Dual Opteron vs single Pentium 4 EE

[resle]

Hi,
I am on the verge of putting up a high-end personal workstation, but having been away from
the CPU market for ... well, years, I am slightly puzzled by the possibilities.

Now, my first goal is to have the most "responsive" system possible, something that will never
freeze because of 100% processor occupation. Hyper Threading seems good at this, but far from
optimal, what I search for is definitely a dual processor system.

So my first choice was the Xeon family, but I immediately became aware that the low FSB is a horrible bottleneck, and in a lot of "desktop" applications like games, spreadsheets and more, 2 Xeons perform worse than a single P4 EE.

In the end, here are my questions:

1) What about a dual opteron system? How does that perform compared to a single P4 EE?

2) In case I want to mount 4gb memory, should I split it in 2 banks per processor?

3) What Windows family OS would perform better?

4) The 64 bit architecture of Opterons has a serious impact on 32 bit apps?

Thanks in advance for every suggestion

- andrea

 

[InlineFive]

1. I would think the Opteron would hold even or pull ahead of the P4EE.
2. I don't honestly know but I think it doesn't matter since the Opterons can access other processor's memory banks at 3.2GB/s.
3. As of right now you will have to stick with a 32-Bit OS, either WinXP Pro or Win2000 Pro.
4. No. The additional registers are left unused, however with the addition of a 64-Bit OS, 64-Bit Drivers and 64-Bit software then they will become used.

 

[Nickel020]

Well one Opteron 248 is pretty much the same as a FX-51 and so beats a P4EE 3.2 in most applications. I would imagine that adding a second Opteron 248 will be more than enough to easily beat a P4EE 3.4 in any application.
As for the memory, yes you should split it evenly, and I think it doesn't really make a noticeable difference performance-wise whether you use Win2000 or WinXP.
And the 64-bit part of the chip won't have any impact in 32-bit Windows since it's not in use. But I think you can expect a 15%+ performance increase with the upcoming WinXP 64-bit and a 64-bit application.

 

[resle]

By the way, what's the most advanced version of the Opteron?

 

[Nickel020]

The most advanced Opteron right now is the 248 (2,2 GHz) but yesterday AMD announced the 250 which will run at 2,4 GHz (same as the FX-51). The first digit in an Opterons name always stands for how many CPU you can use with it (e.g. 148 is only single CPU capable, 248 dual etc.). The price of the Opteron 250 will be about $850 each. It's launch day will be the 18th May.

 

[Duvie]

Well the bandwidth is true but remember most benchmarks have been ran on the older non i875 platforms for dualies....So they is a slight enhancment...

Another thing even with the bandwidth the xoens did quite well against the opteron and their eqauly priced opteron rival.

The reason I am looking at the xeons is to overclock them and get the fsb up so I can run dual chanel pc3200 and be at or near same levels as current p4c and p4ees....

One thing you don't seem to be figuring out is 2 things...

1) Most games and even quite a few applications are not multithreaded so NO DUH they are not gonna perform better then single processor setups. This is not a shock and a terrible comparison. If you run CAD apps like Autocad or ADT2004 you will see a huge increase that no single processor out today is gonna overcome.

2) HT and responsiveness in multitasking is rarely ever tested so you don't get a fair idea of how it will be...even more so with the xeons where each processor has its own HT portion so apps that are able will see 4 "virtual cpus".....


I priced an opteron setup and with comparable speed ones in the CAD apps I will be running the price for 2 chips a board with features I need and the ECC memory are comparable to 2 xeons of comparable speed/performance, a i875 Asus PC-DL mobo and non ECC memory (since the board doesn't require it)....

If it came down to no OCing I would likely have gone the OPteron and AMD route for A/ something new, B) hedging bet 64bit will eventually be released in windows and offer some sort of enhancement in the apps I run.

However I am cheap and want the bang for the buck so I am buying 2.66 (533fsb) xeons (M0 stepping chips) from a guy who will test them to run 3.32 ghz at default vcore....For the cost no opteron setup will beat it in about anything, and everything I have researched says opterons are a tough to null to overclock so that is out....

 

[resle]

Originally posted by: Duvie
1) Most games and even quite a few applications are not multithreaded so NO DUH they are not gonna perform better then single processor setups. This is not a shock and a terrible comparison. If you run CAD apps like Autocad or ADT2004 you will see a huge increase that no single processor out today is gonna overcome.

Of course, this is clear. I don't expect such a configuration to perform better than a single processor
one with games or common office applications. The thing I need to understand, is if a dual configuration
will peform worse in such cases.

 

[Duvie]

Originally posted by: resle

Originally posted by: Duvie
1) Most games and even quite a few applications are not multithreaded so NO DUH they are not gonna perform better then single processor setups. This is not a shock and a terrible comparison. If you run CAD apps like Autocad or ADT2004 you will see a huge increase that no single processor out today is gonna overcome.

Of course, this is clear. I don't expect such a configuration to perform better than a single processor
one with games or common office applications. The thing I need to understand, is if a dual configuration
will peform worse in such cases.

It shouldn't perform worse with equal speed (true mhz) opterons....However a Xeon you are right will be less then a p4c 800fsb model due to the fsb. It would be more directly comparable to the p4b line of 533fsb but has a slight advantage of lower speed chips with the HT.....

Like I said ocing get the xoens cause there are plenty of guys at 2cpu.com who are getting 2.4ghz M0 stepping xeons and getting them to 3.4-3.6ghz and so therefore you can get to that 800fsb but have dual chips that are not even available....Put that to use with an app that can handle multithreading and you got the best on the planet....

Ultimately don't even look at dual cpus if all yo run is games and want faster desktop feeling and responsiveness...I don't think you will be happy for the cost. If you have legitament multithreaded apps then get a dual cpu system...otherwise it is waste, IMO...

 

[joe2004]

Dual Opteron 248 will eat any Pentium alive.

 

[Duvie]

Originally posted by: joe2004
Dual Opteron 248 will eat any Pentium alive.

Go read the reviews the 3.2ghz 533fsb (the 800fsb are on the near horizon) does quite well in the normal areas like rendering and encoding that Intel normally does well in....

Eat alive is stuff ill-informed AMD fanboys spout...Not warranted and not true....

However with the xoens like the single p4 chips M0 stepping chips are allowing 2.4's, 2.66's. and 2.8's to egularly hit 3.4-3.6ghz like the p4c chips....That being said I can take down a 248 opteron in the apps I run while paying 1/3rd to 1/4 the cost...How do you argue with that???

 

[clarkey01]

no, i think if you check anandtech reviews of opteron, its "eats" alive at the xeon

somebody hit metallica " SAD BUT TRUE"

 

[Duvie]

Originally posted by: clarkey01
no, i think if you check anandtech reviews of opteron, its "eats" alive at the xeon

somebody hit metallica " SAD BUT TRUE"

Oh look another AMD fan on anandtech who doesn't knwo what he is talking about...SAD BUT TRUE there...

Fact...

I searced anandtech and the only opteron 248 review that popped up compared the 248 versus the 2.8ghz xeon in web server and IT stuff...Not what we are talking about here...

Here is from one of my threads back a few weeks...

Go to benchmark links....

fact there is they tested the 3.06ghz 512kb chips when now they have a 3.2ghz 1mb L2 cache version available.....

Look at them and count the number the 3.06 wins......Also look at the ones running actual program test versus synthetic benches...for example look at autocadd AUGI test...opteron wins there but that test all around things like memory and hDD susbystem cause it simultaes opening files...When you actually render in autocad the advantage goes to the xeons cause rendering is HT aware.

I think you (insert expletives) amd fanboys need to stop using blanket statement and words like "eat alive" cause it is clear like the guy above many of you are illiterate and beyond gaming probably know very little else about other applications....

I have tons more benchmarks I can pull as I researched this to death 2 weeks back....The fact is once again each has their strong points when we compare flagship to flagship chips.....Guys running the Xeon 2.4's at 3.6ghz with 800fsb are noticing large increases and the author of this thread is right the bandwidth is the limiting factor here and shows up in apps where we have known for a long time that bandwidth matters...the 800fsb Xeons will be arriving shortly and can help this and close the gaps.

I am sure when the 250's are reviewing in mass they will pull ahead even more in those areas they already lead and close gaps in others. But the cost will be the factor there....

 

[Duvie]

Originally posted by: clarkey01
no, i think if you check anandtech reviews of opteron, its "eats" alive at the xeon

somebody hit metallica " SAD BUT TRUE"

I think you must be confused by some other site...Fact is search of "opteron" in AT database of pcu reviews shows that other then the 4 part review 1 year ago in which he mainly tested an opteron as a SINGLE cpu and compared a 244 to a single 3.0c which defeated it in a lot of the non gaming apps...Other then the IT review and database shootout AT has yet to do a opteron to xeon review I see involing dual systems and other workstation apps.....

Other sites have reviewed this much more completely....

 

[clarkey01]

ok im a fan boy..lol..whats the matter ? your fave CPU maker Intel running into a few problems? with tejas etc Xeons are not as good as opterons, granted there will be area's where xeons beat the, due to there sheer clock speed, but its nothing to cry about. Only in singal configuration does the xeon edge over the opteron, but come 2 way, 4 way, above , game over Xeon.

There price/performance ratio is more then enough for me to embrace opteron over xeon's, even tho the performance.

Is it me or are all the Intel fany boys got there knickers in a wtist cos there "savior" of a CPU tejas has been cancled, FADE TO BLACK

lol, chill, its only a forum ;-)

 

[clarkey01]

least we agree on the Nehalem forum Duive ;-) , see Intel and AMD fan boys can get along

 

[PetNorth]

Originally posted by: DuvieHere is from one of my threads back a few weeks...

Go to benchmark links....

fact there is they tested the 3.06ghz 512kb chips when now they have a 3.2ghz 1mb L2 cache version available.....

Wrong. Xeon 3.2 has 512kb L2 and 1mb L3 (top model in Xeon DP). Exactly equal than Xeon 3.06 tested by ExtremeTech in that link (second model in top Xeon DP). So, the only different between them is 140mhz more in 3,2. Expect a mere 2% more perfomance with 3,2 than with that 3,06.

Originally posted by: DuvieLook at them and count the number the 3.06 wins......Also look at the ones running actual program test versus synthetic benches...for example look at autocadd AUGI test...opteron wins there but that test all around things like memory and hDD susbystem cause it simultaes opening files...When you actually render in autocad the advantage goes to the xeons cause rendering is HT aware.

Any link about your last statement?

 

[PetNorth]

I forgt it about bench AUGI Gauge with AutoCAD 2004 in ExtremeTech review:

http://www.extremetech.com/article2/0,1558,1403001,00.asp

We can read:

AUGI Gauge is a benchmark developed by the AutoCAD User Group International. AUGI Gauge tests the performance of either AutoCAD 2002 or AutoCAD 2004, Autodesk's highly popular, general purpose CAD application. AUGI Gauge actually is two benchmarks, a synthetic test that exercises the hardware and one that simulates real usage. We used the real-world test, since we wanted to see how AutoCAD would perform in a reasonably realistic scenario

 

[resle]

*sigh*

My worst fear was to start another flamewar.
All I wanted was to know if a dual opteron system, compared to a dual xeon system,
keeps better up with single processor systems in those tasks which are generally run
better on the latter.