Paid more to go slower, doesn't that suck

[Tomer]

I recently upgraded my older Soyo Dragon MB and 1600+ processor to a Soyo SY-KT600 Dragon Plus and a Barton 2500+

While I gained a significant boost in performance over the old setup (Around 30%, I have an ECS K7S5A running a 2400+ sitting next to my new box that is kicking this new setups ass.

I have ran several CPU benchmarks comparing the 2 machines at stock FSB settings and the 2400+ is beating the 2500+ by about 7 or 8%

I have jacked up the FSB on the new board to 180Mhz which is giving me comparative performance between the two machines now, but it really burns my ass that my 6 month old machine is faster than my new setup.

Was I a rube for spending more for the 2500+ processor instead of just getting another 2400+?

It was my fault for not checking closer, but after seeing the 2500+ runs at a 1.83Mhz compared to the older 2.4 that is running at 1.99Mhz it really kind of pisses me off. Seem like false advertising to me.

Tomer

 

[formulav8]

What you have to understand is that some app's like clockspeed more than the extra cache. In cache intensive apps the 2500+ will beat the 2400+ due to the 2500+ having double the L2 cache. Sometimes the apps perform better when the clockspeed is higher. But for the most part, the 2500+ will beat the 2400+ hands down.


Jason

 

[mechBgon]

Let me guess... you're running benchmarks in :roll:Sisoft Sandra:roll: Now run some UT2003 Botmatch benchmarks.

(also, I hope your RAM is PC2700 or higher, and not PC2100? The 2500+ needs at least PC2700.)

 

[Tomer]

Thanks for the feedback

Video encoding (DVD Shrink) and Video Editing are my main CPU eaters. I will do a side by side encode of Revolutions this evening to see what happens. I know I did that awhile ago with a 2000+ then dropped in a 2400+ and the numbers were like 5 minutes difference in 6 hours of encoding.

The numbers I looked at were from both Sandra and Passmark. They were consistantly about the same % difference.

The RAM is 2 sticks of 512Mb PC2700.

Tomer

 

[mechBgon]

Good to hear that your RAM is PC2700 Encoding apps may not be the best showcase for the 2500+ with its lower clockspeed, I agree. The extra cache of a 2500+ helps in some stuff, the extra MHz of a 2400+ helps in others. I think it's a bit optomistic to expect noticable differences when even AMD has rated the two CPUs only 100 points apart, but at least use your real-world tasks as the final judgement and not synthetic what-do-you-want-to-see-today-sir? benchmarks like Sandra.

 

[Tomer]

Yeah, I agree. Didn't really expect a big difference, but you can imagine my surprise when the new one ran quite a bit slower in my inital test. As an update, I rolled up the 2500+'s FSB to 180. Seems to still be stable and now BM's about the same as my 2400.

Next time I'll do better research.

Now if I could just have a word with the a-hole that positioned the IDE connectors on my Soyo Dragon Plus as far away from the drives as possible, thereby making a standard IDE cable about 1/4' too short.

 

[SickBeast]

hahaha what a joke you're complaining about a 2500+ being SLOWER than a 2400+!? the 2500+ is vastly superior. it has double the L2 cache. if you had better memory it would more than likely run 2200mhz which is 3200+ speeds. you do realize that the 2500+ chip has probably 20 million more transistors than the 2400+ right?

try running a UT2003 benchmark as someone else noted then come back and complain.

 

[bootoo]

I had my 2400+ Tbred B (at 2.25 ) until I got a 2500+ barton. At stock I noticed a nice difference with the barton on lots of app's thanks to the L2 cache. You will eventually be glad.

 

[Tomer]

Although you may consider it superior, and it may well be in certain cases, in mine case it simply is not.

I just completed a DVD encoding session with both processors. Basically the CPU goes to 100% and stays there. The 2400+ took 4 hours 28 minutes to complete. The 2500+ doing the exact same task took 4 hours 52 minutes.

I will run the UT2003 benchmark and see what happens.

 

[Soulkeeper]

most 2500+ can run over 2ghz on default voltage btw
just bump it up to 2ghz or so

the extra cache is always good, but clockspeed means faster cache (more bandwidth, lower latencies)

some apps can manage a 10% or more boost from the barton at the same clock speed, but generally speaking they are pretty close clock for clock

 

[rogue1979]

Originally posted by: SickBeast
hahaha what a joke you're complaining about a 2500+ being SLOWER than a 2400+!? the 2500+ is vastly superior. it has double the L2 cache. if you had better memory it would more than likely run 2200mhz which is 3200+ speeds. you do realize that the 2500+ chip has probably 20 million more transistors than the 2400+ right?

try running a UT2003 benchmark as someone else noted then come back and complain.

The biggest advantage of the stock 2500+ over a stock 2400+ is the fsb speeds. 333MHz beats 266MHz handily. The 2500+ runs 1833MHz and the 2400+ runs 2000MHz. If you put them both on the same fsb speeds still running at stock cpu speeds the 2400+ would be faster. The extra L2 cache of the Barton is only worth about 3-4% performance increase across a wide range of benchmarks. The biggest difference you will see by reading the benchmarks out there is the Athlon 266MHz fsb speed compared to a Barton at 333 or 400MHz, only a few accurate benchmarks show them competing on the same fsb speed.

I have both and don't have to depend on any benchmarks. I have done extensive testing with applebred, t-bred and Barton. Believe what you want, but it seems like most people here also falsely believe that the Barton is much faster than an Athlon. A good Athlon running 2200-2300MHz on a 200MHz+ fsb will run very close to a Barton at the same frequency, only the mobile Bartons that are consistently hitting 2400MHz+ have a noticable advantage.

 

[selfbuilt]

Originally posted by: Tomer
Although you may consider it superior, and it may well be in certain cases, in mine case it simply is not.

I just completed a DVD encoding session with both processors. Basically the CPU goes to 100% and stays there. The 2400+ took 4 hours 28 minutes to complete. The 2500+ doing the exact same task took 4 hours 52 minutes.

I will run the UT2003 benchmark and see what happens.

You should definitely see an improvement with UT2003 due to the higher FSB ... but I'm not surprised to hear its slower at stock speeds for DVD encoding.

I have several systems similar to yours, and I know my son's Barton 2500+ (Asus A7V8X, 512MB PC2700@2-3-2-6) at stock (166x11) is only about 13-15% faster at MPEG encoding than my daughter's Palomino 2000+ (133x11.5, ECS K7S5A, 384MB PC2100@2.5-3-3-7) ... very similar arrangement to yours. Of course, it blows the K7S5A's socks off for gaming ... and it's an unlocked core, so he runs it at 2800+ speeds (166x12.5).

In comparison, my mobile 2500+ run at ~2.5 GHz (198x12.5, Asus A7N8X, 1GB PC3200@2-3-2-6), which SiSandra estimates as a "3600+" is still only about an 35-40% faster than the K7S5A system at MPEG encoding! Of course, at everything else it blows it away! (BTW, those % are "real-world" - they're based on actual timings from a selection of home made DVs).

It just goes to show you that MPEG/DVD encoding is a very brute force CPU function ... you don't really get to see the difference of the extra processor cache of the Barton (or the greater amount of higher speed, lower latency system DDR RAM for that matter!). But you will definitely see the FSB benefit in games, and the Bartons are superb overclockers.

 

[howdyduty]

The 2500+ has some future with many OC'ing to 2.2ghz (11x200). Works great for me

 

[MDE]

If you want your video encoding to go faster why don't you have a P4?