• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Performance/Voltage Index

R

robberbaron

Supreme [H]ardness
Joined
Sep 6, 2004
Messages
6,101
This is something wholly unimportant, but since I'm bored and procrastinating homework, I figure I would write it up. What I'm trying to find is the best Performance/Voltage ratio there is on A64's. This score can be called the PVI. I don't want to make cache an issue with this, as that will lead to arguments I'm sure, so why not have both 1024kb and 512kb cache chips have the same test score? I mean, sure, it's more of a challenge I'd think for a 1mb core to get lower voltages since they have more transistors, but whatever.

For instance,



This would be a PVI of 1704. 1800/1.056. At stock speed and voltage, it's 2400/1.5=1600.
I don't think I can do better than 1704 with this chip, though I might be able to do 1900MHz.

Anybody willing to see how high they can go(speed) while seeing how low they can go(voltage)?

edit:


Got 1766. I want to see this beaten, badly. Be sure to say what week you have and of course the core. Posting a cpu-z shot would be best.
 
nifty idea.
i've done testing on my cpu a while ago. sadly i don't have any cpu-z shots for proof, but here's a chart from the excel document i saved all the stuff to (baron, you've seen this, i'm just showing everyone else)

Code: 
volts   mhz
1.55	2310
1.5	2272
1.45	2224
1.4	2176
1.35	2110
1.3	2042
1.25	1975
1.2	1881
1.15	1765
1.1	1657
1.05	1556
1	1429

the highest PVI score is at 1.25v and 1975mhz, giving 1580
not bad.. but not as good as your 3700+

one thing though... those voltages are in the bios. at 1.25v, i actually get less. i'll reboot and come up with a screenshot.
i'm also assuming it has to be a stable speed. ;)
 
Yeah, stable as in, not half killing windows system files and whatnot, hehe. Hm, I need to make a table of my max speeds for each volt.
 
undervolts a tad over .03v

cpu-1975.png


so 1975.9 / 1.216 = 1624.9

that's better :D
 
Pretty nice. I don't mind my processor being a full 10 degrees cooler while folding :D


1826.
I doubt anyone other than us will be posting in this thread, haha. I mean, maybe iddqd would or someone else that's into this sort of meaningless A64 related stuff.
 
nice. i guess it helps that you have a cpu that's rated for an extra 400mhz at stock voltage :p

and.. i'm sure we'll get some people. we might wanna split the categories between 130nm and 90nm though, cause it's unfair to compare different processes ;)
 
(cf)Eclipse said:
nice. i guess it helps that you have a cpu that's rated for an extra 400mhz at stock voltage :p

and.. i'm sure we'll get some people. we might wanna split the categories between 130nm and 90nm though, cause it's unfair to compare different processes ;)
Click to expand...

Yeah, that's true.

As for the differences in processes... that venice at 2850 with 1.38v has a PVI of a whopping 2065.
 
(cf)Eclipse said:
hell, an fx-55 under phase change at 3600mhz and 1.8v is 2000.
Click to expand...

To match that, I'd have to do 2200 and 1.1vcore. That ain't gonna happen. What I love and hate about this "score" I invented is that there's not a damn thing you can do coolingwise or whatever to improve your score.
 
sure there is. i can probably do 2600mhz with phase change and 1.65v... scale that down into a better range, and i've got a good score. :D
 
(cf)Eclipse said:
sure there is. i can probably do 2600mhz with phase change and 1.65v... scale that down into a better range, and i've got a good score. :D
Click to expand...

Yeah I wasn't thinking of it that way. I was just thinking of keeping it close to 1.0v to get the best score. I can do 2400 at 1.35 which gives 1778. 2700/1.7v wont exactly help my score...


But you're right, phase change could help, but I doubt water or air will ever.
I think PVI should be a competitive value to work on. If it ever caught on, that would be dandy.
 
robberbaron said:
I think PVI should be a competitive value to work on. If it ever caught on, that would be dandy.
Click to expand...
sig time. it's a good way to spread the word..
omg! i need to post something important. :D
 
Whats the meaning of it?

I mean Uve created a score but what does it mean.

If trying to build a quiet system I can see some justification of it since U will have a good heat/performance ratio, likewise If U try to build for less powerconsumption (then one of the Passive water coolers would be nice, but don't even think about phase :D).

So if those are the only uses this score has PVI isn't a ggod name U need something that tells what it is.
 
Frallan said:
Whats the meaning of it?

I mean Uve created a score but what does it mean.

If trying to build a quiet system I can see some justification of it since U will have a good heat/performance ratio, likewise If U try to build for less powerconsumption (then one of the Passive water coolers would be nice, but don't even think about phase :D).

So if those are the only uses this score has PVI isn't a ggod name U need something that tells what it is.
Click to expand...

There are many uses.
It shows the quality of the transistors and the CPU.
It gives you something to do in the dark of the night.
It gives an idea of how well A64's scale with voltage.
Finally, it lets you have a score that matters as much as 3dmark.
 
robberbaron said:
It shows the quality of the transistors and the CPU.
Click to expand...
that's the kicker right there. higher PVI = better clock speeds with same voltage, which would imply a better core :cool:
 
(cf)Eclipse said:
that's the kicker right there. higher PVI = better clock speeds with same voltage, which would imply a better core :cool:
Click to expand...

Yep. What would you rather have? An FX-55 that does 3.4GHz with 1.8v or 3.4GHz with 1.6v? PVI comes into play in many cases, it just hasn't really been given a name or conciously paid much attention to.
 
robberbaron said:
There are many uses.
It shows the quality of the transistors and the CPU.
It gives you something to do in the dark of the night.
It gives an idea of how well A64's scale with voltage.
Finally, it lets you have a score that matters as much as 3dmark.
Click to expand...

Acctually it has some uses which puts it at least two notches ahead of 3D mark
Volatge scaling - well yaea but why?
In the dark of the night Ive got "other" things to do/play with (wishful thinking Ill just start a PIV-table up :()
Transistor quality - thats a good one especially if U could use it to bench the diffrence between Cores/manufactoring processes/Brands (say Intel/AMD)...
 
I would put this in the intel forum, but I dont really have a Pentium4 with a motherboard that allows for this sort of tweaking. Plus the scores would be kind of incompatible, since the clockspeeds are so different. Maybe we could establish a common PR between the two processors? I mean, they're good for different things, and a common PR would take longer to ratify than a constitution with how passionate people are about their brand names.
 
I have to leave shortly but I'll try later and see what I can come up with. Interested to see how little voltage it can take before CRASH. :D
 
my pvi:

2474 / 1.45 = 1706

And thats not even the max of my chip on that voltage, just the max of my motherboard. my CPU wins you robberbarron
 
then back down the voltage to get a higher PVI ;)
he's still 60pts ahead of you.. actually more.

also, i don't think 90nm and 130nm are comparable. 90nm should own 130nm :D
 
wonderboy_ said:
my pvi:

2474 / 1.45 = 1706

And thats not even the max of my chip on that voltage, just the max of my motherboard. my CPU wins you robberbarron
Click to expand...


Meh. 1798, stability tested and all. 92 points ahead of you, though you should be able to pull ahead, having 512kb of cache and 90nm and having a core rated at .1v lower than mine.
 
My Winchester is 1581.25 (2530/1.6). I suppose I need to work on it.
 
(cf)Eclipse said:
then back down the voltage to get a higher PVI ;)
he's still 60pts ahead of you.. actually more.

also, i don't think 90nm and 130nm are comparable. 90nm should own 130nm :D
Click to expand...

As far as i know, i cant turn the voltage down on a 9nda3+
 
tsuehpsyde said:
use clockgen to lower the voltage.
Click to expand...

That would only work if theres a clockgen for it and the bios doesnt freak out when you do that.
 
if it's an nf3-250, there's a clockgen for it :D
does it mean it works though? i dunno. i know voltages don't work with the dfi, which makes me really sad
 
(cf)Eclipse said:
if it's an nf3-250, there's a clockgen for it :D
does it mean it works though? i dunno. i know voltages don't work with the dfi, which makes me really sad
Click to expand...

Yeah, the DFI hates clockgen with a passion. Even the Soltek was easily tweaked with it.

Hah, forgot the NF3 250 was "universal."
 
Well anything that software tunes your system... DFI hates it with abandon, really one of the only ways of KILLING a DFInF4 MoBo is to run nTune on it. Ull never get it back LoL...
 
Frallan said:
Well anything that software tunes your system... DFI hates it with abandon, really one of the only ways of KILLING a DFInF4 MoBo is to run nTune on it. Ull never get it back LoL...
Click to expand...

I poked at nTune just to see what if anything it could do...it said it optomized me and showed me a bar vs. a "baseline" one and said it did something...but nothing had actually been changed...useless bundle of bits.
 
burningrave101 said:
There are actually burn-in steps you can take to try and lower the voltage requirement of your CPU and make it respond to voltage better which can possibly give you a better overclock as well.

http://www.ocforums.com/showthread.php?t=341296

Alot of people say it works. I havn't tried it myself yet.
Click to expand...


I just tried a little bit. It makes 2650MHz/1.7v stable in prime now for 3 passes (so far, it's still going) with Small FFT. Before it died in like 10 seconds. iiiiiiiiiinteresting..

hah, 5 minutes after I posted this, prime hit an error and the system rebooted immediately. Still...
 
I had 2.4ghz and 1.6v yielding only 1500 :(

I'm testing 2.03ghz and 1.2v, stable so far, yielding a pvi of only 1687 ( had 1.175 but not stable or 1723 pvi).

maybe not too bad for an old 3000+.

maybe i can get 2.1ghz out of her at this voltage.

I might just stick this 3000+ in a silent sff box if if's stable at stock and 1.2v
 
it seems that the best power/speed ratio is around 1800mhz to 2000mhz.. for the lower rated cpu's at least.


edit: no, around 1.2-1.3v
 
I've got 2500/1,33=1880!!!



My CPU is am 90nm Athlon 64 3500+, DIK4BI CBBID week 0503

Voltage: Default
HTT: 250x4
CPU: 10x
RAM: "180" MHz

AMD3500+@2500 MHz w/box cooler, Corsair 3200XL @ 227MHz
MSI K8N-SLI Platinum
 
I don't see how this has anything with the "quality" of the transistors. That's just clever sounding speculation. It probably has more to do with other variables such as PSU/mobo.
 
psu and mobo don't have much of an effect on your overclock unless you got crap stuff ;)
 
Either way, I doubt most people here know so much about the inner workings of the microprocessor's components or the manufacture process to make any claim as to this having anything to do with the "quality" of the transistors.
 
You must log in or register to reply here.
Back
Top