Mattspalace's 540J Review with benchies [Archive]

mattspalace

02-26-05, 02:22 PM

So Mr. Sen and I have been talking about 4 hours a day for the last three days or so and we've both been tweaking our rigs non-stop. Yes, we are true o/c'ers. :)

Initially with my 540J I was running 3.9GHz with a 1:1 ratio and a GB set of Mushkin PC3200 Level II (the BH-5 stuff) with 2-2-2-6 timings. Not bad at all, but I knew my CPU had more in it, so I swapped it out from some PQI PC4400 I've had for a long time now. Not the greatest memory made, but it will run a solid 280FSB error free in memtest.

So - I swap out the Mushkin for the PC4400 (no it's not TCCD)

4GHz was a snap. I did that at 1.4v, but I had to find the true ceiling of this CPU. I'm comfortable going up to ~1.5v on air, as I've got very good air cooling (case included), and my puter room stays about 70F.

After some serious tweaking, I have topped out at 4.2GHz 1:1 with this PQI PC4400 running 3-4-4-8 timings (timings are not a huge deal on PIV's if you didn't know unlike A64's)

So here's a screenie at 4.2GHz and some benchies to go along with it.
I'm satisfied with a 1GHz overclock, so I'm finished.

If you want any other benchies, or have any questions, lemme know.
I've had this CPU for a couple of months now.. I'm impressed. It's made in China too.

Matt

http://www.pixhost.com/pixa/agent_008/4200mhz.jpg

Benchies:

1M digits in superpi = 32sec
32M digits in superpi = 27mins 03secs
Sandra bandwidth = 6200mb/s
3DMark2001 = 28059
3DMark2003 = 14779
PCMark2004 = 6562
Sandra CPU Arithmetic = 12472 MIPS / Dhrystone
Sandra CPU multimedia = Integer 30131 / Floating Point 40303
AquaMark3 = 82,629

Temps:
36c idle
58c load

mattspalace

02-26-05, 02:29 PM

Sen and I do plan on doing a comparison once he's finished tweaking. We are both curious about the differences between 1mb of l2 cache and 2mb of l2 cache.

Sentential

02-26-05, 03:42 PM

Sen and I do plan on doing a comparison once he's finished tweaking. We are both curious about the differences between 1mb of l2 cache and 2mb of l2 cache.
You are gonna have to scale yours back some and loosen the RAM timings for a fair comparison... my CPU jsut flat out wont do 4.2 :bang head

Overclocker550

02-26-05, 04:26 PM

"3DMark2001 = 28501 (non oc'd x800 PE)"

good god thats one heck of a good score, is yours a 1mb cache? even better!
I dont see your score on the orb so please publish it and link me.

Sentential

02-26-05, 04:27 PM

Sen and I do plan on doing a comparison once he's finished tweaking. We are both curious about the differences between 1mb of l2 cache and 2mb of l2 cache.

http://img.photobucket.com/albums/v180/Sentential/p4bench.jpg

mattspalace

02-26-05, 06:57 PM

OC550 - I never registered the score, but I'll run it again and post a link for you.

mattspalace

02-26-05, 10:28 PM

AquaMark3 scores coming soon - downloading it now.

Sentential

02-26-05, 10:29 PM

AquaMark3 scores coming soon - downloading it now.
If you could do an 01SE score @ 252 with 3-4-4 i'd appreciate it. That should give us an idea if the extra meg cache helps or not

mattspalace

02-26-05, 10:30 PM

Can do!

hawtrawkr

02-26-05, 10:34 PM

Sen and I do plan on doing a comparison once he's finished tweaking. We are both curious about the differences between 1mb of l2 cache and 2mb of l2 cache.

what?!?!?!?! you mean im not the only one curious about this!

when i saw sens thread go up i thought great ill get to see some first hand results from someone who knows how to oc well. then i think bah hes never had a prescott before to compare it to! :temper:

Sentential

02-26-05, 10:37 PM

what?!?!?!?! you mean im not the only one curious about this!

when i saw sens thread go up i thought great ill get to see some first hand results from someone who knows how to oc well. then i think bah hes never had a prescott before to compare it to! :temper:
LOL... Ill find somone to get a comparision with. Unfortunatly all I can do is a run vs a Winchester

hawtrawkr

02-26-05, 10:40 PM

LOL... Ill find somone to get a comparision with. Unfortunatly all I can do is a run vs a Winchester

hell id send you one of my prescotts to test but theyre all s478 :shrug:

mattspalace

02-26-05, 10:48 PM

Sen, I suggest we use Sandra Arithmetic scores for comparison since we have different motherboards and different videocards..

that work?

AquaMark3 scores posted.

Sentential

02-26-05, 10:53 PM

Sen, I suggest we use Sandra Arithmetic scores for comparison since we have different motherboards and different videocards..

that work?

AquaMark3 scores posted.
I get 11,843 on the arithmetic. However I still want to see your 3dmark reguardless. There are tests that are 100% CPU driven and not vcard based

mattspalace

02-26-05, 10:55 PM

Okay, I'll do both

BigStan

02-26-05, 10:57 PM

I get 11,843 on the arithmetic. However I still want to see your 3dmark reguardless. There are tests that are 100% CPU driven and not vcard based
What fsb are you getting 11,843 at?

mattspalace

02-26-05, 11:06 PM

26,570 on 3DMark2k1 http://service.futuremark.com/compare?2k1=8451571

Sandra Arithmetic = 11,943

That's with your requested 252FSB and 3-4-4 timings

mattspalace

02-26-05, 11:31 PM

I think I had a typo in my first 3DMark2k1 post. I think it was actually 27501 marks, not 28501.

I did just install the latest catalyst drivers and now my score is 27734

Here's a compare link: http://service.futuremark.com/compare?2k1=8451592

mattspalace

02-26-05, 11:38 PM

man_utd

02-26-05, 11:42 PM

Recommendations: Drop resolution in all video type benchies to lowest it can do, to eliminate the video card as much as possible. Then again, on two high end cards, it might just make it worse.

mattspalace

02-26-05, 11:54 PM

Sandra multimedia benchmark maybe?

warlock110

02-27-05, 02:25 AM

i would do sandra, it's good when comparing the CPU cuzz that thing don't do nothing with graphic card. and post all the data in 1 thread :) it's so hard reading it like this.

glock19owner

02-27-05, 02:27 AM

Matt...nice OC :thup:

But how come so low in Super Pi?

"1M digits in superpi = 32sec"

To me this should be under 30 seconds with your OC...

jenko

02-27-05, 04:13 AM

Matt...nice OC :thup:

But how come so low in Super Pi?

"1M digits in superpi = 32sec"

To me this should be under 30 seconds with your OC...

I would say maybe the loose memory timings?
How about downloading 3dmark05 and just both run the 2 cpu tests?

hawtrawkr

02-27-05, 08:00 AM

32s sounds right for 4.2ghz even with 2-2-2-5 i think hed be at 31 or 30 at the lowest.

Sentential

02-27-05, 09:22 AM

Well that didn't explain much. My cpu wins in Sandra Arithmetic and your's wins in 3DMark2K1..

Which is more accurate? Maybe I should overclock my PE just a little and see how much my score increases..
Thanks for the bench. Im gonna give the numbers to oc550 he can decipher them. I can say right off the bat already that the 600s are faster, abeit not THAT much faster, but it looks to me like its about 5-7%

Sentential

02-27-05, 09:26 AM

i would do sandra, it's good when comparing the CPU cuzz that thing don't do nothing with graphic card. and post all the data in 1 thread :) it's so hard reading it like this.
Not really. Look at lobby they are damn near equal indicating same memory speed. Ignore dragothic and nature. Look car low. Mine pulls ahead.. by what you ask? 5% Just as Intel said it would

Sentential

02-27-05, 09:34 AM

You gotta take a look at the benches like this for it to make sense:

http://img.photobucket.com/albums/v180/Sentential/540Jvs640.jpg

Zebbo

02-27-05, 09:53 AM

Nice to see some nice overclocks from both of you guys. But 3DMark01 is really something you should not use when comparing CPU's. I suggest something like Quake3/Doom3 with low 640x480 resolution with 16-bit colors and SuperPi 8M / 32M.

I know you guys will figure something out :)

hawtrawkr

02-27-05, 10:09 AM

01se would be a better gauge if they both had the same videocards and memory clocked the same as the diffrence would then be nothing but cpu. even with the exact same system if you go from a 6800 to a x800 or vice versa the scores are going to vary though.

PCM0nk3y

02-27-05, 11:02 AM

I think that both should run 3dmark03 because it will give you a CPU score, which will tell us which is the faster of the 2.

mattspalace

02-27-05, 11:10 AM

Hey aquamark3 does that too. I just didn't write it down.

mattspalace

02-27-05, 11:18 AM

Yeah I don't really like going with 3dmark2k1 since we have different video cards.

Sen, can you run Aquamark3 and note your cpu scores for me?

mattspalace

02-27-05, 11:38 AM

Sen, Here are a couple of CPU scores from AQM3

AT 252FSB = 11,987
AT 262FSB = 12,511

Run your 640 at 252 and see what your CPU score is. I think this is going to be a better comparison than 3DMark2k1..

Sentential

02-27-05, 12:51 PM

Zebbo

02-27-05, 12:58 PM

hawtrawkr, you are right there. Would be good if both of guys could get same memory and video card before this kind of comparing.

No matter what, Sen's result would be better in both, Doom3 and Quake3 if they'd bench 'em. NVidia kicks ATI's ass on those games.

glock19owner

02-27-05, 01:21 PM

32s sounds right for 4.2ghz even with 2-2-2-5 i think hed be at 31 or 30 at the lowest.

How is that right when a AMD A64 running at 2.75Ghz beats it by almost 2 seconds :confused:...To me that just doenst sound right when I have seen other 4G setups run well below 30 seconds...

http://home.insightbb.com/~myglock19/SPMod.JPG

mattspalace

02-27-05, 01:29 PM

they must be running a divider with tight timings..

Actually that score is just over 1sec faster..

hawtrawkr

02-27-05, 02:17 PM

i havent seen many p4 systems run well below 30 seconds except on the prescott patched super pi.

PCM0nk3y

02-27-05, 02:25 PM

hey have you guys thought about maybe trying out PCmark04? that's pretty CPU oriented and very little to do with graphics card.

glock19owner

02-27-05, 02:30 PM

they must be running a divider with tight timings..

Actually that score is just over 1sec faster..

I am talking about some members from XS...not here...that could be the main reason why right there though ;)...

But it just doesnt seem logical that a 4.2 CPU gets beat by a 2.75 CPU...

(not making this into a A64 vs Intel post, so dont make it sound like this is)...but was more curious as to why a A64 would/can beat a 4.2 in calculations :confused:

hey have you guys thought about maybe trying out PCmark04? that's pretty CPU oriented and very little to do with graphics card.

PC Mark 02 would be better since it list CPU, Memory, HDD's serperatly...

(Sent, Matt...IM me if you two do run 02...I have something that you both would want/need for it :D)

hawtrawkr

02-27-05, 02:42 PM

my winchester @ 2700mhz can score 30seconds but it takes me 4.4ghz on my intel to get the same.

glock19owner

02-27-05, 02:55 PM

So in the long run...when it comes to the calc part of the CPU's...the 4+G Intels are more for bragging rights with the OC then it is for actual calculations (given that Intels wipe A64s on multimedia)...when in fact a A64 is going to be physically faster with lower Ghz in CPU?

Once again...not making this into a A64 vs Intel post...but trying to figure out a few things...debating on building a Intel...but if the A64's will still be faster in CPU at almost half of what a Intel will run...then its kind of a waste for me...

Sorry to hijack the thread Matt...but your thread brought up some good questions for me...

mattspalace

02-27-05, 02:57 PM

No problem glock. I think A64's run SuperPi so fast because of hypertransport.

glock19owner

02-27-05, 03:01 PM

Thought the new Intels supported that also...like yours and Sents?

hawtrawkr

02-27-05, 03:01 PM

PCM0nk3y

02-27-05, 03:05 PM

honestly I think that benching isn't the best way to compare intel and amd. I think it's kind of like the comparison between a 6800 ultra and a x800xt, both of them have games that run A LOT better than the other card. and I think that some benchmark programs are the same way. some lean more towards AMD and others more towards Intel, just my thoughts tho.

glock19owner

02-27-05, 03:05 PM

i think amds have a advantage on the smaller superpi calculations and the higher clocked prescotts are better for the bigger ones.

i see alot of fx-55 with a little over 3ghz scoring 28sec 1m

If thats the case...I wouldnt have a issue seeing how these 2 compair on a 32mb calc if either sent or matt would be willing to run a 32mb calc...

I would be interested in seeing the results...becuase if a A64 is still faster or close to even...then it is a waste on my part to spend the extra money on a Intel setup right now...

And I have seen FX's under 3G running at around 26seconds....sucka's

mattspalace

02-27-05, 03:06 PM

Thought the new Intels supported that also...like yours and Sents?

Nope, they still use a northbridge. A64's don't need it though...less communication outside the cpu = faster.

mattspalace

02-27-05, 03:07 PM

Yeah, I'll run a 32mb bench

glock19owner

02-27-05, 03:08 PM

Nope, they still use a northbridge. A64's don't need it though...less communication outside the cpu = faster.

Ok...this is where I was getting confussed...because I was thinking both supported HT...now this is making more sense to me...:D

honestly I think that benching isn't the best way to compare intel and amd.

Once again...not making this into a A64 vs Intel post...but trying to figure out a few things...debating on building a Intel...but if the A64's will still be faster in CPU at almost half of what a Intel will run...then its kind of a waste for me...

Please try to choice your words a little better...because around here..this is all it takes to start this thread into a flame war...

I personally would hate to see Matt's thread get turned into this...this is why I stated what I did a few times...because I am more on a fact finding mission...then a compair/war mission ;)

Thanks :)

Sucka

02-27-05, 03:32 PM

i see alot of fx-55 with a little over 3ghz scoring 28sec 1m

I score 26 at 3.1Ghz, only needs like 3.0 for 27.

mattspalace

02-27-05, 03:37 PM

glock, here's my 32M superpi calc time: 27min03sec

took that sucker awhile to run!

mattspalace

02-27-05, 03:38 PM

Hyperthreading and HyperTransport are VERY different.

PCM0nk3y

02-27-05, 03:57 PM

Ok...this is where I was getting confussed...because I was thinking both supported HT...now this is making more sense to me...:D

Please try to choice your words a little better...because around here..this is all it takes to start this thread into a flame war...

I personally would hate to see Matt's thread get turned into this...this is why I stated what I did a few times...because I am more on a fact finding mission...then a compair/war mission ;)

Thanks :)

um....sorry my opinion about benching gave you the effect that I was trying to turn this into an AMD vs Intel flamewar again. I was simply trying to say that I think some programs favor a certain processor over others, I don't recall trying to say that either Intel or AMD was better than the other...I think my statement was totally unbiased and not worthy of being called a flamewar comment.

my apologees matt, I'm not trying to turn this into any type of flamewar

Sentential

02-27-05, 04:07 PM

(Sent, Matt...IM me if you two do run 02...I have something that you both would want/need for it :D)
Sure thing. Next time I see you on AIM I will

glock19owner

02-27-05, 04:56 PM

um....sorry my opinion about benching gave you the effect that I was trying to turn this into an AMD vs Intel flamewar again. I was simply trying to say that I think some programs favor a certain processor over others, I don't recall trying to say that either Intel or AMD was better than the other...I think my statement was totally unbiased and not worthy of being called a flamewar comment.

my apologees matt, I'm not trying to turn this into any type of flamewar

My bad...was just leery on this turning into a A64 vs Intel thread...I will applogize for my comment too you PC :D...

And Sent...you know how I am on AOHELL ;)..

Hyperthreading and HyperTransport are VERY different.

I understand these are 2 totally different structurs...but I was under the impression that the new EE's was using hypertransport...

This is where I was getting confussed with everything...this clears everything up for me now...

Also I will post my 32mb calc later tonight...Dont think super pi and dvd encoding will play together very nicely...even with a Intel ;)...

mattspalace

02-27-05, 05:27 PM

No idea about the PIV EE's. Them thar CPU's are too much money for me. :D

mattspalace

02-27-05, 06:10 PM

Just FYI, I'm adding all my bench scores to the original post.

mattspalace

02-28-05, 04:15 PM

glock, did you ever run that 32M superpi?

mattspalace

03-02-05, 06:21 PM

Reran 3dmark2k1 with my new modded x800xt pe

OC550, here's a compare link for ya - http://service.futuremark.com/compare?2k1=8458455

glock19owner

03-02-05, 06:51 PM

glock, did you ever run that 32M superpi?

I havent forgotten...just been busy with my own record 01 run ;)...problem is...lets just say...I am on X64 now :(...

Not sure what is going on...could be a few things...but w2k is history now and my Pro wont run over 450...have a new PSU comming...should be here by saturday...so hopefully the new PSU and reformatting and reinstalling W2k will get my system back up to normal...then I will finish try to finish my 01 runs and then get on Super Pi...

I want that 27k world record on a air cooled, non v modded Pro in 01 first...I am sure you can understand :)...

Just seen your score...my little air cooled pro is only 700 marks from you ;)