i put -100 delay the skews what i can improve? can i put my vNB lower? vDIMM maybe? what helps this skew?
Tracks/traces on motherboard which transport informations between buses can have different lengths (and they do - of course differences in lenghts are absolutely minimal - the only exception might by DRAMM slots).
When working on nominal clocks MB projects include those differences, corrections are in printed in BIOS and some tolerance is also included. When you OC you push this included tolerance threshold to limits and this is why u need to adjust it yourself.
On motherboard there are data lines and parallel to them control data lines for this data. Both have minimal differences in length, when you over clock you change the rate at which they travel (you can say you speed them up - although technical you do not, because they already travel at almost constant close to speed of light :)) and this is where mismatch can occur - when data pack traveling by one line can reach target (i.e CPU or NB or Mem Stick) faster/slower then control packet for it which traveling by parallel line then errors will occur.
So you add delay, delay sending data by one from pair of parallel line by lets say 100ps which is 0.0000000001 of the second and it helps get all the packets on time to their targets so there are no errors hence lets hope we are stable.
And coming back to question in theory It should not impact positively your volts but who knows? Tho i would not say so :).
So if you are happy with you OC and you do not have problems you probably should not touch it because there is no point of doing so.
But, if as I did, you face some problems or you cant go any higher with OC then Skews might help you.