PDA

View Full Version : SiSoft memspeeds on DDR400


Nutty
03-06-04, 08:09 PM
Can anyone with an nforce2 run the memtest on Sandra with a 200FSB and not in dual channel mode?

I only seem to get 2800/2600 Int/Float for my ram at 2-2-2-6. cpu 2.2Ghz.. Seems a bit low to me..

Thanks.

Smokey
03-06-04, 11:15 PM
I just did a quick run with the latest Sandra

FSB @ 200 2-3-3-8 but dual mode, and I got 2918/2745, cpu at 2.2

Ninja Prime
03-07-04, 12:46 AM
I'm getting 3004/3003 Int/Float @ 200mhz with 2-3-3-6 timings, processor is an Athlon64 @ 2Ghz.

Edit: Bah, just noticed "with an Nforce2." :rolleyes: ;) My brother has one, I'll see if I can get him to run it if hes around.

FastM
03-07-04, 10:59 AM
When i had my Epox 8rda+ i was running 2x256's (dual channel) @ 6-2-2-2 but only at 185FSB. I was getting the same 2800/2600 your getting now.

Now i get 3100/3100 :D

Nutty
03-07-04, 01:49 PM
Hmm smokey's result is close slightly higher I'll put to dual channel.. Just wondering if theres something I've got enabled, or haven't thats slowing me down a bit..

FastM
03-07-04, 02:00 PM
Originally posted by Nutty
Hmm smokey's result is close slightly higher I'll put to dual channel.. Just wondering if theres something I've got enabled, or haven't thats slowing me down a bit..

USB ports? a backround?. damn memory resident things.

Dazz
03-07-04, 02:49 PM
I found 2-2-2-6 to be slower then 2-2-2-7

Nutty
03-07-04, 04:19 PM
Yeah for dual channel, the optimal is 2-2-2-11 apparently.. Baron tried it, said he got higher Sandra scores, but things felt chuggy.

I seem to get IE crashes and more distortion on Energy Core explosions in UT2K4 at 2-2-2-6, so I'm currently 2-2-3-6, which is what Corsair state it should be.

I just assume PC3200 ram should give me readings closer to 3200, rather than 2600..

Dazz
03-07-04, 06:00 PM
NO that is not the case at all. it's top torectical bandwith, it will NEVER be 100% effecient, The Athlon XP is around 92%, Athlon64 95% and the Pentium4 62%. All these can change, from core stepping's and board chipsets.

AncientGamer
03-13-04, 06:09 AM
Originally posted by Dazz
NO that is not the case at all. it's top torectical bandwith, it will NEVER be 100% effecient, The Athlon XP is around 92%, Athlon64 95% and the Pentium4 62%. All these can change, from core stepping's and board chipsets.

Why do you say the P4 is only 62% effecient? I have a p4 with single mode PC3200 ram and I get 2918/2922... (I really am asking, I'm very new to DDR ram, I just got my P4 cpu.)

Dazz
03-13-04, 08:22 AM
Yeah thats due to the P4's bandwith being twice as much as PC3200 but if it runs in dual channel you will find it running well below 6.4GB/sec that it's rated for. The resaon for this is high latency for the long pipeline and also the efficenty of Quad Data Rate bus. Even the bandwith of a 533Mhz processor is more then that of a single channel 400MHz so it can compensate and get close to the bandwith of Single channel Double Data Rate memory.

Woodelf
03-13-04, 01:19 PM
I've alway heard that athlons respond better to latency, while P4's to bandwith. Which is why I go more by the mulimedia test in sandra when comparing to p4's. While I use the mem test just to see where I'm at compared to my previous mem test (or other mem in my motherboard/mem class). :)

Dazz
03-13-04, 04:29 PM
The mulimedia test will always run getter un a dual CPU setup/HT as it runs multiple applications at once.

Woodelf
03-13-04, 04:42 PM
Originally posted by Dazz
The mulimedia test will always run getter un a dual CPU setup/HT as it runs multiple applications at once.
As it should since it's a general benchmark, not a gaming benchmark.

AncientGamer
03-13-04, 06:58 PM
Dazz, I have an 800mhz bus P4, what will my effeciency for Dual channel DDR be?

p.s. Thanks for all the responses in the various threads ;)

NickSpolec
03-13-04, 10:31 PM
When I get my Albatron NForce2 on Monday I will post some benchmarks.

Dazz
03-14-04, 05:04 AM
It depends but the average i have seen is around @ 800MHz DC is 5.1GB/sec bandwith for this benchmark still quite a distance from it's 6.4GB/sec.

GamblerFEXonlin
03-15-04, 06:52 AM
Ah another chance to promote the Socket 939 athlon 64!



http://jooh.no/web/A64FX_memory_band.png
The P4 3.2c is a 800Mhz FSB


http://jooh.no/web/A64FX_dualMemCtrl_latency.png
the latency is very good


http://jooh.no/web/opteron_dualMemCtrl.png
so maybe real-high polycount/geometry detail games run better?

SnakeEyes
03-15-04, 10:19 AM
The results in the GunMetal benchmark graph are whacked! (no, not using a negative term here- just amazes me the difference in performance for that processor / motherboard / memory and the others tested). :eek:

NickSpolec
03-15-04, 11:34 PM
My SiSandra (2002, SP1... yeah, I need to update) memory Benchmarks, with a Mobile Barton @ 2500mhz, 200FSB (whacks out at anything higher.. maybe I need to do the wire trick :P), single channel, CAS2, 3,3,5...

INT: 3018
FLOAT: 2782

Dazz
03-16-04, 02:31 AM
I know a friend that knows a friend that has already tested the Athlon 64 939skt processor and he says that on average it is 5% faster then the Athlon 64 754skt. But GunMetal is THE only game i have seen that shows the most diffrence from single and dual channel.

SnakeEyes
03-16-04, 10:10 AM
How does Quake3 compare between single and dual channel? I was under the impression that it also showed quite a bit of variation based on memory bandwidth.

Dazz
03-16-04, 04:39 PM
Hardly any diffrence.

AncientGamer
03-17-04, 02:43 AM
I have two Kingston Value Ram DDR400s in dual channel mode (1 gig). I overclocked from 2.6c to 3.0c and my Sandra memory score is 5000ish, is that good or really low?