Duallie Tuallie/Double T-Birdies bench it out

[noko]

Interesting comparisons, any comments:

<A HREF="http://www.2cpu.com/Hardware/Dell_Duallie_Tuallie/dell_2550_1.html" target="_new">http://www.2cpu.com/Hardware/Dell_Duallie_Tuallie/dell_2550_1.html</A>

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed<P ID="edit"><FONT SIZE=-1><EM>Edited by noko on 07/29/01 09:22 PM.</EM></FONT></P>

 

[Crashman]

Too bad it's a scam-there were no Tualatins in the benchmakrs, only some old Copermines and a couple Athlons.

Video killed my Radio Card!

 

[noko]

What are you saying? SiSoft indicates P3s with 512kb of cache. Sounds like Tualatins to me. It is from a Dell Server advertise as such also.

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed<P ID="edit"><FONT SIZE=-1><EM>Edited by noko on 07/30/01 01:26 AM.</EM></FONT></P>

 

[Crashman]

Yes, on that page. Then you go to the benchmarks and he's comparing PIII 1000 and XEON 667 systems to Athlon4 1333 and P4 1600 systems. So where's the benchmarks on the Tualatin? Absent! Is it really fair to compare a P4 1600 and Athlon4 1333 to a PIII 1000 and Xeon 667? You be the judge.

Video killed my Radio Card!

 

[peteb]

Crash, the top lines indicate currently tested cpus vs the comparative systems for reference. The top lines are the actual (simulated) dual Tully performance vs expected (simulated) Xeon, Athlon, PIII performance.

Seems okay to me if you put stock in Sandra benchmarks.

-* This Space For Rent *-
email for application details

 

[girish]

the author says the server system are Tualatin capable, and they are offered as options:

* Up to two Intel® Pentium® III processors from 933MHz <font color=red>to 1.13GHz</font color=red> and up to 4GB ECC SDRAM for expandable performance.
* 256KB Level 2 Cache (933MHz & 1GHz), <font color=red>512KB Level 2 Cache <b>(1.13GHz)</b></font color=red>

crash is right, the benchmarks does not have any Tualatins.
there is no 1.13MHz Pentium-III with 512KB L2 cache in the tests.

and what is this actual/simulated/expected things? how can you estimate Tualatin performance by simulating it with a Coppermine???? and what does this " actual(simulated) " mean, are you saying actual means the same as simulated???

so what do we do with these benchmark results? keep them with us to compare when somebody actually does some Tualatin tests with the same Dell server?

girish

<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>

 

[peteb]

SiSoft Picture #1:

Processors - 2 x Intel Pentium 3 Family.... 1133Mhz@1.13Ghz
L2 Cache - 512kb ECC synchronous ATC

SiSoft Picture #2:

Processor (as above)
Model Information - P6T (Tualatin) Pentium III 1G +xxV
L2 On Board Cache - 512kb.....

SiSoft Picture #4&#5

Current CPU scores considerably higher than reference P3 1G setup.

and what is this actual/simulated/expected things? how can you estimate Tualatin performance by simulating it with a Coppermine???? and what does this " actual(simulated) " mean, are you saying actual means the same as simulated???

I put that guff in the post as many people do not regard SiSoft as an accurate real world test of system performance. Also the 1G/Xeon/Atlon comparison figures are purely speculative as they are programmed averages for comparison only.

In my opinion these are real Tualatin Sandra benchmarks. You may want to actually read the article and look at the pics before you decide it is rubbish...

-* This Space For Rent *-
email for application details

 

[kusek]

Apparent not everyone has used Sandra before.
The top line is always your processor and the next 4 lines are average benchmarks for what Sandra considers "comparable". The tester has no choice on what systems shows up on these 4 lines.
The Tualitans are the "test" system on the Results page. The 4 models it chose to compare to are
Dual P3 1GHZ,Dual P3 Xeon 667, Dual Athlon 4 1.33, and Dual P4 Xeon 1.6.
I don't see any "cheating" here unless we can't trust the SiSoft benchmark program.

 

[noko]

Now that we confirm it to be real Tuallies, why do some of you doubt SiSoft benchmarks? The benchmarks are programs that are purely timed on each cpu it is runned on. Then the values are compared to each other to show the relative differences. Now how accurate does that reflect real world programs. If the FPU test is 40% faster in SiSoft Saundra, would you see a 40% increase in 3dsMax? or a somewhat higher % increase since 3dsMax may also use interger math as well?

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed<P ID="edit"><FONT SIZE=-1><EM>Edited by noko on 07/30/01 10:51 AM.</EM></FONT></P>

 

[FatBurger]

The tester has no choice on what systems shows up on these 4 lines.

Actually, you do. There's a pull down box, and you can select any processors (or RAM, or hard drives, etc) you want to compare them to. Except I don't think they have a P75, for instance.

-----------------
Whoever thinks up a good sig for me gets a prize

 

[noko]

This is all mute now, AnAndTech has a review of the .13micron P3. It has very respectful performance and they just used the 256kb cache version and not the 512kb version. In fact in a number of tests it is faster then the T-Bird 1.2ghz processor while others it falls behind. Also interesting is that the Tuallie has data prefetch like the P4 and the Athlon4. Now I wonder how rcf84 512kb version P3 would perform?

<A HREF="http://www.anandtech.com/cpu/index.html" target="_new">http://www.anandtech.com/cpu/index.html</A>

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed

 

[rcf84]

Trust me. I get a 15% boost in Quake 3 :*)

Nice Intel and AMD users get a Cookie.... Yummy

 

[noko]

Yes with the data prefetch, the bigger cache of your version would make a bigger difference in real programs not necessarily pseudo benchmarks. For benchmarks like SiSoft you will not see this performance increase due to the program fitting inside of the Cache which makes the Benchmark faulty as far as I see it since it is not analyzing the the memory to cpu factors that will occur in most programs. This will go for the P4 when it gets the 512kb of cache as well.

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed

 

[noko]

Another thing is the faster you overclock the FSB the better the data prefetch will be for you setup. A double wammy.

It does look like the P3 Tually is clock per clock faster in interger math (business type applications) then a T-Bird.

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed<P ID="edit"><FONT SIZE=-1><EM>Edited by noko on 07/30/01 12:14 PM.</EM></FONT></P>

 

[Crashman]

OK, but these simulations compare a complete system to a base benchmark of another system. The only way you can do a true comparison is to use as many similar or eqaul parts as possible. We all know that Dell pushes stability over performance on their servers. I'd really like to see something like a TUSL2 coppermine vs. 512k Tualatin vs. A7V133/Athlon comparison, with all the cards, drives, and memory being equal.

Video killed my Radio Card!