Really, interesting, could you share them numbers with us plz ?Quote:
Originally Posted by garyd9
Printable View
Really, interesting, could you share them numbers with us plz ?Quote:
Originally Posted by garyd9
To test, I put my 580 back to stock speeds, and ran the only benchmarks I have: unigine heaven, dirt2, and one of the benchmarks that comes with Civ5. All the numbers were lower (I didn't write down the x8 numbers.) The one that really nagged at me was unigine heaven - which dropped from a very consistent 49.5 average fps to the low 40's. (It was consistent over 3 runs.) The others were less significant (or less consistent over multiple runs.)Quote:
Originally Posted by Leeghoofd
I'll admit it wasn't a scientific or very well controlled test, and not one I'd want to publish in any type of review, but it was enough for my own personal use and enough to sway my own opinion.
I ended up taking the xonar essence STX out of the machine and I'm using the on board realtek sound for now. I'm hoping that the next round of P67 (or Z68) m/b's have more PCIe options for those of us trying to leave legacy PCI behind. (The irony is that the Essence STX board is actually a PCI board with an extra chip bridging it to the PCIe x1 interface... the same is true for my EMU 1212m sound card.)
Take care
Gary
Well, those results sound strange. You shouldn't see such a difference :/ But I don't know of any method to verify if the slot is in 16x, 8x, or less.
About the PCI-E lanes, don't expect too much as it is inherent to the P67 arch (8 PCI-E lanes according to wikipedia)
GPU-Z shows the number of lanes being used, but I don't know of anything that actually measures the bandwidth.Quote:
Originally Posted by Khoral
As for the limitation on the number of lanes, there are actually enough lanes to support what I want, but the motherboard (or perhaps the BIOS) doesn't let me allocate them the way I want (or has the slots in a physical location that isn't working for me right now.) I started to consider the p8p67 deluxe (as it has more options with the PLX chip), but it would have the same PHYSICAL issue as the Pro. Perhaps the WS Revolution would work for me, but it'd be cheaper for me to buy the legacy PCI version of the Xonar Essence!
I could also stop being so spoiled about my sound cards. ;)
Your English is much better than my French, so no apology needed.Quote:
Sorry for my bad english, I'm trying to improve it
Take care
Gary
?Did you look at some other manufacturers too ? Like Gigabyte with their UD5 and UD7 ?
Well for me I would just say, look at your voltages when running linx, something funny happened to me and the board had volt spikes beyond 1.5v, which after a short while fried the cpu and board.Quote:
Originally Posted by total90
If volts and temps are good under load then I would say all is well.
Running a 2500k at 4.9ghz, running Linx, under load the voltages peak at 1.458v - is this okay for everyday use with Watercooling, temps are 68degrees @ max load
.
well thanks mate :)Quote:
Originally Posted by maverik-sg1
4.8GHz is a no-go for my CPU, tried everything, Prime95 Blend usually shuts down the computer around 3 hours, Small works without problems.
Altough it seems stable for 24/7 usual work, didn't shut down even once.
These are all the settings I tried, don't seem to make difference from default:
VRM Frequency: Manual
VRM Fixed Frequency Mode: 350
Phase Control: Extreme
Duty Control: Extreme
CPU Current Capability: 130%
I use LLC Regular, but I don't see how this could be a problem, works fine with 4.6GHz, even more voltage doesn't help Prime Blend stability.
Edit: actually, it seems 1T/2T memory command rate was my difference between 13xx and 15xx bios. with 1T I lost 1 multiplier at every voltage range as apposed to 2T. so, ignore some of my comments below. though I'm not sure if one bios was better than the other still.
my Deluxe/i5 2500k has been rock stable 4.5GHz for about 4 weeks now. I used offset but required more voltage during load, so I use offset off now. still waiting for new BIOS from Asus. seriously where are the bios updates? I liked 13xx bios better, seemed to get better results. (lost about 100mhz in every voltage range, I think related to memory somehow) but can't downgrade now. and I still want the offsets fixed, it overshoots the mark too much and still drops. gigabytes are more on the mark and less jumping around. so, come on asus...
Anyways here's my 24/7 4.5GHz settings. I'd love to get more like 4.7-4.8 if not 5.0 though. waiting for new BIOS though.
Vcore: 1.355 Bios, 1.64 idle, 1.352v under load w/ prime95
RAM: 2133MHz, lowered to 1T stable. no extra VCCIO required.
DIGI+ VRM Load-Line Calibration: Ultra High
DIGI+ VRM Current Capability: 140% (I could probably lower this)
DIGI+ VRM Frequency: Auto w/ Spread Spectrum
DIGI+ VRM Phase Control: Manual; Ultra Fast.
DIGI+ VRM Duty Control: Extreme
EPU: Enabled/Auto
CPU Spread Spectrum: Disabled
C3/C6/EIST: Enabled
my previous 2500K had memory controllers crap out too after awhile. combination of 1T and high volts/overclock I think did it. I'm hesitant to OC my new one past 1.39v because of that.Quote:
Originally Posted by Darkstormz
I kinda do. Have you tried other lvls for LLC?Quote:
Originally Posted by Gr1mR34p3r
I will change to Sandy next week but have a question: I grab a i2500k but dont know if I go for -M Pro or Deluxe.
Can any one help me here? I will OC the system ....
How important are the extra 4 digi on deluxe compared to pro for OC?
people seem to be reaching 5GHz even on the sabertooth (8 phase) so, I wouldn't think it matters so much. maybe only in the real extreme 5.2+ ranges. Asus has a guide here somewhere and they pretty much say the same thing.Quote:
Originally Posted by Vega
Edit: also since you have a 980X, it might be best to wait for ivy bridge and skip sandy altogether.
Yeh, but I cant use my 980X on ivy, right? It will be a complete new socket or is it the new Sandy (1155)? If no then its no way i sell it! This cpu is a golden one....Quote:
Originally Posted by LeetMiniWheat
Will the new X68/x78 handle my 980x?
You are correct, the 980X CPU will only work on the X58 chipset on socket 1366. Sandy Bridge, is as you mentioned on socket 1155 and SB-E will be on socket 2011 using the X68 chipset ( http://www.techpowerup.com/138087/In...-Pictured.html ). As you can probably surmise, the X980 CPU won't work on either of these chipsets (it's uses a different socket). Ivy Bridge is the die shrink of Sandy Bridge, but whether IB will work on exists Sandy Bridge MB with a BIOS/UEFI update remains to be seen. If you're really set on keeping the 980X, then you are pretty much stuck with your current MB as it won't work in the any of the motherboards designed for either SB or IB.Quote:
Originally Posted by Vega
BTW, I take back anything bad I said about 1503 bios. I thought I was getting worse overclocks compared to my 13XX, but it turns out I just needed some more VCCIO to run my memory @ 1T. (I'd mostly tested 13XX with 2T)
I'm 12hrs prime stable 4.7GHz @ 1.374v load w/ 1.1v VCCIO (still testing lower VCCIO though)
This is about the max VCore I feel safe with (under 1.38v) after my last sandy's mem controller died.
wonder what else I can try to squeeze more MHz without going over this voltage wall? I'm thinking of tweaking VCCSA and CPU PLL, but not sure which direction I should be going in (up? or down?)
also, i've heard people say they can get higher clocks with lower load volts with LLC off, but wouldn't that mean higher overall non-load volts? or higher half-load volts if the load isn't enough to vdroop significantly?
Yes, non-load voltage is higher, but only by a small margin, around 0.01V, but load voltage with LLC off is 0.03-0.04V lower than LLC High.Quote:
also, i've heard people say they can get higher clocks with lower load volts with LLC off, but wouldn't that mean higher overall non-load volts? or higher half-load volts if the load isn't enough to vdroop significantly?
I would say to 4.6GHz you are best with LLC Regular, but 4.8GHz isn't stable with Regular and requires LLC High, haven't yet tried LLC Medium though.
Lower as far as possible, but these voltages only bring lower power bills, nothing more. Actually, running Prime at high CPU frequency to figure how low you can go, you're actually doing nothing good, not to costs and not to CPU health, I guess it's just for personal pleasure.Quote:
I'm thinking of tweaking VCCSA and CPU PLL, but not sure which direction I should be going in (up? or down?)
To give you an idea how much, I am stable with:
4GHz
VCCSA Voltage: 0.80000
VCCIO Voltage: 1.00000
CPU PLL Voltage: 1.52500
PCH Voltage: 0.92000
4.6GHz
VCCSA Voltage: 0.81250
VCCIO Voltage: 1.00000
CPU PLL Voltage: 1.52500
PCH Voltage: 0.94000
4.8GHz
VCCSA Voltage: 0.81250
VCCIO Voltage: 1.03750
CPU PLL Voltage: 1.53750
PCH Voltage: 0.95000
Determining CPU PLL is the easiest, if it boots you are 100% stable, Prime 12h+, LinX 20 passes etc., regardless of the CPU frequency.
sorry maybe a little off topic, but I've had no luck getting my L041C123 past 4.6-4.7 with <1.38v.
I got a 3rd 2500k today, a costa rica chip, batch # 3101A127. first impressions: not impressed. been up all night with it and it seems to clock identically to my L041C123 malaysia chip. (see sig for clocks/volts range) but it refuses to do 5.0GHz <1.45v (probably needs 1.5v) my first 2500k was horrible and would barely even do 4.6-4.7 before it's memory controllers died.
how hard is it to find a chip to do 5GHz under 1.44v? /sigh (preferably 1.4v)
hope my P8P67 Deluxe isn't limiting my clocks somehow. I'm still not convinced 1503 bios is better than 1305 wonder why Asus hasn't released a new BIOS lately, they must be working on Z68/X68?
how hard is it to find a chip to do 5GHz under 1.44v? /sigh (preferably 1.4v)[/QUOTE]
Pretty hard, I know someone that went through 20 chips to find one that could do that.
What specific problem are you having on 1305 or 1503? If everything is working ok for you, it's a good sign.Quote:
Originally Posted by LeetMiniWheat
-Raja
From my experience with ~12 2600k chips. Only 3 have been under 1.45V loaded for 5ghz prime and 3d stable. Only 1 of those has been linx w/ avx stable which was at 1.42V loaded.
My 2500K does 5ghz linx w/ avx stable @ 1.42V and 4.8ghz @ 1.35V (my 24/7 oc). The batch # is L045A848. This is the only chip I bought. It'll boot windows and run superPi at x54 multi with 1.5V. I'm using a P8P67 Pro and have the same results with 1305 and 1502 bios. Guess I got fairly lucky.
my new 2500K was priming fine for 6 hours at 4.9 when it was brand new, and now it locks up almost instantly. more volts eventually BSOD's 101/124 after 2-6 hours. can't even get 4.8 stable either on it. seems like junk, or something else I dunno. and my other 2500K was having problems doing even 4.7/1.376 which It used to prime 12+ just fine. I'm starting to wonder more and more if maybe something else in my machine is causing problems. or is it normal for chips to degrade like this? sofar i've seen it in all 3 chips i've owned. (first one stopped doing 2133MHz memory even at stock)
default volts for this new 301A127 is 1.165v/3.4GHz so it looked promising.
and BTW, i envy those of you with enough cooling for LinX+AVX lol
no problems @all with my 24/7 settings written in my sig. maybe u have some other troubles @ur sys. itīs strange that all ur 3 chips r degrading on that fast way. which vcore etc.. did u use for benching ?
btw. i have done the LinX stess test and joined the club ;) no degradation unitl now :D
I was priming 4.9@1.43v and 4.8@1.4v and it seemed fine until I left it overnight.Quote:
Originally Posted by Sauron_Vie
anyways, I'm done for now until I get better cooling. sold my 2nd 2500k and my corsair H50, so I'm on stock HS/F @ 4.5GHz/1.31v with my L041C123
It could be that you are killing the memory controllers in your chips. Your sig says you're running 1.66V, 2133mhz 1T. That's about as hard as you can push the IMC. The stock voltage for Sandybridge ram is 1.5V I read that it really only should be overvolted up to 5%, which is 1.58V I'd suggest that if you get another chip, you keep the ram voltage at 1.55V or less.