• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Core or Memory OC better?

A

Alex41290

[H]ard|Gawd
Joined
Feb 20, 2004
Messages
1,570
Which nets a better performance gain? I was wondering because I know if you lower memory, you can raise core, and vice versa (to an extent). I would like to know which is better for an Aquamark benchmark, 3DMark 01 and 03 benchmark, and games such as UT2004, Doom 3, etc. If anyone knows, I think it would help a lot of people, not just me.

Thanks!

Alex
 
as far as i know, memory overclocking has far better results, taking the memory on my x800 from 555 to 580 got me over took me from 11700 to 12300 in 3dmark03
 
It actually depends on the card. In general the core is more important. I've done some testing to back that up for my personal setup.

3DMark03 - GT4 Mother Nature

Stock - 519.8/560.3: 76.7
Mem +20 - 519.8/580.5: 77.3
Core +20 - 540.0/560.3: 78.5
Both +20 - 540.0/582.4: 79.0

Mem +20 by itself gained 0.78% FPS, in the both -> core test it made a difference of 0.65%.
Core +20 by itself gained 2.35%.

In other testing, the ratio seems close to 3:1, clock-per-clock. Clearly core is more important for real performance, as expected.
 
thanks everbody...one more question:

I just installed an Arctic Silencer Revision 3 on my MSI 9800 Pro, and my OC is less than with stock!! why would this be? I tried reseating the thing, and tried new thermal paste, with no avail. Thanks in advance.


alex
 
I dunno if you use ATI tool, but try fixing your Fan speed at %100...
 
sorry if i sound stupid, but how do u do that in ATI tool? i looked around in it and i couldn't fine anything...am i not looking in the right place? i have the newest version on the website, and i couldn't find anything about the fan speed. I also checked on the switch on the arctic silencer. If i'm correct, having it "down" is the highest speed, correct?


thanks

alex
 
You need the 9800XT PCB (for a 9800 Pro) in order to take advantage of the temperature monitoring and fan control capabilities.

In ATI Tool (newest one, preferably), go to Settings and then select the appropriate option from the drop-down menu at the top. This works for the Silencer if you have the fan connected to the card, as its RPM is based off of input voltage.

Your OC should be equal or better, especially for the core. The only thing I can think of is that it may have a poor interface with the core.
 
There is no option in the settings menu that has anything to do with clock speed or anything of the sort, so i'm assuming that i don't have the XT PCB. Umm...what would cause a bad contact with the core? would that be my bad thermal paste job? or did i not screw it in well enough? thanks for the help.
 
It should show temperature monitoring and fan control options, if so. But you can still flash it.

Bad contact can be caused by the cooler being on too tight, or if it's too loose. It's pretty tough to really mess up thermal paste application.
 
haha, but ur forgetting that ur talking to me...this was my first "thermal adventure" so i probly did mess it up...but i followed AS's instructions...so, should i try loosing or tightening it? i might have it too tight. But i OC'ed my card to like 420/375...is that a good OC? i mean, stock its like 378/338, so i guess mine is pretty damn good...lol


thanks for all the help...


alex
 
Later-gen 9800 Pros seem hit-or-miss with the OC, at least in my experience (although the R360 ones seem decent). I got one of the very first ones (original) and it OC'd like a champ. It's running at 436/386 in my secondary gaming PC right now. But without a new cooler or ramsinks, it only did around 420/365. But I've seen some of the newer ones do worse. So 420/375 is at least average. :)

I've seen people put on Silencer too tight before, and it did increase temperatures...that might have been over at TR or another board, I forget already, lol.
 
okay, well, i'll try loosening it a tiny bit...i got like 45000 on aquamark, so i figure i'm doing well with it no matter what
and i can't hear anything in my case over my side fan, so i guess that's a good thing....


thanks for all the help,


alex
 
Yep, sounds good, bro. :)

I haven't worked with Silencers too much or had any problems during the few times I did, but I know I recently read of a similar problem on another forum where the guy solved it when he realized it was on too tight. If I come across it again, I'll let you know.
 
Since the stock goop has "bulk" its posible your shim is high and using AS3 the "gap" isnt filled as well and you end up at higher temps. If you used a diff cooler, the stock cooler has a dip to reach the core, aftermarket ones dont, so again you may be suffering from a gap.

The R360's Ive run across all peak'd in the 420'sMhz while my R350 does 485Mhz on air....after Vgpu mod. Vgpu mod'ing the R360's didnt help more than 5Mhz worth. Just a data point.
 
Well, for those that care, in ATi Tool, I changed the tRAS value from 10 to 11, and i got like 10MHz more out of memory, but then i went to 13.....:eek:!!! don't ever do that anyone, just to forewarn you....haha, i haven't gotten to loosening it yet, and how to you v-mod ur vid. card? it might be running too hot already to volt mod it, but if you could just let me know, that would be cool....


anyway, thanks again everyone



edit: and one more thing, are mem heatsinks worth it? i mean, if they are, by all means i'll probly get them...can u direct me to some good ones? and also, how much would they cost, like 5 bucks? :p, i don't have oodles of money right now, but it might be a good investment...i do have an old athlon xp heatsink, could i use that? thanks!
 
You can set the CL up one, too, but more than that causes problems. I messed a bit with the memory timings and found most do not produce desirable results.

You can vmod with a pencil or with some soldering. But you need better cooling if you plan on doing that, I would go as far as to say water or better...unless you have radical air cooling. The pencil mod is over here: http://www.vr-zone.com/?i=817&s=1

(works on the XT, too)

Ramsinks rarely give large gains, and really only benefit situations in which heat is the primary limiter. They can't hurt if you apply them correctly, and can be helpful with RAM that runs particularly hot or when voltmodding. I have ramsinks on the rear of my XT just for the heck of it, I chose copper OCZ ones. They can run $10-15 for 4 I believe but the copper ones need airflow to be efficient (get the aluminum, otherwise). You can also cut up old heatsinks and such, use arctic silver epoxy to apply the sinks (mixed with regular arctic silver if you want them to be removable down the road).
 
So I won't see a super-noticeable change? If so, I don't know if its worth the 10-15 bucks if there is no real performance gain...or anything greatly improving...is there anything else i can do to increase my overclock? and keep it from artifacting?


thanks everyone for all the great help!
 
You can always do a voltmod. :)

Most likely you won't see a huge gain, although ramsinks did help out my 9800 Pro. I felt it was worthwhile at the time because my card was $400 and I gained a healthy 6.85% on my memory overclock (or technically 3.42% overall)...which relates to about $13.70. Considering it was the best card at the time, it was a good deal. No voltmod.

Without a voltmod, that RAM tends to max out around 385-370, so I figure you have 10-15 MHz to gain at best with just ramsinks added. Probably not worth it unless you find something really cheap or make your own...or voltmod.
 
when u up the mem clock the core is stressed as well, the memory controler is on the gpu if I'm not mistaken (could be wrong, I'm no expert).

I'd say that the mem clock is best on the 9800 series because the gpus are pretty strong. But both help alot.

~Adam
 
Yes, the memory controller is a significant portion of the GPU die (<10%), although in some ways it might be better to think of it as four separate controllers (quad-channel 64-bit = 256-bit memory interface). In any case, the stress put on the GPU from memory overclocking would be more or less heat-limited (not voltage-limited), and certainly not massive in normal operating circumstances. So with relatively satisfactory cooling on an otherwise stock card, the difference would be minimal if even existent. Depending on other characteristics of the card, memory or core could be more vital in certain circumstances, but for pure performance the core is significantly more important. The distinction becomes more blurry in synthetic benchmarks and can vary from setup to setup.
 
wow Maxx, I got about 20 words in that paragraph :p...but cool, thank you for the help, i'll go with you in that ramsinks are not really worth it...i'll try maxing out the core, which will probably be like 430 if i go with like 360 memory, or maybe 425/365? Who knows, I'll see what I can do...


thanks again,


alex
 
Your best bet is to max out the core, then the mem, separately, to get an idea of what each can do alone. Then max out the core (to the value you found) and while it is there, find the max mem clock again. If it's significantly lower, then max out the mem and find the max core again. Most likely there won't be a great disparity, but if there is one then you have to make a call based on the performance difference (3DMark03 game tests should be sufficient).
 
Maxx said:
Your best bet is to max out the core, then the mem, separately, to get an idea of what each can do alone. Then max out the core (to the value you found) and while it is there, find the max mem clock again. If it's significantly lower, then max out the mem and find the max core again. Most likely there won't be a great disparity, but if there is one then you have to make a call based on the performance difference (3DMark03 game tests should be sufficient).
Click to expand...


I agree, sounds long but good
 
You must log in or register to reply here.
Back
Top