Cheap Processor Upgrade? - Page 26

http://searchnetworking.techtarget.c...887152,00.html

Noodle

He's talking about CPUmarks. You can filter the results to just show what the two cpu intensive tests' results were. I think you'll find if you try it you'll match a few fx-55s and fx-53s.

3DMark Score4976 3DMarks
CPU Score5736 CPUMarks

Yes, 3dmark05 stresses the GPU almost independently from the CPU in the Game tests. But with the CPU tests it reverses the test and reduces the GPU load to nearly zilch by turning off loads of fancy effects and instead putting lighting and shadowing on the CPU, along with lowering the resolution to like 640x480. Video card will make little difference here, unless you are running like a Geforce FX 5200 or lower.

LucidChaos, how did you score 5,736 CPU points?!!? That's like 3,000 higher than my 2.13GHz! That seems off a bit lol. You would be in like 5GHz P4 territory there haha. I see you have newer beta video drivers than I. They might be doing something wrong..
http://service.futuremark.com/compare?3dm05=894338

BTW, last night I went in and changed my CPU's thermal pad for some Arctic Silver 5. I think it dropped 3-4 degrees. A small improvement. And, I strongly recommened everyone buy some ArticSilver Arcticlean because that stuff cleans the old pad and paste off miraculously (and it's not carcinogenic, narcotic, caustic terrible crap. In fact, it's citrus-ish! It smells GOOD!). For $6 and the fact you only need a few drops (!) to clean anything, it's AWESOME! Within a couple of seconds it just dissolves the old pad and it looks like you have a brand new surface to work with. Part 1 just annihilates that nasty old pad or paste and Part 2 cleans finally and prevents oxidation.

Only two things I can think of that you would be significantly different on:

1. I did bench at 2.66ghz... thats worth a cpumark or two

2. In order to keep up with the people on top, I did change my LOD setting in rivatuner to +3. That makes everything look like absolute hell, but things do run a tad bit faster. For screenies (and some good reading) visit this thread on xtremesystems: http://www.xtremesystems.org/forums/...ad.php?t=64078
This should explain why some scores on ORB are so much higher than others with similar configurations.

Well that's depressing. Pretty much makes 3dmark useless as a comparison. All my scores are with defaults across the board, well except for clocks. I use 3dmark as a way to collect benchmark info for myself. Especially for 3dmark2001 as I have about 50 results collected for various systems.

You're absolutely right. 3dmark no longer measures your actual performance, but instead measures how much time you spend researching how to score higher in 3dmark. Before reading the 9300 5k club thread, I maxxed out at 4512 3dmarks. Now with some tweaking I can just about hit 5000.

Sad that 3dmark is the de facto standard for 3d performance benching; that there isn't some bench program out there that won't allow you to make it look god-awful just to score a few more points.

If you are gonna only bench the CPU why not use PCMark04, won't it give a more detailed analysis?

I did. Look below.

Just an update but those of you who know me then you guys know that I turned a lovely 1.7ghz cpu into a 2.26 Monster well id like to update by saying ive gotten it to run at 1.324 V now compared to the 1.340 V it used to run. Anything lower resulted in BSOD (blue screen of death) considering im running it 133mhz past the best pentiumM available from intel I think thats good stuff! those who are afraid of 1.7ghz cpus dont be! there great!

Now that's strange... My 1.8 750 ran at 1.308v by default. You would think the faster chip would have the higher default voltage.

Thanks to this thread I am officially Overclocked! Pentium M 725 1.6GHz to 2.13GHz.

I agree, I had mine running at 1.292, but decided to give it a little voltage buffer, so I am also running at 1.324. Power to the M 735s!

Woot! just intalled the 1.7 last night for a sweet 2.26ghz. Im currently running at 1.308 and it ran just fine in prime95 for a little less than 10 hours. I havent gone any lower though, might try i dont know. will do a full 24 hour test soon at my current voltage and see how it does.

do u think 10 hours is long enough or should i continue on my plan for a 24 hour test? Also which setting under tortue test should i use? I used blend, but should i be using Small FTTs or In-place large FTTs instead?

REAVER, How long did you run prime95 @ 1.292?

Horsefly, I would say 10 hours is enough. Since if anything you're going to get an error from too little voltage, based on my experience, the small FFTs are what will give you errors. I would still run the blend test though, as it tests everything.

k, thanks lucidchaos.
I'll keep everyone updated on low i can get the voltage w/o errors

Just under 10 hours on the small FFT torture test.

There is a 1.6 engineering sample on ebay for a buy it now price of $99.
ES chips are multiplier unlocked and if you can find a way to change the multiplier, it might be a cheap upgrade.

http://cgi.ebay.com/ws/eBayISAPI.dll...774415970&rd=1

Doesn't matter, they still only go up to multiplier 16x

Also, I recieved my 1.6 400fsb chip yesterday and popped it in my 9300. Wire trick and all. 2.13ghz without a problem, all for the low low price of $140 shipped.
I'm running it full load with 1.2v now. I'm trying to bump the vcore down everyday until to find the lowest Vcore it will run at stably.

with this wire mode, are you guys still able to run at 6x multiplier/ 0.7v?

I can't. I just set mine to .7v and it gave a blue screen and rebooted on me.

.9v runs fine though. I just fired up NFSU2 and tried it for 5 minutes and it didn't lock up. Maybe I can go lower...

With battery 1.2v seems comfortable for 2.13ghz. If I drop to 1.1v it freezes up. 1.15v freezes up too.