3.0GHz vs. 3.3GHz

S

Shorty

Gawd
Joined
Aug 5, 2003
Messages
791
I finally got up the balls to attempt to overclock my Pentium 4 2.4C up to 3.6GHz. Regularly, it is at 3.0GHz (1.5375 vcore) and I wanted to juice it up. I'm using DDR400 so I switched from a 5:4 ratio to a 3:2 ratio. I threw up the vcore to 1.6 and noticed it fail Prime95 stress test before the first test would even complete. I then noticed how CPU-Z listed my vcore at around 1.552, so I upped the vcore to 1.7 and now CPU-Z reads a vcore of 1.668. I ran Prime95 and it failed again. I kept the voltage and switched the bus down to 285MHz (190MHz memory). It failed, so I threw down the bus to 275MHz (183.4MHz memory) and ran Prime95. I watched the rest of the Rose Bowl and the entire Orange Bowl, and I arrive to see Prime95 still stressing my processor a little more than four hours later. I like what I see, especially for a 1.6 vcore (1.552~1.568). Now, my question is... which situation would performance better in?

#1
2.4GHz @ 3.0GHz
1.0GHz Bus (250MHz)
250MHz X 12 = 3.0GHz
5:4 Ratio
Memory ~ 200MHz
Latency: 2-3-3-6

OR

#2
2.4GHz @ 3.3GHz
1.1GHz Bus (275MHz)
275MHz X 12 = 3.3GHz
3:2 Ratio
Memory ~ 184MHz
Latency: 2-3-3-6

I tried Sandra and I get a 5134/5140 memory benchmark with situation #2, which is funny since in situation #1, I get in the high 4000s. Will the gain of processor MHz make up for the loss in memory MHz. Or will the memory perform faster now at 184MHz along with the faster processor than it would at 200MHz with the slower processor? Basically I'm worried about the memory and whether that is going to handicap my computer. I'm going to try to squeeze out a 2-2-2-5 latency at 184MHz with my RAM... try 2.8 volts this time.
 
#2
2.4GHz @ 3.3GHz
1.1GHz Bus (275MHz)
275MHz X 12 = 3.3GHz
3:2 Ratio
Memory ~ 184MHz
Latency: 2-3-3-6


im runnig really similiar to that right now and i think it runs perfectly fine but then again my timings are a lil slower. but up to a point ure cpu speed shouldnt really matter i'd try to get tighter timings just my 2 cents.
 
Good question. (And on a side note, very nice job of laying out the situation.)

I'll also vote for option two, but it's a qualified endorsement. In the great majority of applications CPU speed is more important than memory speed. (And memory speed is more important than tight memory timings. More on that in a moment.)

Now you're talking about a ten percent improvement in processor speed, which is barely enough to show up on your screen in application performance. But while memory speed isn't as important as processor speed in most cases, you've induced a hell of a gap between the memory and CPU speeds in these two examples. That will take some of the edge off the processor performance.

I'd predict that you'd see a slightly better system benchmark score with option two, but I don't think you'd ever be able to tell the difference in a real application.

But what can you get if you loosen up the memory timings? If you could get your memory to run at 220 (275FSB on a 5:4 divider) you'll probably see the best overall performance you're going to find with this setup.




BHD
 
2-3-2-6 at 2.75 volts looks like it. I tried 2-2-2-5 at 2.8 volts and Windows loaded up at 800x600 with default video drivers and messages stating that Windows has recovered from a fatal error. I might try 2-2-2-6 and 2-3-2-5 tomorrow, but I think 2-3-2-6 is where it's at...

The memory is Corsair XMS 3200C2 (2x256MB). The farthest I can go without going total error ridden in Memtest86 is 208MHz at 2.5-3-3-7. Can't go any higher even with 3-4-4-8 without Memtest86 giving me tons of errors. Even with a 5:4 divider, I'd only get 3.12GHz.
 
Yeah, on its best day Corsair XMS3200 isn't good overclocking memory for an 800MHz system.

You have a sweet rig but I'm actually glad I'm not in your shoes. Having memory that hit the wall at 208 would drive me nuts if I had a proc that could run 275 on relatively low voltage.




BHD
 
You must log in or register to reply here.
Back
Top