San Diego and Venice comparison at same clock speed

[ben805]

This is just a comparison between the 3700 San Diego and 3200 Venice at the same clock speed at 2862 Mhz, identical setup and configuration. Both cpu were being clocked at the same FSB and Multiplier

But first off, I wanted to show you how high can we overclocked them to passed the SuperPI 32M. After the SuperPI 32M shootout, the rest of the benchmarks are running at 318 X9 on BOTH CPU.


3200 Venice: The highest OC'ed speed I can run SuperPI 32M:


3700 SD: The highest OC'ed speed I can run SuperPI 32M:



OK, now the rest of the benchmarks were all running at 318 x9 on both cpu, on the same setup and same everything.



3200 Venice PCMark04:


3700 SD PCMark04:


3200 Venice Sandra:


3700 SD Sandra:

[ben805]

3200 Venice 3DMark2001 SE:


3700 SD 3DMark2001 SE:






The Venice idle around 26~27'C, underload 42~43'C.

The SD idle around 35~36'C, underload 59~60'C.

3200 Venice: USD$199.00 at Monarch

3700 SD: USD$344.00 at Monarch

At the same clock speed, the San Diego's 1MB L2 Cache 'suppose' to have an advantage over the Venice, but I'll let the numbers speak for themselves and let you be the judge. Personally, if I were to run air/water cool, I would stick with Venice. But when phase change is in the picture, then THIS(notice I said 'this', NOT 'all') San Diego should have a slightly advantage over the Venice since it requires less vcore to run at the same clock speed, but then it has much higher heat load so the verdict is still out. I'm awaiting the 'Bunker Buster' single stage phase change from PCICE, as soon as I get it I will push the envelop and see how far would it go....and hopefully I didn't wasted the extra $145....

[J-Mag]

The venice and SD are at different resolutions. They are different computers... You aren't using the EXACT same components are you? (besides cpu )

[ben805]

same setup, same configuration, same bios, same memory timing, same hardwares (including motherboard, memory, video card, psu, and all that jazz). The Venice belongs to my brother...and I have the San Diego.

[tristancarton]

1mb a64 chip getting ousted by a 512kb chip in 3dmark2001 at the same clock speed? i don't think that is possible. something must bet different. 3dmark2001 eats that 1mb cache up.

edit: didn't see the fsb difference. you should really run them at the exact same fsb and memory timings if you make a title like this. still seems strange though. usually the extra 512kb of cache adds quite a bit onto 3dmark2001.

[trakslacker]

well assuming that the exact same components are being used, why is it taht the Diego is losing out almost every benchmark?

[saiamne]

318mhz FSB on the Venice

285mhz FSB on the SD

[trakslacker]

same setup, same configuration, same bios, same memory timing, same hardwares (including motherboard, memory, video card, psu, and all that jazz). The Venice belongs to my brother...and I have the San Diego.

dayum...well then my previous question stands. How in the crap is the Sandy losing every last bench?

[ben805]

1mb a64 chip getting ousted by a 512kb chip in 3dmark2001 at the same clock speed? i don't think that is possible. something must bet different. 3dmark2001 eats that 1mb cache up.
edit: didn't see the fsb difference. you should really run them at the exact same fsb and memory timings if you make a title like this.

I've been asking the same question myself, same as my brother....I'm hoping someone can explain why is this happening and what the hell is going on...

well assuming that the exact same components are being used, why is it taht the Diego is losing out almost every benchmark?

I hope someone can answer us as well.

[tristancarton]

ben any chance you could do the same performance speeds at exactly the same fsb and memory timings instead of using different fsbs on each chip? and repost the results.

[ben805]

318mhz FSB on the Venice

285mhz FSB on the SD

WRONG, look at the post again, the HIGHEST achievable SuperPI 32M with the Venice is 318 x9, my SD crapped out and got BSOD when I tried to keep up with my brother's Venice, underload temperature was 60'C and there's no way I can get it to complete 32M, already tried that many times, running at the same speed as my brother's Venice at 318 x9...I can run the rest of the benchmarks all day long, superPI 32M just not doable!! The HIGHEST achievable SuperPI 32M with my SD is at 285 x10. So the rest of the benchmark OTHER THAN SuperPI 32M are all performed at the same setting, same clock speed, same timing, same everything.....318 x9. Someone ought to tell me if my SD is defective because "technically" and "theoritically" my San Diego suppose to smoke my brother's Venice!!!!!!! well...not smoke, but perform slightly better!!

ben any chance you could do the same performance speeds at exactly the same fsb and memory timings instead of using different fsbs on each chip? and repost the results.

that's EXACTLY what I did, how else can I do a fair comparison if I didn't match everything to the teeth? FSB, Multiplier, HT Ratio, EVERYTHING.....100% identical. I'm pulling my hair out trying to figure out WHY and HOW the hell did my San Diego perform the way it is, defective or what? bad memory controller? bad L2 cache???

[Wiggy McShades]

put a divider on the ram to get them both at the same fsb it seems like extra fsb is alot more beinificial than the 1mb cache.


edit: but dam thats alot better of a score for teh venice wiht only a 30mhz fsb advantage. maybe amd used higher latecy cache on the sandiago

[ben805]

put a divider on the ram to get them both at the same fsb it seems like extra fsb is alot more beinificial than the 1mb cache.


edit: but dam thats alot better of a score for teh venice wiht only a 30mhz fsb advantage. maybe amd used higher latecy cache on the sandiago

no divider needed, my GSKILL can take up to about 340FSB @2.5-4-3-7.
There is no FSB advantage!! look at the rest of the benchmarks, look at those picture's description. Again, both cpu are running at the same FSB, same multiplier, same everything!!

[tristancarton]

that's EXACTLY what I did, how else can I do a fair comparison if I didn't match everything to the teeth? FSB, Multiplier, HT Ratio, EVERYTHING.....100% identical. I'm pulling my hair out trying to figure out WHY and HOW the hell did my San Diego perform the way it is, defective or what? bad memory controller? bad L2 cache???

ben, i assume you wanted to see the difference between the two chips at exact same settings to see how much the extra cache matters.(based on title) don't worry about getting the highest speeds. just lower the san diego to a 9x multiplier and bench between 2750-2800 for each chip.(highest the san diego can go on the 9x mult) that will give you a much better look than running different fsbs and multipliers.

edit:nevermind read your response. i have no idea what the problem is.

[Wiggy McShades]

DOH i get it now sry u poseted ur highest oc for both first dam im stupid really sorry bout that.

[ben805]

Don't be sorry, I hope I didn't confused you guys

[Faceless Rebel]

Now that we have cleared that up, does anyone have any theories yet? I want to be reassured I didn't just waste $200 on a 1MB L2 cache that is somehow slower than 512k L2 cache, even though that sounds completely impossible...

[ben805]

I don't know what the answer is as I'm with all of you....still trying to figure out what the heck is going on, the only possibilities are: 1)Either my 3700 SD or my brother's Venice is defective. and 2)Heat killed!!

Assuming neither one of the processor is defective, the only theory I can come up with...is my 3700 SD underload temperature was at around 60'C while my brother's 3200 Venice ran much cooler and underload only 43'C, I'm not sure if the 17'C difference would make an impact in performance even at the same clock speed, but since both of them are running on the same XP-90 heatsink and the same 2800rpm fan, then I guess the Venice perform a tiny little bit better than the San Diego?!?

Now, IF...and IF...I can find a way to keep BOTH processors at the exact same underload temperature consistently, then maybe the benchmark numbers will be slightly different...again, it's just theory and I don't have any equipment or tools to test this theory out.

[Wiggy McShades]

is there any way amd pulled an intel god forbid and used higher latecy cacche on the sandiago? and im not really an amd guy but i like to learn so is there any type of throtilling on amd cpu's that would kick in at 60c? and maybe look for a backround program running on the sandiago rig although u probally already did worth another look maybe?

[trakslacker]

as far as I know AMD does not have any sort of thermal throttling logic built into the A64s. Basically if you overheat it, too bad. Although I really doubt that hitting 60C completely killed his SD. Also, why the crap is the SD temp so darn much higher?

Ben, how long would you estimate the SD was running at those load temps and did the HSF have good contact with the proc?

[i found nemo]

I've been asking the same question myself, same as my brother....I'm hoping someone can explain why is this happening and what the hell is going on...




I hope someone can answer us as well.

more efficient arcitexture????

[ben805]

as far as I know AMD does not have any sort of thermal throttling logic built into the A64s. Basically if you overheat it, too bad. Although I really doubt that hitting 60C completely killed his SD. Also, why the crap is the SD temp so darn much higher?

Ben, how long would you estimate the SD was running at those load temps and did the HSF have good contact with the proc?

couple minutes on each benchmarks, sometime it would dropped down to 56~57'C while the cpu load wasn't 100% but most of the time it was way up there 59~60'C. I remounted the heatsink and reapply the AS5 twice, you know that heatsink bracket on the motherboard where the heatsink sits on? it was sand down couple mm for my previous IHS-less winnie, so the heatsink sits pretty tight on the Venice and SD, contact is not an issue...

the main reason why I bought the San Diego was all because of the extra L2 Cache, this is going to suck if I were to stick with only air cool...maybe the cache or memory controller doesn't like those toasty heat, I don't know....I didn't expected this, and the numbers should of been the other way around Oh well, I guess I'll have to wait for the phase change and see how it would perform then....I thought about removing only the San Diego's IHS to compensate the extra heat, but then I don't think it would be a fair comparison unless both IHS were taken off. The only advantage that my San Diego has over the Venice, is the memory score on PCMark04, Cache/memory on Sandra, and the Cpumark99, his Venice only did 332 and mine was 368, here's another screen shot of his Venice running SuperPI 32M(unpatched) then cpumark99:

[Ugly_Jim]

He clearly says that they're both running at 318x9.

OK, now the rest of the benchmarks were all running at 318 x9 on both cpu, on the same setup and same everything.

[agenda2005]

Can you run both of them at 10X multi and the same frequency, HTT and memory timming? Because lowering multiplier has been know to cause loss of efficiency, especially in Sandra.
The only way we can clear this up is when we have both of them at thier highest multiplier posible, hence 3500+ is required for comparison with 3700+.
Secondly, can you change your fan RPM so that your 3200+ run with low rpm and hence higher temps, with 3700+ running with high fan speed/rpm and hence lower temps. Try and balance the temps as much as you can to help clarify the temperature/heat issues.

[ImportantAwareness89]

This is a bad benchmark. Raise your hand those of you who have seen the 3700+ vs the 3800+ where the 3700+ whoops @$$ in virtually all the benchmarks? Ok, enough said. Do the test on the same system plz. Kthx.