s939 3500 oc experinces?

[bias_hjorth]

Hi guys

I´am up for buying this wonder (at least I think I am) but I would like to hear if some tried one of these yet.. ?
The mobo I choose to it should be a gigabyte right..

thx in advance

[Pejsen]

I'm really interessted in this too, as i am also going s939 soon

[kakaroto]

this is a pic from mangoose420.

The 3500+ is truely a good overclocker But at this moment still to expensive. A good 3400+ NC will kill this baby

[=w=]

Or mobile OC'ed .

[Master_G]

What cooling is that screenie with?

G

[bias_hjorth]

Originally posted by kakaroto


this is a pic from mangoose420.

The 3500+ is truely a good overclocker But at this moment still to expensive. A good 3400+ NC will kill this baby

come to think of it you´re right.. I assume this was with air or water..?

[Mariachi]

Must be Air / Water, or even stock cooler

[Mongoose420]

thats with the stock hsf. and according to hexus's pifast challenge its good enough for 37th in the world, 2nd best aircooled(slk948 and tornado on a fx51 looks to be the best aircooled rig at 45.30) and best retail box cooled on the list. Granted not everyone uploads there results.

Ive since got 2x512 ocz3700EB that does 250 3-2-2-10 1T and i'll have my JPI v5 fitted with JPI xlr120 rads and a Dtek WW on the cpu as soon as i get everything in and i'll be picking up a DDr booster as soon as they become available. 260~275fsb 1:1 3-2-2-10 1T and 2.7~2.8ghz on the 3500+ will pull down some nice numbers.

And ive yet to see a skt745 newcastle use'n comparable cooling turn in faster numbers in pifast or super pi than my 3500+.

[bias_hjorth]

very nice with air!! those new s939 seems to really clock well.. wonder what your cpu would do under a phase-c..

[Mongoose420]

not a practical chip for phase change and would more than likely need a 300HTT capable board. being stuck with the 11 multiplier and if i had a board that did 300htt i'd still only get 250mhz out of my memory. But if you were to spend the 500~1000 dollars on a phase change set up you would buy a fx53 and not a 3500+. Not just for the multipliers but the extra cache.

[StormPC]

Definitely! The Newcastles suck on 3DMark because they have been robbed of cache. They do PiFast and SuperPi well though.

[kakaroto]

Originally posted by StormPC
Definitely! The Newcastles suck on 3DMark because they have been robbed of cache. They do PiFast and SuperPi well though.

they not suck. The NC do pretty wel under 3dmark01. Ok...not as fast as a clawhammer 1MB but wth, under overall games and graphics, they also do fine.

btw: NICE 3dmark score for stock stock setup

Mangoose420: Hopefully i will get my 3400+ NC (200x12) from USA. Then i will test it under stock hsf.
I think i can only beat youre score only by 1:1 around 250-260MHz 2-2-2-5 memory speed.

Dual channel give you alot of boost

[lostark374]

ill let you know in about a week

[StormPC]

Originally posted by kakaroto
they not suck. The NC do pretty wel under 3dmark01. Ok...not as fast as a clawhammer 1MB but wth, under overall games and graphics, they also do fine.

btw: NICE 3dmark score for stock stock setup

Mangoose420: Hopefully i will get my 3400+ NC (200x12) from USA. Then i will test it under stock hsf.
I think i can only beat youre score only by 1:1 around 250-260MHz 2-2-2-5 memory speed.

Dual channel give you alot of boost

Thanks.

Well sucking is a relative term. The 939 Newcastle gets it's a$$ handed to it in games and 3DMark by the 754 ClawHammers because the DC does not make up for having only half the cache. If all you do is calculate pi and look at Sandra mem bandwidth you'll be happy with the 3500+. If not you won't be.

The only fast 939 is the FX-53.

[Mongoose420]

this is new, the 3500+ sucks in games now? I'd hightly doubt the diffrence in actual gaming is noticable. And to put a finer point on it the s939 newcastles suck in car low and drag low just like the skt754 newcastles do when being compared to clawhammers, only diffrence is dual channel on s939 makes up for the cache in memory dependant applications. The diffrence clock for clock is about 5~6% in 3dmark01

[misteroadster]

Originally posted by StormPC
Thanks.

Well sucking is a relative term. The 939 Newcastle gets it's a$$ handed to it in games and 3DMark by the 754 ClawHammers because the DC does not make up for having only half the cache. If all you do is calculate pi and look at Sandra mem bandwidth you'll be happy with the 3500+. If not you won't be.

The only fast 939 is the FX-53.

i think you're wrong.
i tried 3800+ stock with 9700pro stock and have 19550 , it's 550 more than FX53 940 with same set up.

[StormPC]

Now you're getting silly. The FX-53 S940 (you know, the fastest computer on the FM 2k1 ORB?) will slap a 3800+ so bad it's not even a contest. The 3700+ will smoke the 3800+ also. Post your best 2k1 score and I'll run my 3700+ on air at the same clocks and memory timings and we'll see though, no need to take my word for it.

Oh, and testing high powered systems by limiting them with 3 or 4 generation old videocards is not very scientific because you are penalizing the faster processors.

[bias_hjorth]

Originally posted by StormPC
Now you're getting silly. The FX-53 S940 (you know, the fastest computer on the FM 2k1 ORB?) will slap a 3800+ so bad it's not even a contest. The 3700+ will smoke the 3800+ also. Post your best 2k1 score and I'll run my 3700+ on air at the same clocks and memory timings and we'll see though, no need to take my word for it.

Oh, and testing high powered systems by limiting them with 3 or 4 generation old videocards is not very scientific because you are penalizing the faster processors.

plz. define "smoke" cause I would love to see that.. I honestly believe their are on pair.. but couldnt you guys both do a stock run (stock everything) and with similar timings.. That would settle it once and for all

[iboomalot]

just playing with mine at this point

aquamark CPU at 10,200
Pi was 37secs at 1M
haven't run 2001se yet
PCmark2002 = 8000 low 10,000 high on memory and 7900 on CPU

had a problem with my RAID setup and going to a single WD740 and see if that helps.

would give you more info but re-installing XP so those #s are off brain power or what little I have

[StormPC]

Smoke=beat (and not just by a little bit).

What problems did you have with the RAID? You using an ASUS A8V?

[saaya]

i also noticed that the 940 fx is slightly faster in some benches than the 939 fx... weird... is it the buffered memory?

[StormPC]

AMD's in general have been proven to not benefit from more than a certain amount of bandwidth. That's why the single channel S754 is not far behind the DC S939 (w/FX-53) and S940. Once you get PC5000+ bandwidth it makes no difference at all, so the registered memory is not the handicap many thought it was.

[misteroadster]

when you run at 2400 12X200
or 2400 10X240 there is a good gain on 3800+ , bandwith is not so useless than that.
do you want i do the test ?

[StormPC]

There is a good gain in pi calculations and Sandra mem bandwidth but the gain in 3DMark is far less.

What video cards do you have? We can do our own little test if you like.

[misteroadster]

for sure your gain will be much higher with increasing 200->240 , but you can't obtain 7K BP
and my 9700 is a bottleneck