A64 3700+ SD... how good these overclock?

[Sorrento]

Ok, I think I have another very dumb question, sorry in advance:

 

The CPU revision is written on the heatspreader surface... right?

 

I keep forgeting to scan the surface and get an image of every CPU, AMD or Intel, I use... arghh, and when I need it I need to remove heatsink and clean CPU to get it. If its somewhere else, I really would like to know

[Cylok]

I'm getting prime stable clock of 550 MHz over stock with mine. I've been running @ 2750 24/7 with everything I would normally run, and haven't one glitch related to the O/C. On the other hand though, only single problem I have had, has been out of my grfx card, which I am trying to narrow down now. Driver related I think. So, IMHO, I believe the 3700+ (SD) O/C's very well.

[Angry_Games]

Ok, I think I have another very dumb question, sorry in advance:

 

The CPU revision is written on the heatspreader surface... right?

 

I keep forgeting to scan the surface and get an image of every CPU, AMD or Intel, I use... arghh, and when I need it I need to remove heatsink and clean CPU to get it. If its somewhere else, I really would like to know

it is laser-etched right onto the heatspreader

 

it will look something like

 

CAA2C 0504

 

lemme see if i got a pic around here somewhere...

 

 

ah here is one

 

ADA2800AEP4AX = processor specifications...can be found here:

http://www.angrygames.com/AMD_CPU/AMD_64/AMD_64.htm

 

CBAPC 0404VPMW

 

CBAPC = stepping code (this is the code that clockers look for)

0404 = week/year code (4th week of 2004) (they also look at week/year codes too)

 

 

by the way, this is my 2800+ Newcastle 754 and my 3400+ Newcastle 754 together just so you dont get confused lol

[Punster]

I'm getting prime stable clock of 550 MHz over stock with mine. I've been running @ 2750 24/7 with everything I would normally run, and haven't one glitch related to the O/C. On the other hand though, only single problem I have had, has been out of my grfx card, which I am trying to narrow down now. Driver related I think. So, IMHO, I believe the 3700+ (SD) O/C's very well.

 

What voltage is that? Mine buggeres up over 2620mhz - i think its ram related.

[Sorrento]

Thanks for answering the dumb question AG

 

I found a review of the 3700+ using a DFI Lanparty nf4 SLI DR, and they arrived to the same conclusions you did AG: 3.0 ghz was imposible to achieve with BIOS 510-2... but could be posible to reach 3.2 ghz with BIOS 310p but still like the 510-2 with memory divider problems.

 

I am not sure about this conclusions and issues they came up with, what do you guys think:

 

http://www.bleedinedge.com/reviews/process...n-Diego_01.html

[Cylok]

What voltage is that? Mine buggeres up over 2620mhz - i think its ram related.

 

 

CPU = 1.5v @ 104% = 1.54v overall

[THunDA]

My 4000 is prime stable @ 2900mhz on air..

 

*note this is running 4x512mb's of tcc5 thats why its running 2T*

 

Im curently priming away with the bh5 setup in my sig.. I think that between the 3700 and 4000 you are gone get somewhere between 2800-2900mhz.. I can get into windows at 3000mhz but to make it stable it needs more voltage that im not wanting to give it..

[Quad Master]

I just wanted to ask.

What sort of a perf. difference will be there with

 

3000+ Venice - Clocked to 2800 Ghz [Air Cooled]

3700+ SanDeigo - Clocked to 2800 Ghz [Air Cooled]

 

Both Prime Stable

 

As you know 15th & 17th Venice are easily capable of reaching 2.9 on Air.

The Venice Will OC by a greater percent than SanDeigo [% Compared to there stock speed]

 

What sort of a difference will be there with Venice having 512Kb L2 & SD having 1MB L2.

 

Which one wud be better , i also know that it wud depend on ram speed but still.

If Venice 3000+ reached 2.8 & SD also reaches 2.8 then why waste more money.

[Sorrento]

Yes, both will be able to run at same speeds... but there are two things to consider: Multiplier and/or cache size.

 

San Diego has 1 mb L2 cache size, Venice has 512 kb L2 cache. If compared to the 3200+ Venice, the 3700+ has an 11x, and if its used will have higher FPU power than with lower multi and higher HTT. I am not sure why where what here, but the results are always better using higher multiplier for raw CPU power, while higher HTT gives you better 3d power.

 

The image I posted of an FX55 CH doing 3.2 ghz was with cpu at 16 x 200. In SuperPi that resulted in a little less than 26 seconds (25.92). If the same system runs at 10 x 320 it scores a little more than 26 seconds, almost 0.6 seconds slower. 200 mhz test was with BH5's, while 320 mhz test was with TCCD's at 1:1 1T.

 

3700 has same multi than the 3500+ Venice... only diference will be the cache size, nearly 200 mhz cpu speed advantage.

[Quad Master]

Thanks for the info.

 

But wont increasing the FSB from 200 to a higher value result in ram running at higher

speed and yeild better perf than running ram at 200Mhz.

 

BH-5 are capable of doing 250 @ 2-2-2-5-1T with proper VDimm.

Wont these BH-5 do better with 250 than at 200.

 

also i didnt knew than instead of increasing the FSB by increasing the multiplier

we yeild better perf. on FXSeries of AMD.

[LazyBum]

My 4000+ San Diego is running at 240x12, same as what AG got with his. I had trouble getting that last 5 out of the FSB, 235 was my limit until I changed to the 510-2FIX bios. I should do a toaster shot with it for the OC database, I suppose. Maybe sometime this week, I'm just running on a half-assed install on a partition until my Raptor shows up from RMA and I can redo my RAID array.

[Sorrento]

Thanks for the info.

 

But wont increasing the FSB from 200 to a higher value result in ram running at higher

speed and yeild better perf than running ram at 200Mhz.

 

BH-5 are capable of doing 250 @ 2-2-2-5-1T with proper VDimm.

Wont these BH-5 do better with 250 than at 200.

 

also i didnt knew than instead of increasing the FSB by increasing the multiplier

we yeild better perf. on FXSeries of AMD.

The guy with the FX55 I posted the information only had a MSI K8N Neo2 back then, faulty as it didn't let him rise the HTT (its HTT not FSB btw) frequency to 240 only. With his DFI NF4 Ultra D he is able to run at 320 mhz now.

 

However, the same result happens at tests like SuperPi: 200 x 16 results in 25.9 seconds, while at 320 x 10 the result is 26.4 or something, half a second slower.

 

The BH5 does better indeed at 250, same does his TCCD at 320, but for tests that use raw cpu power lower multis are better. I think 3D tests benefit from the extra HTT, even if its not at sync with the ram.

 

Not sure if other cpu's behave this way, but Athlon64's do... thats why higher multipliers cost a fortune. The diference is cache size is around 200 mhz, maybe more since at 2.7 ghz a 3200+ Venice isn't as fast as the stock FX55 in all tests, thats only at 2.8 ghz. But the 3200+ is 10x, maybe a 3500+ Venice with 11x would give a diferent result... who knows. The 3700+ SD has 11x too, but with 1 mb cache, at 2.6 ghz I think it would be exactly as fast as the FX55, except in SuperPi and other CPU tests as one has 11x multi and the other 13x