3700 San Diego vs. 3200 Venice: Round 2!

Printable View

Show 100 post(s) from this thread on one page

05-12-2005, 03:39 AM
s7e9h3n

Quote:

Originally Posted by ewitte

It could easily be 27s if I could pull 2-2-2-5-1T.

Eric

AFAIK, that's not possible @ 2800....but please feel free to prove me wrong ;)
05-12-2005, 03:41 AM
s7e9h3n

Quote:

Originally Posted by ewitte

It could easily be 27s or even 26s if I could pull 2-2-2-5-1T.

Eric

Ok, 26's are impossible @ 2800- I know that for sure :D
BTW- Here's what it took for me to get into 27's with the 3800+ (albeit untweaked, unpatched):
http://i137.exs.cx/img137/8564/10x310m3zm.jpg
26's:
http://img86.exs.cx/img86/4919/11x290m29153cu.jpg
05-12-2005, 04:12 AM
Zebo

Wow what kinda cooling? strange temp 25C..water chiller?
05-12-2005, 04:35 AM
s7e9h3n

Quote:

Originally Posted by Zebo

Wow what kinda cooling? strange temp 25C..water chiller?

Mine? Mach2 for these shots, stock hsf on first page. These shots were from the review I did on the Venice about 2 months ago here
05-12-2005, 06:02 AM
i found nemo

tha's some nice benchin' and comparison, glad you shot these two out make 'em think
05-12-2005, 06:10 AM
ewitte

Quote:

Originally Posted by s7e9h3n

AFAIK, that's not possible @ 2800....but please feel free to prove me wrong ;)

Its not going to be me because my TCCD does not like dropping below 2.5CAS. Someone with UTT or BH and excessive voltage needs to jump in ;) If it were possible I'd say 26-27s is within reason. BTW I do know someone that can get 1.5-2-2-5 @ 265x10. I really think there is some kind of BIOS/OS issue for those getting lower #'s with San Diego.

Eric
05-12-2005, 06:43 AM
ben805

At this point, I suppose it's safe to say that my brother's Venice gave my San Diego a run for the money, actually outran it in certain area. I can not explain why is this happening because the results goes against the theory and expectation that most of us had, technically...more cache should be better. I wouldn't put the blame on this SD either, just take a look at Dumo's FX-55 result from the other thread, he was running at the same clock speed with the same timing and our scores are very close, so...to say that I have got a defective or sucky San Diego, it's like saying his FX-55 is just as sucky as my SD! :D
I'm willing to accept the fact that my brother has got an excellent Venice, but I refused to believe my SD is out of whack or perform any worst than any other SD at the same clock speed out there, its VERY easy to verify this....280 x10 is not an impossible number to achieve, anybody who has a 3700 SD with any TCCD ram can run the exact same clock speed with the exact same memory timing, they can then use the same benchmarks and compare their results with mine. In the mean time, I would like to thank you all for the thoughts, comments, and suggestions that you gave me. ;)

andL64 and eva2000:

I do not know why the SuperPI 1M is slower, more than likely due to the memory timing I used, I wasn't aim to achieve the best SuperPI score so I didn't bother to alter the settings, If my intention is to get the highest SuperPI score, I would of pump more vmem and run much tighter timing on the ram, and I wouldn't of chose 280 x10 either. :)

Quote:

Originally Posted by Mastakilla

the strange part is:
ur PWMIC, toghetter with ur cpu temp, of the san diego is muuuuuch higher then the venice..?!
so i guess it has to do something with clockgen of ur san diego?

looking at the heat load and overall temperature, I believe this San Diego draw more juice from the PSU and put more stress on those mosfet and other components.

Quote:

Originally Posted by s7e9h3n

Great job Ben. Notice that people actually are commenting on the results of your tests rather than crapping on how you ran the tests? :D . Hey your work's not over. Get some rest tonight and get ready for the onslaught of questions tomorwow :toast:

Thanks for the suggestion and I'm glad I rerun the test again ;)
05-12-2005, 06:56 AM
ewitte

Quote:

Originally Posted by Mastakilla

the strange part is:
ur PWMIC, toghetter with ur cpu temp, of the san diego is muuuuuch higher then the venice..?!

I wish I knew what PWMIC was because its running hot on my PC. The CPU only gets into the lower 30's, lower 40's for chipset yet PWMIC can hit 55C.

Eric
05-12-2005, 06:59 AM
ewitte

Quote:

Originally Posted by ben805

I'm willing to accept the fact that my brother has got an excellent Venice, but I refused to believe my SD is out of whack or perform any worst than any other SD at the same clock speed out there, its VERY easy to verify this....280 x10 is not an impossible number to achieve, anybody who has a 3700 SD with any TCCD ram can run the exact same clock speed with the exact same memory timing, they can then use the same benchmarks and compare their results with mine. In the mean time, I would like to thank you all for the thoughts, comments, and suggestions that you gave me. ;)

I'd like to compare results. Can you post a screenshot of A64Tweaker? There are a lot more memory variables than you listed above. I am getting 2s lower than you in SuperPi but I run 255x11 not 280x10.

Eric
05-12-2005, 07:12 AM
ben805

Quote:

Originally Posted by ewitte

I'd like to compare results. Can you post a screenshot of A64Tweaker? There are a lot more memory variables than you listed above. I am getting 2s lower than you in SuperPi but I run 255x11 not 280x10.

Eric

The screen shots of my memory timing can be found on my signature, that's something I do not touch once I'd found the best combination for my TCCD....but that is about to change tonight when I put those Mushkin Redline in!
05-12-2005, 07:26 AM
pershoot

Quote:

Originally Posted by s7e9h3n

AFAIK, that's not possible @ 2800....but please feel free to prove me wrong ;)

my SD 3700+ @ 2.75, pulls 28-29s super pi (1.1e??)1m, 1.5-2-2-7 (with some other timings slghtly looser at this stage). so ya i agree, @ 2800 you probably wouldnt get 27s. but ya never know:)
05-12-2005, 07:33 AM
Alec

Have you altered your registry in windows so it uses 1024kb of L2 cache??
05-12-2005, 07:38 AM
pershoot

that should not be necessary, as windows will auto detect and use whatever onchip has.
05-12-2005, 08:07 AM
BlackArchon

I can't believe that there is no benefit from 1 MB L2 Cache. I assume three reasons for it:

- BIOS problems
- he used apps which don't benefit at all from a big L2 cache (some games like UT2004 do so very well)
- AMD changed the memory controller with revision E dramatically

Are there in the web some benchmarks from Venice and SD cores with equal settings?
05-12-2005, 08:35 AM
ben805

Quote:

Originally Posted by Alec

Have you altered your registry in windows so it uses 1024kb of L2 cache??

nope, where in registry can you change that? and have anybody seen any improvement on these FX or SD by doing registry tweak like that?

Quote:

Originally Posted by BlackArchon

I can't believe that there is no benefit from 1 MB L2 Cache.

Me neither, but I wouldn't say absolutely no benefit....but only in certain area would it outshine the 512KB, if you look at the cpumark99 score, it's about 10% gain!

Quote:

Originally Posted by BlackArchon

I assume three reasons for it:

- BIOS problems
- he used apps which don't benefit at all from a big L2 cache (some games like UT2004 do so very well)
- AMD changed the memory controller with revision E dramatically

those could be the possibilities.....

Quote:

Originally Posted by BlackArchon

Are there in the web some benchmarks from Venice and SD cores with equal settings?

I'm starting to wonder exactly how did other previewer(s) conducted their test!! Or was it only the "chery picked" samples can yield that kind of results??!?
05-12-2005, 09:01 AM
ewitte

Quote:

Originally Posted by pershoot

that should not be necessary, as windows will auto detect and use whatever onchip has.

Its been known to autodetect incorrectly before.
05-12-2005, 09:36 AM
Bennah

We are back to this ;)

Should have some little results of my own this weekend ;)
05-12-2005, 09:47 AM
andL64

ok being so "ugly" wasnt ok, so i apologize!

u ve done some work for us and that should be appreciated!

i said the (p)review is weird is because the cpumark score got better with SD but in 3d01 theres no performance gain?! 3D01 was known to respond best for larger L2 cache so dunno?!

Otherwise ure right with the venice/SD Spi thats ok cause scores are near beneath only SD is little bit better. So you re right with changed timings it will get really better.

but its still strange.

seEn
05-12-2005, 09:53 AM
s7e9h3n

Sorry for the semi-hijacking Ben, but I promise I'm not gonna post about this much more: :D

Quote:

Originally Posted by ewitte

Its not going to be me because my TCCD does not like dropping below 2.5CAS. Someone with UTT or BH and excessive voltage needs to jump in ;) If it were possible I'd say 26-27s is within reason. BTW I do know someone that can get 1.5-2-2-5 @ 265x10. I really think there is some kind of BIOS/OS issue for those getting lower #'s with San Diego.

Eric

No need for someone else ;) :
http://img115.exs.cx/img115/9205/2754wx.jpg

BTW, I've run SPI with bh-5 @ ~270, but got nowhere near 26's @ 270x10.
05-12-2005, 09:54 AM
s7e9h3n

Quote:

Originally Posted by Bennah

We are back to this ;)

Should have some little results of my own this weekend ;)

It's about time :rolleyes: ;)
05-12-2005, 10:06 AM
Bennah

Quote:

Originally Posted by s7e9h3n

It's about time :rolleyes: ;)

I still havnt recieved the CPU :p:

Fedex arrived yesterday with it but I was at work. Collecting it tomorrow :D
05-12-2005, 10:12 AM
s7e9h3n

Quote:

Originally Posted by Bennah

I still havnt recieved the CPU :p:

Fedex arrived yesterday with it but I was at work. Collecting it tomorrow :D

Well that explains everything..... :D
05-12-2005, 12:49 PM
ImportantAwareness89

What I don't get is why you have to overclock.

If a 3200+ runs at 10 x 200, why don't you jsut run the 3700+ at 10 x 200...

Well, seems to be a better setup, but I stand by:

http://www.compressorcooling.com/htm...go_vs__ve.html

That is the 3700+ vs the 3800+. So its a 2.2 vs a 2.4. You would think the 2.4 (3800+) should win with the extra clock, but the 1MB cache of the 2.2 (3700+) is enough to make it win in most benches.

If 2.2 > 2.4, then SURELY SD >>> Venice clock for clock. Maybe your SD is jacked up? =(
05-12-2005, 01:23 PM
ben805

Quote:

Originally Posted by dmo580

What I don't get is why you have to overclock.

If a 3200+ runs at 10 x 200, why don't you jsut run the 3700+ at 10 x 200...

Well, seems to be a better setup, but I stand by:

http://www.compressorcooling.com/htm...go_vs__ve.html

That is the 3700+ vs the 3800+. So its a 2.2 vs a 2.4. You would think the 2.4 (3800+) should win with the extra clock, but the 1MB cache of the 2.2 (3700+) is enough to make it win in most benches.

If 2.2 > 2.4, then SURELY SD >>> Venice clock for clock. Maybe your SD is jacked up? =(

Why do I have to overclock? :lol: eerrr.....am I in the wrong forum or what?? :rolleyes:
05-12-2005, 02:05 PM
viccyran

Quote:

Originally Posted by ben805

Why do I have to overclock? :lol: eerrr.....am I in the wrong forum or what?? :rolleyes:

He was referring to why do you have to overclock to compare clock-for-clock between SD and Venice. If you were to run both at a given stock speed, heat (if it was a problem) would have not have been an issue, and maybe the results might show something different :]

And yes, I was :confused: when I read the first part of his post too :)

Show 100 post(s) from this thread on one page