Dimension 8300 CPU and memory performance

mariopezzo,

Does the system have onboard video or an expansion card?

what average are you talking about?   a lot of the machines that they pull those averages from have very little software on them.  Make sure that you unload all running programs, background programs, turn off virus scan, etc.  then re run benchmark

The system uses a Radeon 9800XT graphic card, an Audigy2 and a Raid controller.

Nothing more, nothing less...

 

 

I already unloaded almost everything through the Task Manager.

My impression is that the MB (Intel?) Dell uses is not a "top performer". Or maybe that BIOS and/or memory settings are a little bit "conservative" (but stable).

BTW, I love this PC and I am not blaming Dell. Simply I would like to make it even better, if possible

Good point.

But running DMark 2003 the CPU score is really much lower than the average of the compared Dell computers .

http://service.futuremark.com/compare?2k3=1729888

For example my project (before ATI overdrive, on the left) shows 160 points of difference! The MB is not the same, nor I could fine an identical one (OK, my PC comes from Dell Europe: is this the reason why?). Searchig out for other similar systems things don't change: almost all the other projects (excluding obviously the Dimension XPS with the Intel P4 Extreme Edition) score higher in the CPU test.

I repeated the benchmark many times, w/wo HT activated, no process running in background but the essentials: same result.

BTW: which MB could be mine? Different from the ones sold in the States or simply a different code?

I would question where Sandra got the comparative measures from.  If they are using samples provide by board makers, it's quite common for the maker to tweak the snot of the the board they send for evaluation to insure optimum scores. 

One other thing to check it HT- if this is enabled, it can add a bit of overhead that can result in slower measures on many benchmarks.

Dell uses Intel made motherboards using standard Intel chipset.  The quality of the stuff is on par with what Intel sends out in retail form, so I don't think it's sub-standard stuff.

They definitely tend to run a bit on the conservative side. The HT will tend to give you varied results as the exact same sequences tend not to repeat. For better pure results try running what benches you can in safe mode. I merely use benchmarks for determining the overall effect of any changes I may make to the system. The raid will most likely effect your results on anything but disk IO.

How's your memory configured?  According to this:

http://www.anandtech.com/memory/showdoc.html?i=1839

The configuration of memory can have an impact on memory bandwidth.  You're on an 8300, so you're not running integrated video, right?  Integrated video has a large negative impact on memory bandwidth.

FWIW, On my 4600 2.8ghz 800mhz FSB machine with 2 256mb single sided PC3200 chips and a cheap AGP card, I'm running scores of about 4300 on SiSoft Sandra's Memory Bandwith, 3800 on AIDA's memory read test, 1250 on Adia's memory write and CPU of 6691 and Memory of 7315 on PCMark 2002. 

I think you got the point.

I am running two double-sided 512 Mb PC 3200 modules. Windows XP Professional and the photo retouch programs I use would eventually benefit of 2 Gb of memory...

But, considering that all the memory modules must be matched, I am wondering if Dell could eventually accept to take back the existing two and sell me four, adding only the cost for the upgrade...I'll ask technical support!

Thank you!

Mario

BTW, when you talk about populating the Dimension 8300 four RAM slots with double-sided *matched* modules, do you mean that all of them must be matched or simply pair by pair?

 

---

@mariopezzo wrote:

BTW, when you talk about populating the Dimension 8300 four RAM slots with double-sided *matched* modules, do you mean that all of them must be matched or simply pair by pair?

---

I don't know - I've never gotten a firm answer on that.  I don't think they have to be exact, though - only of the same size, configuration and timings.

In the meanwhile I have checked the Intel 875P Chipset Memory Configuration Guide and what Thomas Pabst says: the better performance is obtained using 4 double sided RAM modules, totally matched. At the second place you find two DS modules, matched (my situation).

But if you mix modules, matched in pairs, you can expect a...decrease in performance, up to 25%!!!

Monday I'll ask the technical support if it is possible to get *four matched RAM modules*. Otherwise I think I'll stay with my config.

I called Dell Support Europe. I could not send an e-mail because the service is "not active yet..." (???)

They are completely unaware of the facts about memory performance reported on the Intel's 875P White Paper and on Thomas Pabst website. They offered me to buy four new memory sticks, *matched two by two* (-20 - 25% performance). And they did not believe in what I was trying to explain!

I think they should be at least aware of the facts discussed here. Consequently now I have three choiches left:

• I can stay with my two, DS matched 512 Mb modules. What a pity I cannot add 1 Gb more... 
• I must find some commercial brand offering *4 matched* DS, approved, 512Mb memory modules. And eventually sell the ones I have...
• I can add two more "matched modules" and benchmark to see how much of performance I lost, when I don't need to use all that RAM

I am very curious to see what Dell USA is going to do, especially now that they are offering a "Free Quadruple Memory Upgrade".

Will they offer matched modules? Paired two by two? I do understand that in Europe they do not sell the Dimension XPS, but still I cannot understand this policy: it is like selling a Ferrari and then providing only 195/65 R15 tires

Same "friendly complaining" about the on-board, disabled MB Raid controller: the Promise solution is nice, but still a bottleneck on the PCI bus (150 Vs 133 Mb/sec - even if the RAID 0 theoretical  max. transfer rate will probably never be reached...). And probably it also lowers the CPU performance (as someone said along this thread).

But this is another story...

Not surprising that front-line tech support wouldn't know about such an item.  Even a 25% decrease in memory throughput is something a user would be unlikely to notice outside of a benchmark, and these tech support guys are trained to  support users - not benchmarks.

Why not just order two more matching memory stick to what you have now?  In reading the article, I didn't get that all memory in the system had to come from exactly the same lot - only that mixing different memory configurations (ex: 2 512mb sticks with 2 256mb sticks, or two double sided with 2 single sided) would reduce performance.

There are some other threads on this forum where folks posted some benchmarks of various memory configurations.  If I remember correctly, there was a drop when mixing two different sizes of matched pairs, but otherwise there wasn't any significant difference.

You are right!

I thought they should have been from the same lot but it is not true...

Thank you