![]() |
![]() |
#23 |
Mar 2003
New Zealand
115710 Posts |
![]()
Dual channel memory is not worth paying much extra for if you are just running Prime95, it will give less than 5% improvement. But then i865 boards are not much more than i848 ones now anyway.
Onboard video is handy, but not really worth paying extra for. If you are actually using the video then a real AGP card is much better, and if you are running a dedicated number cruncher and just want the video for setting the BIOS then a second hand 2Mb PCI card is easy to find for a dollar. |
![]() |
![]() |
![]() |
#24 |
Oct 2004
232 Posts |
![]()
Well if you want 3D games performance forget onboard video chipsets. However my cluster does not need this at all and the power of the cards would be wasted.
You say stick in an agp or cheap pci card into your motherboard and avoid the need to buy a mobo with integrated graphics. Well secondhand, yes they are cheap, but I find the price difference to get a mobo with integrated graphics as opposed to without is very small maybe £5-10 extra. Well the reason I avoid this is that without a 90 degree riser card, you increase the size of your system components to about 4 inches vertically because the card is at right angles to the mobo. Also an integrated chipset is likely to use less power than say a pci card which requires additional logic circuitry to interface with the bus etc. Dual channel will increase prime95 performance, even if not dramatically. However, on different cluster applications, the dual channel might help performance by a greater (or smaller) percentage depending on how much the application uses memory outside the L2 cache. At least worth having the option by getting a board that supports it even if you only run it with one stick of memory in single channel configuration for prime at the moment. |
![]() |
![]() |
![]() |
#25 |
Jan 2003
North Carolina
111101102 Posts |
![]()
Two more details.
On board video that uses RAM slows prime95. If on baord video is a must, it would be better for it have its own onboard video memory separate from RAM. When looking at mobo with memory slots, check for dual channel operation limitations. My mobo runs DDR333 with all three memory slots in use. But DDR400 is limited to the first and third slot only. Also, last time I ordered hardware from newegg.com, I was eligable for OEM priced norton ghost, system works, etc. You might want to check that out too. |
![]() |
![]() |
![]() |
#26 |
"Mike"
Aug 2002
24×499 Posts |
![]()
I have a Dell Dimension 3000 here that gives the same iteration times with the onboard graphics as it does with a PCI video card... It has a 2.8GHz Prescott without HT and it has 2x256 DDR333 in dual channel mode... The non-HT 2.8GHz Prescott has a 533MHz FSB...
|
![]() |
![]() |
![]() |
#27 |
Oct 2004
232 Posts |
![]()
Nomadicus, thanks for pointing out that using some onboard video solutions could slow RAM access and therefore prime95.
I had in fact considered that this effect might come into play. You are right that this could happen because of the video controller and the processor both wanting access to the memory they share. (because the video doesn't have its own dedicated memory). Well I can suggest running your gui desktop in a low resolution and low colour depth which will reduce the amount of memory accessing caused by screen refresh and updating the screen buffer when something on screen changes. Secondly I think some interleaving effect may happen minimising the effect (if the memory controller and/or processor is clever?) As prime95 keeps memory read/write within the Level2 cache as much as possible, the main memory should not be contended with the video accesses most of the time. Both my machines with Intel 865G chipset do share video/main memory this way. I would be interested to know of other chipsets and/or motherboards where the graphics has dedicated video ram and thus avoids this problem. My prime95 benchmarks with 865G integrated graphics seem to be comparable with other people running on dedicated card with own memory. I therefore do not think this effect is major. If you have any benchmark results indicating a wide variation between onboard and plug-in graphics please post them or summarise your experiences. For applications other than prime95 programs may well use memory beyond the L2 cache, so the difference would be more noticeable. |
![]() |
![]() |
![]() |
#28 |
Oct 2004
232 Posts |
![]()
Like xyzzy above who compared onboard versus dedicated graphics on his P4 above, I've done the same on my 3GHz P4 (using an Intel brand motherboard).
The prime95 processing doesn't seem to vary significantly when I remove the 9600XT card. I have not run such comparative benchmarks on applications other than prime95 though. |
![]() |
![]() |
![]() |
#29 |
Jan 2003
North Carolina
111101102 Posts |
![]()
I'm not aware of an AMD based mobo that has onboard video with video RAM.
My one AMD based mobo with onboard video, prime95 slows down noticably. I was surprised that changing resolution didn't make much difference. "My prime95 benchmarks with 865G integrated graphics seem to be comparable with other people running on dedicated card with own memory. I therefore do not think this effect is major." "The prime95 processing doesn't seem to vary significantly when I remove the 9600XT card." Cool. |
![]() |
![]() |
![]() |
#30 | |
"Mike"
Aug 2002
798410 Posts |
![]() Quote:
http://www.ocworkbench.com/2004/giga.../K8S760M-1.htm |
|
![]() |
![]() |
![]() |
Thread Tools | |
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
Best Prime95 bang for my bucks? | joblack | Hardware | 21 | 2012-04-01 00:40 |
1 buck an hour | crash893 | Hardware | 6 | 2009-06-18 01:45 |
best bang for the buck? | crash893 | Software | 6 | 2007-06-06 07:10 |
Bang per buck | Brucifer | Prime Sierpinski Project | 7 | 2006-09-29 14:55 |
Best bang for the buck? Athlon XP vs. P4 | dans | Hardware | 9 | 2003-09-10 11:18 |