Dare To Compare

aristide1 · Post by **aristide1** » Fri Jan 18, 2008 1:06 pm

http://www.madshrimps.be/?action=getarticle&articID=603
An article on cpus with same speed but different cache and FSBs.

http://www23.tomshardware.com/cpu_2007.html
On Toms Hardware you can go to the cpu page and compare an AMD 5200+ Windsor with a 5000+ Brisbane. Both processors run at 2.6 MHz stock, but the Windsor has 2*1MB of L2 cache, while Brisbane has 2*512K of L2 cache. Then check the PC mark and Sis Sandra tests. You can also compare an E4xxx Intel at 1.8GHz with an E4xxx also at 1.8GHz, the only difference there is 1MB total versus 2MB total cache.

http://www.madshrimps.be/?action=getarticle&articID=636
This article shows that quad Penryns use a little less power than a comparable E6xx series, which is nice because you pick up 2 more processors for no power increase. But both get a little hungry when pushed passed 3.2 GHz.

djkest · Post by **djkest** » Fri Jan 18, 2008 2:09 pm

It's interesting the Windsor @2.6 GHz and 512 cache is still faster than the brisbane. But in any case, the difference in 1MB cache vs. 512 cache appears to be 1%... and that is with the different core also. The windsor is also using more power.

Hmmm... no difference w/ intel?

Then again, this is Tom's Hardware. These guys have been caught cheating at least once, so I'm not sure how much I trust them.

aristide1 · Post by **aristide1** » Fri Jan 18, 2008 4:01 pm

djkest wrote:It's interesting the Windsor @2.6 GHz and 512 cache is still faster than the brisbane. But in any case, the difference in 1MB cache vs. 512 cache appears to be 1%... and that is with the different core also. The windsor is also using more power.

I believe this matches what the AMD engineers said when they were asked why only 512K cache? Doubling it didn't bring huge increases in anything except cost.

djkest wrote:Hmmm... no difference w/ intel?

There are more than a few who will tell you flat out the E21xx feel slow. Sure they OC like crazy, but they have a lot to make up for.

djkest wrote:Then again, this is Tom's Hardware. These guys have been caught cheating at least once, so I'm not sure how much I trust them.

Not sure what specifically you are referring to. I doubt they would be the first either. One thing is certain, they often review things like motherboards and make a huge deal about incredibly small differences in performance.

djkest · Post by **djkest** » Fri Jan 18, 2008 6:33 pm

What I'm referring to is when the Athlon first came out Tom's was busted for rigging tests to make the AMD CPUs CATCH ON FIRE, then smearing AMD over it(they posted a video), until the truth came out. It was proven that they had disabled some sort of thermal protection or put something flammable on the CPU, I can't remember which. There have been a few other cases, but I can't recall them off the top of my head.

I started folding on T5450 2x 1.66GHz and it's a little slower than I thought it would be.. but oh well. It's still faster per clock cycle than the brisbane next to it.

aristide1 · Post by **aristide1** » Sat Jan 19, 2008 4:28 pm

djkest wrote:What I'm referring to is when the Athlon first came out Tom's was busted for rigging tests to make the AMD CPUs CATCH ON FIRE, then smearing AMD over it(they posted a video), until the truth came out. It was proven that they had disabled some sort of thermal protection or put something flammable on the CPU, I can't remember which. There have been a few other cases, but I can't recall them off the top of my head.

That's just ridiculous. Frankly AMD should have sued. I only use TH for the PC Mark and SIS Sandra results, stuff that anybody could spot as fraudulent in a heartbeat.

djkest wrote:I started folding on T5450 2x 1.66GHz and it's a little slower than I thought it would be.. but oh well. It's still faster per clock cycle than the brisbane next to it.

Hmmm, I'm going to ask you to be more specific again. You're telling me your Black Edition takes longer per checkpoint than your T5450?

djkest · Post by **djkest** » Sat Jan 19, 2008 5:00 pm

No no... sorry. I'll give you a breakdown. These are all on the same types of work units (the 500 pointers)... console client. I need to get SMP up and running again.

T5450 @ 1.66GHz C2D, 2GB DDR2 667.... 34.8 min
5000+ @ 3.0 GHz bris, 2GB DDR2 800 ... 25.5 min

and now...
E8400 @ 3.3GHz, C2D, 2GB DDR2 800 ... 16.7 min

aristide1 · Post by **aristide1** » Sun Jan 20, 2008 7:28 am

djkest wrote:No no... sorry. I'll give you a breakdown. These are all on the same types of work units (the 500 pointers)... console client. I need to get SMP up and running again.

T5450 @ 1.66GHz C2D, 2GB DDR2 667.... 34.8 min
5000+ @ 3.0 GHz bris, 2GB DDR2 800 ... 25.5 min

and now...
E8400 @ 3.3GHz, C2D, 2GB DDR2 800 ... 16.7 min

OK, but I don't think that if you doubled the speed of your T5450 you would decrease the time of eack checkpoint by half.

The E8400 is impressive for non-SMP points.

djkest · Post by **djkest** » Sun Jan 20, 2008 2:08 pm

I had to quit the SMP client because I kept getting these monster work units. I'd crunch on them 3 days, get to about 85%, and then the deadline would pass and the whole 3 days was wasted. So I'm doing the command line client again, for now.

What I really need is 1) get rock stable cooling/ configuration for my new setup maximizing CPU speed 2) figure out what the heck kind of SMP settings I should be using, cause whatever I had before wasn't really working out for me.

aristide1 · Post by **aristide1** » Sun Jan 20, 2008 8:24 pm

djkest wrote:I had to quit the SMP client because I kept getting these monster work units. I'd crunch on them 3 days, get to about 85%, and then the deadline would pass and the whole 3 days was wasted. So I'm doing the command line client again, for now.

What I really need is 1) get rock stable cooling/ configuration for my new setup maximizing CPU speed 2) figure out what the heck kind of SMP settings I should be using, cause whatever I had before wasn't really working out for me.

Their should be some way to avoid those 2530 pointers. Some option in the advanced options that you need to say yet to access. Have you tried asking on the FAH forums????

djkest · Post by **djkest** » Sun Jan 20, 2008 8:35 pm

I'm not even sure how this compares, but this is a PMD core.... this thing is ripping through the checkpoints at 3.80 GHz... 8.5 min per checkpoint. This has got to be one of those crappy 300 pointers... and yeah, this thing is running only 50%. My tests with 1 hour of Prime 95 25.5 shows it should be stable once I start up the SMP client.

The new 45nm quads are gonna be sick.

aristide1 · Post by **aristide1** » Mon Jan 21, 2008 4:37 pm

1.264 volts. Ain't that a lot?

Where did you get it? I won't ask what you paid for it.

djkest · Post by **djkest** » Mon Jan 21, 2008 9:28 pm

$189.99 at a "local" store called Micro Center. I had to drive 70 miles each way to get it. Total trip took about 3 hours.

Stock voltage is 1.225V .. which shows up as 1.15V at the motherboard. I think I've seen people running like... 1.4V with these already.. seems like a lot.

Unfortunately at 3.8GHz it crashed after folding for 4 hours... yay... it was prime stable for an hour, I thought that was good enough. So I bumped it down to 3.6Ghz for a little while.

aristide1 · Post by **aristide1** » Tue Jan 22, 2008 7:19 am

djkest wrote:$189.99 at a "local" store called Micro Center. I had to drive 70 miles each way to get it. Total trip took about 3 hours..

You drove 3 hours to get one? You're serious about this.

djkest wrote:Unfortunately at 3.8GHz it crashed after folding for 4 hours... yay... it was prime stable for an hour, I thought that was good enough. So I bumped it down to 3.6Ghz for a little while.

"Round" errors in Prime 95 can produce invalid results in folding, according to the documentation that comes with Prime 95. While my one system seemed stable I ran Prime 95 and it got a rounding error after eight hours. A small increase in the NB voltage and the system passed 24 hours of Prime 95.