Which of these motherboard combos is the fastest?

C5Dave

Junior Member
Dec 28, 2002
8
0
0
I'm looking to buy a "sweet spot" level system that is going to give me the most value/performance for my dollar today. I'm going to be doing about 70% gaming and the rest normal business application stuff.

I may try some light over-clocking but nothing extensive.

So here's my question... with all other components being equal....which one of these two CPU/mobo/memory platforms will give me the fastest overall performance?

1) AMD Athlon XP 2400+
ASUS A7N8X Deluxe
512kb Corsair XMS-3500 DDR

or

2) P4-2.4mhz 533fsb
Gigabyte GA-8IHXP
512Kb PC800 Rambus RDRAM

Thanks for your help.
 

mechBgon

Super Moderator<br>Elite Member
Oct 31, 1999
30,699
1
0
Hi C5Dave, welcome to the Forums. :D It's not good form to put duplicate posts in more than one section of the Forums, so you're aware...

As for your question, look here for some benchmarks of a similar Abit board with the CPU at 2.0GHz in both 133/133 memory/FSB and 166/166 memory/FSB conditions, versus a 2.53GHz P4 with DDR333.

AnandTech also did a comparison you'll find semi-relevant: P4's with PC1066 RDRAM versus AthlonXP's with nForce2/dualDDR The P4 is faster with PC1066 than with DDR, so there will be a 5%-10% performance hit moving down from these marks to P4 on a DDR platform in many cases.
 

C5Dave

Junior Member
Dec 28, 2002
8
0
0
Mech,

Thanks for the help and suggestions. These articles are just what I was looking for.

I wonder why the anandtech review did not include the TWO processors I was looking to compare (2400/P4-2.4)?? Oh well guess I can estimate besteen the two that they did compare.

From looking at the tests it appears that the AMD is a bit faster overall... so I think I'll go with the Athlon.

But I was wondering if you think it would be worth it to spend $120 more to go with the 2600+. If I did ...could I use the 333fsb version of the 2600+. I'm not sure if the ASUS board could make use of the 333fsb.

This is a great forum...

Thx
 

nick1985

Lifer
Dec 29, 2002
27,153
6
81
Originally posted by: C5Dave
Mech,

Thanks for the help and suggestions. These articles are just what I was looking for.

I wonder why the anandtech review did not include the TWO processors I was looking to compare (2400/P4-2.4)?? Oh well guess I can estimate besteen the two that they did compare.

From looking at the tests it appears that the AMD is a bit faster overall... so I think I'll go with the Athlon.

But I was wondering if you think it would be worth it to spend $120 more to go with the 2600+. If I did ...could I use the 333fsb version of the 2600+. I'm not sure if the ASUS board could make use of the 333fsb.

This is a great forum...

Thx




the asus a7n8x can handle the 333 FSB.
also, if you did get an athlon 2400, if you are not planning to overclock, pc2100 is just as fast as pc3500. you can save a whole lot of cash if you do that. but if you do plan on getting a 333, get pc2700 so it will match for FSB. if you plan on overclocking your cpu to hell, then go for higher speeds.


ps-its not 512KB, its 512 MB. lol, 512kb of ram aint very much :)

 

C5Dave

Junior Member
Dec 28, 2002
8
0
0
Nick,

Wow...I'm somewhat a newbie when it comes to understanding all the matching of memory speeds ect.
from the numbering scheme I just expected PC3500 to be the fastest and best.

Can you explain to me how and why PC2100 is just as fast as PC3500?

I'm going to do some limited overclocking.. but need a very releiables systems so I won't go too far.
And I do want to save $$$ and not buy memory OCing capabilty that I won't ever use. But I do want the best performance I can get out of a 2400+ with limited or no overclocking.

 

mechBgon

Super Moderator<br>Elite Member
Oct 31, 1999
30,699
1
0
PC2100 memory is designed to run at a 133MHz double-pumped bus (Double Data Rate 266MHz or DDR266). PC3500 can also run at the DDR266 speed, but is capable of being operated much faster (up to DDR433 guaranteed and probably beyond). At DDR266 speeds, if all the other timings are the same too, the PC3500 doesn't actually perform any differently than the PC2100.

Regarding those "other timings," take Column Address Strobe Latency as an example. CAS Latency is sometimes further abbreviated to CL, so you will see CL2 or CL2.5 commonly mentioned. 2 has a quicker response, just like a hard drive with a lower seek time is going to have faster average performance. Where a PC2100 module might be rated for an overall 2.5-3-3-6 set of timings, the PC3500 might be able to run (on an A7N8X, anyway) at 1-1-1-4 (and for the skeptics who've never seen a 1 option, yes, they're in there).

If you took the two modules and raised the speed to 166MHz (DDR333) then the PC2100 module might make it (I have a couple that do) but chances are you would have to add some slack to the timings, while the PC3500 isn't even breaking a sweat. Obviously, good-quality PC2700 is a good compromise if you don't anticipate pushing things a lot further than this. The PC3500 may still offer tighter timings that the PC2700 can't handle, but the law of diminishing returns is setting in.

Where PC3500 would make a lot of sense is if you want to boost the "pipe" between the CPU, northbridge and memory as much as possible by overclocking the CPU and memory buses synchronously to 200MHz (DDR400) or higher. Wait, you say, the CPU isn't going to let me go from 133MHz bus to over 200MHz bus, that would be overclocking it 50%+! :Q Yeah, but with nForce2 and a Thoroughbred, you can begin lowering the CPU's multiplier to keep it within its comfort zone. The article at Hexus illustrates very clearly that using this technique to boost a 2400+ from 133MHz to 166MHz resulted in about a 5-7% performance boost even without raising the total CPU MHz from its stock 2.0GHz. So with DDR433, or PC3500, you could conceivably keep the CPU within 80MHz of its intended speed while supercharging its memory bus to the 220MHz (DDR440) level, provided everything else in the system can hack it.

The question is whether today's nForce2 boards themselves can handle that kind of speed. Jury seems to be out on this question. Some people are going up to 210MHz and higher, I saw a 225 reported, but Evan found that the boards in his roundup were peaking around 195. I picked PC3200 CL2 from Corsair in the end.
 

nick1985

Lifer
Dec 29, 2002
27,153
6
81
i picked mushkin 2700 with 2:2:2 latency(which means that it will smoke corsair 3200) corsair doesnt offer 2 2 2. mushkin is my brand
 

mechBgon

Super Moderator<br>Elite Member
Oct 31, 1999
30,699
1
0
Huh? Corsair PC3200 can certainly run 2-2-2 at PC2700 speeds. :confused: In fact, if you want, I'll try 1-1-1-4 when I get to work tomorrow... :D Not knockin' your Mushkin, it's fine stuff too.
 

nick1985

Lifer
Dec 29, 2002
27,153
6
81
Originally posted by: mechBgon
Huh? Corsair PC3200 can certainly run 2-2-2 at PC2700 speeds. :confused: In fact, if you want, I'll try 1-1-1-4 when I get to work tomorrow... :D Not knockin' your Mushkin, it's fine stuff too.

yeah with the corsair you have to tweak it to get to 2 2 2. my mushkin comes stock. and ill bet a million dollars i can tweak this quality stuff($110 for 256) very high.
 

Originally posted by: nick1985
i picked mushkin 2700 with 2:2:2 latency(which means that it will smoke corsair 3200) corsair doesnt offer 2 2 2. mushkin is my brand

rolleye.gif
 

nick1985

Lifer
Dec 29, 2002
27,153
6
81
Originally posted by: Phocas
Originally posted by: nick1985
i picked mushkin 2700 with 2:2:2 latency(which means that it will smoke corsair 3200) corsair doesnt offer 2 2 2. mushkin is my brand

rolleye.gif

what? dont believe me or somthin?
 

TroutFish

Junior Member
Dec 12, 2002
19
0
0
Hi mechBgon,

I hear what you are saying about CAS latency. Has anyone ever tested memory transaction performance with the same memory set to different CAS latencies to see if CL really makes a difference? The site below at Mushkin seems to indicate that CL is not important for DDR type RAM. But then of course they market different levels of DDR, which seems to indicate the opposite point. Just wondering if there are hard numbers anywhere that show CAS latency makes a difference in DDR....

Mushkin FAQ on CAS latency

Thanks,
TF
 

TroutFish

Junior Member
Dec 12, 2002
19
0
0
c5dave,

I'm looking at a similar GA8-IHXP system. If you buy a 533 FSB motherboard, make sure you buy PC1066 RAM. knowledgemicro.com seems to have pretty good prices on that type of RAM. make sure you use the 184 pin version.

TF
 

CraigRT

Lifer
Jun 16, 2000
31,440
5
0
I'd go for the AMD rig too.
I went Intel recently after being die-hard AMD.
gonna go back to AMD again.
 

chocoruacal

Golden Member
Nov 12, 2002
1,197
0
0
Originally posted by: TroutFish
Hi mechBgon,

I hear what you are saying about CAS latency. Has anyone ever tested memory transaction performance with the same memory set to different CAS latencies to see if CL really makes a difference? The site below at Mushkin seems to indicate that CL is not important for DDR type RAM. But then of course they market different levels of DDR, which seems to indicate the opposite point. Just wondering if there are hard numbers anywhere that show CAS latency makes a difference in DDR....

Mushkin FAQ on CAS latency

Thanks,
TF

I've benchmarked my board, Abit AT7, using different CAS latencies (2 and 2.5) and the differences were negligable, as in, CAS 2 scores were slightly higher, but increases in real world performance (gaming/encoding) weren't noticable. I'm sure a lot of factors come into play though....RAM speed, board, etc.

If money is no concern, I would definitely go with the Intel/Rambus option, but with 1066 Rambus. NForce2 is still new, you might not get the stability you're looking for.
 

TroutFish

Junior Member
Dec 12, 2002
19
0
0
Thanks Chorucal, that's the first "real-world" comparison I've seen of CAS latency effects. I appreciate the response.

TF