3000+ venice

senorstealth · 4 Jun 2006 at 22:24

Whats the maximum overclock people have managed to get from this processor.

I don't know what the stepping is, as cpu-z won't read it and i don't want to remove the HS (again

)

Minstadave · 4 Jun 2006 at 22:28

senorstealth said:
Whats the maximum overclock people have managed to get from this processor.

I don't know what the stepping is, as cpu-z won't read it and i don't want to remove the HS (again )

On air you're looking at 2.6-2.8 really, maybe higher on decent cooling.

Global_Inferno · 4 Jun 2006 at 22:29

Minstadave said:
On air you're looking at 2.6-2.8 really, maybe higher on decent cooling.

Really? with only a 9x multiplier?? He'll need some AMAZING RAM to pull THAT outta the bag! Thats more than my 4000+ does! :rolleyes:

senorstealth · 4 Jun 2006 at 22:33

Was thinking that 2.2 - 2.3 was a realistic figure

Minstadave · 4 Jun 2006 at 22:35

Global_Inferno said:
Really? with only a 9x multiplier?? He'll need some AMAZING RAM to pull THAT outta the bag! Thats more than my 4000+ does!

Not if you use a memory divider......and I stand by 2.6-2.8 thats normal for a Venice, especially anyone bought recently.

2.6 on stock air with volts around 1.5-1.55V.

2.8 if you're lucky with the chip and have a decent aftermarket aircooler and give 1.55-1.6V.

(Edited for knee jerk reaction to rolleyes and being talked to like I don't know what I'm talking about.)

Global_Inferno · 4 Jun 2006 at 22:40

Yeh but dividers lower performance

So I think i'd rather have it clocked at 2.2GHz without a divider than 2.4-2.6 (ok.. maybe 2.6) with a divider.

Minstadave · 4 Jun 2006 at 22:42

Global_Inferno said:
Yeh but dividers lower performance So I think i'd rather have it clocked at 2.2GHz without a divider than 2.4-2.6 (ok.. maybe 2.6) with a divider.

Completely wrong I'm afraid. A64 does not suffer in any significant way with dividers, its not memory bottlenecked at all. Performance on A64 is simply about clock speed, 2.7Ghz (300x9) with the memory running at 200mhz on a 133mhz divider will outperform any 1:1 clock that you'd be able to get otherwise.

Try benching it yourself, only takes a few minutes, A64 is all about raw clockspeed, memory clocks gain you very little.

Global_Inferno · 4 Jun 2006 at 22:43

Minstadave said:
Completely wrong I'm afraid. A64 does not suffer in any significant way with dividers, its not memory bottlenecked at all. Performance on A64 is simply about clock speed, 2.7Ghz (300x9) with the memory running at 200mhz on a 133mhz divider will outperform any 1:1 clock that you'd be able to get otherwise.

Prove it.

Because from my experience, dividers SUCK ass.

EDIT: On closer searching of google.. seems, your right! I think I must have been having a blonde moment and was thinking about ye' olde Barton cores *blink* :confused:

In which case, lets see what it can do

Minstadave · 4 Jun 2006 at 22:45

Global_Inferno said:
Prove it. Because from my experience, dividers SUCK ass.

You don't have enough experience then, I'll find some figures.

Global_Inferno · 4 Jun 2006 at 22:48

Dont worry I've realised the error of my ways! However dividers DO LOWER PERFORMANCE, it is just by a negligable amount..

senorstealth · 4 Jun 2006 at 23:03

Thanks for the advice guys, least u know wht its capable of now

Minstadave · 4 Jun 2006 at 23:05

Ok, glad we've got that sorted out, after I did all the work

Here's what dividers can do on the 9x multi.

All have memory at 235Mhz:

1:1 Pi 1mb = 40.9secs

235x9= 2115Mhz

166 Divider Pi 1Mb = 33 secs

288x9= 2595Mhz

150 Divider Pi 1Mb = 31.2secs

312x9= 2808Mhz

Minstadave · 4 Jun 2006 at 23:10

32% increase in clock speed = 25% reduction in Pi times

This is all at the same memory speed, so using a memory divider makes a negligable decrease in the scaling of the system, even at 1:1 at 311x9=2808Mhz you'd be seeing 30sec pi time.

Here's a shot of 2700Mhz at 300x10 1:1 I did a fair while back with 2x512mb of Gskill LE TCCD, thats a massive increase in the memory bandwith and it still only pulls 31 secs in pi (though the CPU is a Winchester, so is slightly slower than the Opteron 144 I used for todays pics):

Defcon5 · 4 Jun 2006 at 23:40

Global_Inferno said:
Dont worry I've realised the error of my ways! However dividers DO LOWER PERFORMANCE, it is just by a negligable amount..

Very true. the performance is only deemed insignificant due to the boost given by being able to use lower latencies

Minstadave · 4 Jun 2006 at 23:53

Defcon5 said:
Very true. the performance is only deemed insignificant due to the boost given by being able to use lower latencies

No its not the lower latencies that are important, its the higher CPU clock you can achieve without being held back by memory restraints. I wouldn't use a divider in order to get lower latencies, thats not going to get you significant performance gains (unless you changing from very slack to very tight).

PapaLaz.uk · 5 Jun 2006 at 09:15

As above, if RAM constraints aren't an issue, then obviously no point using a divider. However, if you're RAM limited, then it makes sense to chuck on a divider and etch more out of the CPU. Also running a divider can help keep the timings v.tight, which the AMD chipsets seem to love.

The current 3000+ Venice chips are clocking nicely - bought one the other day and it's currently running @2.6 on stock volts - memory is on a divider, though still slightly clocked and running with tight timings.

senorstealth · 5 Jun 2006 at 11:35

PapaLaz.uk said:
The current 3000+ Venice chips are clocking nicely - bought one the other day and it's currently running @2.6 on stock volts - memory is on a divider, though still slightly clocked and running with tight timings.

Now thats what i wanted to hear, think its time to buy some new ram

Minstadave · 5 Jun 2006 at 11:38

senorstealth said:
Now thats what i wanted to hear, think its time to buy some new ram

Whats your current RAM?

senorstealth · 5 Jun 2006 at 11:56

You really don't want to know, tried underclocking it the other day so i could push the cpu and it really didn't like that very much.

It Crucial pc3200 3.4.4.8 @ T1, though its not a matched pair (ones a double sided stick)

Looking at something along the lines of G.Skill 2GB DDR HZ PC4000

Minstadave · 5 Jun 2006 at 11:58

senorstealth said:
You really don't want to know, tried underclocking it the other day so i could push the cpu and it really didn't like that very much.

It Crucial pc3200 3.4.4.8 @ T1, though its not a matched pair (ones a double sided stick)

Looking at something along the lines of G.Skill 2GB DDR HZ PC4000

The HZ or ZX are both very nice from Gskill

Good to see you going for 2 gigs too, a wise move