Ati Overclocking

I see a lot of bleeding at work, it doesn't acutally harm the screen or the board, that we know of. Usually they are known issues with the business models with integrated chipsets. If it causes hardware damage, it will result in a need to replace the motherboard, but since it's occuring with software settings, it's only temporary. I'm still playing with it a little, but like I said.. I need to wait for that cooler to push it any higher

I have the memory clocked as high as ATT will let me drag the slider - 375. Is there a way to let it clock higher? Or is that the absolute max I am going to be able to push my memory clock? It's holding my scores back and I know it can go higher -_-

Any ideas?

 

You could try a different overclocking utility. PowerStrip by EnTech Taiwan will do the trick; I've used it before.

 

I've got my max overclock without touching latencies at 378.00/371.25 GPU/MEM respectively.

It lets me play FEAR at 800x600 with almost everything set to high except those light density settings, those at high and medium.

Shader's at minimum or medium I forget and AF at 8x.

AF doesn't seem to affect my performance much at all, which is awesome~!

AA on the otherhand affects me big time.

I'm usually at 80%+ over what is it? 60fps or whatever? You know how FEAR benchmark splits the results into 3 sections? majority of my framerates are in the last section giving me very good stable playable framerates without ANY frames dropping below 20fps.

Wicked

 

i must cry at this point. i have an acer 1691 with x700 64mb (with 64bit bus). after lots of tweaking and overclocking...

3dmark05: 1500
3dmark03: 4100


it's the f***ng memory width... maybe i solder two chips more on motherboard and get 128bit bus with 128mb memory...

 

awwwww no....dj_pirtu....yes it is your memory...oh well it's still a good card...oh dude an overclocked x600 beats you....dayam

It's still okay, you have a good card.

don't dwell on stuff like that...enjoy your laptop.

Cheers,
Mike

 

This x600 isnt overclocked lightly. Lots and Lots of hard work and comming from someone who works with laptops every day for a living . On top of that I come from years of experience with desktops and a little dabling in overclocking. I'm not done playing with my clocks, and plan on going for that 2200 score that I need to equal out to an x700 stock . Keep trying, you can always go higher. Just remember the rule of thumb I've learned so far: Memory <= (less than or equal to) Core speed. And if you want, play with latencies, they will actually help. Play with them carefully and let us know what you get.

Edit: I have to give some credit to ikovac, if it weren't for him I probably wouldn't have thought to play with the latencies. ikovac, THANK YOU!

 

Thank you mate!

I'm so proud! I'm glad if I managed to clear things up and even make some of you guys actually try it and see the benefit of it. I also work with computers for a looong time, overclocked every single graphic card I've ever had, and never, never had any problems whatsoever afterwards. Notebooks (my third already) are a new playground. Small, hot and slow. I want to make them faster.
Again, I really enjoy posting to this thread. Feel free to share your experiences about latencies with us.

Cheers,

 

Here's a quick question...

In terms of pixel pipelines, I'm not totally familliar with what the Mobility series has. I know the MR 9800 had 12pipelines if I remember correctly. What does the Xxxx series have open? And is it possible to open any of them up on any of the cards?

One reason why I ask this question is after reading about it and seeing the laregely different scores at the same clock speeds between the x600 and the x700, if perhaps the x600 only has 4 or 8 pixel pipelines open while perhaps the x700 has 12 or 16.

By the way: I got my laptop cooler in today, and I have noticed some slightly different temperatures, but I think that due to the odd shape and poor grill locations of the zd8000 the air is having a hard time actually getting into the notebook. Dually, SpeedFan 4.27 isn't helping much with the temperature readings as I thought it would, Quanta never was much of one for putting very good temperature sensors around the MB. -_-

I think if I can find a way to control the in-machine fans, and keep them running at about half-speed consistantly, I could get that nice cold air into the notebook and keep it running nice and cool. Any suggestions, let me know!

 

You might want to try I9kfangui or i8kfangui...both are great for my I9300....hope this helps.

 

Actually not too well, IXkfangui is designed specifically for Dell Inspirion (sp?) series of noteboks . But, after looking at the software screenshots on the website, it looks to be a very intuitive and useful program

 

so far i know:

x700 has 8 pipelines
x600 has 4 pipelines
mr9700 has 4 pipelines (equal to x600)
mr9800 has 8 pipelines (equal to x700)

i had a heavily overclocked radeon 9700 nonpro on my desktop system and it got 3dmark05: 2400 and 3dmark03: 6000 with 4 pipelines and 256bit(!) memory bus. memory speed really makes sense.

 

Controlling fans. Not easy. On my last Asus I upgraded BIOS at some point and it did what are you seeking for. Quiet fan almost always. I didn't like it though.

Cheers,

 

By using 5.13 I managed to get 3074 with ATT setting latencies and at 415/435 MHz. Not too much, but some of the speed drops I noticed with 5.11 now dissapeared. So it is more "fluent" if you know what I mean. I'll try some more combinations.

I even installed CCC! Powerplay works!

So far so good.

Cheers,

 

Noise doesn't actually bother me, I used to have a desktop fan that sounded like a vaccuum cleaner 99% of the time, could hear it with my door closed 3 rooms away. So fan noise doesn't really bother me.

I'm looking for getting the air in. Right now the Quanta BIOS has practically ZERO options in it. You can turn on the 85 minute HDD test that does nothing useful. If I can convince all 3 fans to run all the time at half speed, the cold air from the LapCool4 would constantly get inside and keep everything cool, the way I like it .

I didn't have much of a chance to look around last night, as I didn't get home from work until late, but so far I can't find *any* program that lets you play with the fan speeds of HP notebooks. I may just be stuck like this until I mod the base assembly or the fans themselves .

 

i just uninstalled ccc it was awful screwd up my system completly half life played at 1 frame per seccond nearly all of the time and the graphics look realy bad (textures looked like they wernt at the right resolution) evan at max settings. now ccc is gone and everything looks brilliant again.

 

that's weird. CCC is just a wrapper for the ati control panel, you can adjust exact the same settings there.. it's just more user friendly and requires .net

 

yeah well everything is working well now the new drivers are actualy quite good.

 

what do you think i could take my Xpress 200M to? Its the version that has no dedicated memory, it borrows all its juice from the main board memory.

 

I am not even sure that you can overclock the Xpress 200M, especially the ones that use shared memory. Anyone confirm?

I have just downloaded and installed the Mobility 5.13's. They are a worthy upgrade to the 5.11's. I noticed that Quake 4 was a little faster. I will try some more titles later.

 

Can you access core and mem clock for 200M? If you can, I'd try it. I think with this integrated systems when you overclock FSB, it should show some speed gain in total. If I understand it right, then FSB +5% will drive CPU 5%, MEM 5%, and integrated GPU will access MEM at higher rate. That in total maybe gives more than 5% gain? It is just my thinking.

Cheers, and Merry Xmas,

 

Hi. I've been reading this board for a while now and thanks to the all great tips I got from here I'm now a happy owner of an Aspire 5024WLMi.

But anyways, to my problem. My x700 was clocked at 358/330 with Bios 1.13 and 358/300 with Bios 1.18. Now, I've somehow managed to get it to keep the 358/330 clocks (I have no clue how it has managed to do that) but I want to get it keep 358/345 (may change in the future) with Powerplay on. The problem I have now is that I use ATITool to enable the overclock but after I've enabled those speeds and disconnect the laptop from the AC, the computer crashes.

Any solutions for this problem?

 

I'm no expert on powerplay or how exactly it functions, but it sounds like powerplay could be kicking in when the AC is disconnected and it may be expecting certain settings, and finding others, thus causing the crash. Although, a little bit more descriptive version of your crash may help us find the problem and solution.

 

Well, I don't really know how I could explain it in a more descriptive way. You are probably right though about the reason why it crashes. Here's what happens in a simplistic way:
Windows starts -> Run ATITool to enable the overclock -> Disconnect AC -> Instant crash.

EDIT: This time around I got artifacts instead and after a moment VPU kicked in and lowered the clocks to normal speeds.

 

i wouldnt have my card oc and on the battery anyway dont want batery life to be any shorter than it allready is.

 

The reason I want this overclock to work with Powerplay on is that it would be overclocked only when the laptop is connected to the AC.

 

Hi Caze,

I see two possible solutions:
1.
Overclocking and powerplay don't like each other. I tried it on few occassions. I use Ati tray tools to OC, and I set extremely high clocks by setting latencies and overclocking. Powerplay in 5.13 is quite benevolent to the OC, but if I even start ATT and afterwards CCC it just freezes. So despite the fact that Powerplay is much more sophisticated way of managing power, I'd try with turning off Powerplay, and then underclock card when it is on battery. It works even at 100/100! I had it for months! Maybe that would satisfy your needs. As you don't need latencies, try with Notebook Hardware Control and there you can overclock and UNDERCLOCK just by ticking one box!

2.
In my opinion it is not Powerplay alone reason for your problem. Ati tool is that problem. If you know how to use Rabit, then set new frequencies into the BIOS, and then use powerplay, without any overclocking tool. I think there should be no problems. On AC it would work at your desired speeds, on battery Powerplay would kick in and do its job.

I hope it helps. I've never had this problem, but after few posts from you I realized that is what I would do.

Cheers,

 

Hi ikovac,

You understood my problem 100% correctly and I thank you for your suggestions. But I still got some questions for you.

1. Using this Notebook Hardware Control to overclock and underclock the GPU, would it be the same as using ATT PWS Overclocking option? And I guess using Powerplay is out of the question with this method?

2. This would be the most ideal solution for me but I've got to ask that how safe is this? I'm guessing that overwriting the BIOS with your own clock speeds could be a little risky if the card decides it doesn't want to work on those frequencies. On the other hand, if you have made a backup of the original BIOS and just start the laptop with AC disconnected it would be using the frequencies given by Powerplay and you should be able to overwrite the BIOS again with the original... Then there is always the question about warranty.

EDIT: I took a look at RaBiT 1.7 (said it was 1.6 when I downloaded it ) and so far I don't have a clue how to use it. Everything was just grayed when I started it.

EDIT2: Ok, I think I'm starting to know how to use it. This is what I think I need to do:
1. Dump the BIOS which I'm currently using.
2. Open it with RaBiT, edit it to my liking and save it.
3. Somehow upload the new BIOS.

I tried dumping the BIOS with ATT but RaBiT just complains that "No BIOS Signature" when I try to open it. Same happens with the file I got from ATITool.

EDIT3: Seems this sucker is harder to dump than what I thought. I've also tried with atiflash but it doesn't even seem to find the adapter(!). Any ideas for this problem as well?

 

Hi Caze,

I would go for the first solution then. The reason for mentioning NHC is that it actually works with Powerplay enabled and overclocks and underclocks on AC without any problems (or I've never seen any). Maybe it is using some less aggressive mode or something, and ATi driver doesn't seem to be affected. On battery it simply says that Powerplay took control and that is it. Check those pictures:
On AC and overclocked
On AC and underclocked
On battery and Powerplay kicks in

Notice that I cannot do that with ATT started.

Cheers,

 

Crap crap crap and crap again. This tools would have been perfect for me if it hadn't been for a couple of reasons.

1. It uses .Net Framework, which is known to be a real resource hog. NHC uses 50mb(!) of RAM for me at startup.

2. It doesn't support changing the volting for AMD processors. (This would just have been a bonus.)

3. It does what you said it would do with the GPU but the amount of resources it uses to do that one thing, doesn't really make it worth it.

 

Hi Caze,

Sorry to hear that. I don't have this issue. Two cli.exe + nhc.exe on my system use around 10-15MB RAM together. nhc alone takes around 2-6MB! If I use them actively then temporarily CCC+NHC would take around 30. In few seconds they are back to 15 or less.



Look for cli.exe (that is CCC) and nhc is already highlighted. cli.exe after some time both have around 2.7 MB.

Maybe something went wrong during installation?

I really wouldn't know why you get so big allocation. I use NHC for a long time (It was CHC before). On my old Asus I could use EVERY SINGLE FEATURE in it. Even controlling fans! Now I cannot do fans anymore on my new notebook, but still I think it is one of the best apps available for Centrinos (and other notebooks). .NET framework is really pain in the ...., but I don't mind. .NET frmwrk 2.0 has 23MB installation file (I don't know how much it takes on the disk), and as I can see few programs need it, maybe even more in the future (unfortunately). I think it is a small price to pay to enjoy (freeware) NHC. Just my opinion.

Cheers,

 

It doesn't use more than that for me either after startup but during startup it uses 50mb, which usually means longer startups. This I can tell from that I tested to boot with and without it starting with Windows.

I decided to test the 1.8 version and while it isn't as compatible as the 1.9, it actually consumes around 30mb less RAM at startup so I'm going to stick with it for now. So, thanks again for your help ikovac.

 

Guys, you can check my article about latencies on http://www.thegamebooks.com/latenci...t55.html?sid=d3fd401654a87d900e4a3df0dbb2713b
It might be quite interesting!



Some new info to read and tested TRFC optimization. Above numbers are BEFORE TRFC setting to 16. Now I get 3098 in 3dmark05, and 6653 in 3dmark03. Pretty good for an old DDR?

Cheers,

 

Excellent article Ivan. Thanks for posting.

I was playing Quake 4 yesterday at 400:400, and in one level I started getting artifacts. Turned it back down to slightly above default, and was OK.

What I really want to do is figure out the correct ratio of core:mem that will give me the best performance. I'm going to keep the default ratio of 351:331 and bump up the speeds, keeping the ratios the same...I'll let you know how I make out.

 

Hi Chaz,
Thanks, actually not many people are into this matter (scared perhaps, and I think this is very easy and has no side effects in terms of heat).
Your idea is good, but actually I think there must be some connection between GPU/MEM ratios and latencies. If you know that latencies in fact mean after how many clocks something will happen, then it is very important at what stable frequency your GPU and MEM work. Also ratio between them will be influenced by latencies.
Try this: leave default clocks, and set latencies so they give you some better results (100 marks more for example in 3dmark05). Then overclock. And check for artifacts. That is the procedure I use, until I get speeds so high when I have to trim latencies some more. 400/400 actually works well without LAT. I also saw that same clock speeds give quite good marks.

If you want to do it without touching default LAT I think top would be 400/400. Above that too many mem refreshes per second occur and I get artifacts if I don't set it to 255 for example. You said 400/400 was good and then started to show some artifacts. Try pulling Mem refresh rate to the end (higher number). See what happens. I think you'll see some speed gain and stability.

Happy overclocking,

 

Hey Ivan - did some testing.

I worked with the memory refresh settings and the CAS Lantencies.

Here is what I found.

The CAS latency has a noticable impact on performance. Default is 2, I set it to 1 without issues.
These tests below are with the memory refresh at the default of 39. I did something different this time as well - I kept the original ratios of 351:331. The clocks are now 400:376.

Bench! in ATi Tray:
CAS @ 2 = 3,387
CAS @ 1 = 3,412

There is a small performance gain, but it is still something.

Next, keeping the CAS @1, I experimented with the memory refresh rate. I do understand that this helps me get higher overlclocking speeds out of the memory, but I wanted to find out if they provided any performance increase, and here is what I found:
Clocks 400:370 (core:mem):

Mem Refresh Default = 39

Bench!

39 = 3,417
99 = 3,436
105 = 3,437
115 = 3,438
129 = 3,438
255 = 3,442

As you can see, the performance increase, although small, cuts off at 99.

Now for 3DMark05 tests:
Clocks 400:376 - CAS =1 - Mem Refresh = 99
SCORE = 3,183

I changed to clocks to 400:390, keeping the above settings.
SCORE = 3,219

Not a huge increase here, it appears that keeping the ratios does work well. I think that at 400:376, the processor and memory are doing balanced jobs. Going out of ratio to 390 does not offer much of an increase at all, perhaps because either the memory or core is working too fast to be efficent.

I'll test out some more later.

 

Hi Chaz,

Good going! I'm glad that you tried. CAS is one of the most important, together with mem refresh and TRFC.

I've never actually done this mem refresh testing. once I understood how it works, I've set it to 255 and that was it. In my opinion refresh happens way too often. As it should be dependant on actual time (in miliseconds), not frequency, then more you overclock, memory can make more cycles without a refresh! So at very high speeds 255 may be also too often!

I did a test where GPU is always at 390, and memory speed ranges between 310 and 430 in steps of 20MHz.
Check this:


ATT is more memory speed dependant than 3dmark05. I think that 3dmark05 is actually very hard on GPU. Games are not so GPU intensive and memory overclocking does show greater performance increase than overclocking GPU. So that is why you get so small differences when you keep GPU at one speed, and change memory.

A theory that manufacturer already set a good ratio between GPU and MEM sounds OK to me, but I'm no so optimistic. I see you have calculated mem speed at GPU 400. Good thinking. But as my card is 350/300 by default, that gives me for example 400/340, and difference between let's say 400/400 and 400/340 is 500+ ATT marks. That is considerable even for memory speed loving ATT bench. In my opinion the difference between GPU and MEM should be inside 20MHz. I still didn't find any proof that GPU should be faster than mem or vice versa. I'd personally go for faster memory than GPU, if I can set it. I still believe that games are using relatevely optimized GPU processing, and bottleneck is memory speed.

The test that would solve all questions should be: 3dmark score measured in all possible combinations of Mem and CPU and heat measured at every mark. Divide score by heat and you get a number. The higher, the better. Now that is a big work to do. Unfortunately 3dmark05 is so long to test. I think I'm not so patient. I might make a mini test. Like from 350/350 to 400/400 in 10MHz steps. That would give me 25 measurements that can be represented in a graph (3D). The peak should show the best combo.

BTW I was playing FEAR today without cooling pad and it froze at 415/435 after 5 minutes of playing. Good old cheap cooling pad really makes a difference! I guess 400/420 or lower will have to do in those (rare) occassions.

Cheers,

 

Thanks for the great info.

Last time I played around with TRFC, I kept getting restarts. I'll try going in steps and then testing stability.

3DMark05 does take a while...not as long as '03, but still long enough for me to say "OK, I'm only doing two or three tests." And, it is very hard on the GPU - uses every part to the maximum. Memory really makes a difference in 3DMark05, and although the core doesn't seem to matter to ATT that much, in 3DMark05, it does have more of an impact.

I'll try what you said and keep the core:mem within 20MHz...let you know how I make out.

Ivan - I think those clocks are a bit high, I wouldn't push the memory that far. 400MHz is as far as I would go with DDR..just my opinion.

 

OK, played with the TRFC a bit, and performance actually decreases steadily as I increase the value. Default is at 15 - I went all the way to 21 without problems, and did not go any further because I was not seeing any performance advantage. I attempted to set it to 13, but my screen messed up instantly and I had to reset it. So, I left it at 15, which is most stable.

The memory refresh set at 99 produces the best results for me, along with the CAS @ 1. The below benchmarks were done with those settings. I fooled around with the clocks a bit:

core:mem

375:375 = 3,410
385:375 = 3,414
400:375 = 3,423

OK, so playing with the core doesn't do much. I knew it probably wouldn't, but with the opimized timings, I decided to give it a whirl anyway. I stopped increasing that. Stepped up the memory a bit below:

400:385 = 3,515 - look at what a difference 10MHz makes! Over 100 points!
400:400 = 3,650 That's almost another 150 points with a mere 15MHz increase in the memory clock.
410:410 = 3,747

I didn't push it any further with that, but I kept seeing the performance climb.

I'll do a few 3DMark05 tests later.

 

435 too fast? But somehow I cannot just let go all that speed. Bad boy.

I'm joking. I guess you're right - 400 should be the top speed for DDR. I usually play games at lower speeds of course. The gain at 400/400 is big already, so one FPS doesn't make so big difference. The secret of successful overclocking is to know when to stop and enjoy the newly obtained speed, without thinking if everything will blow up in a cloud of smoke. . On the other hand you remember that I used to play at this high speeds when my heatsink wasn't placed well on the GPU and memory, at over a 100 degrees C! Now temps are around 80-90 degrees. So I think that maybe heat wasn't the main issue in this freezing today. TRFC was also in the game. It was at 16, and at 28 I never had this problem. I think I'll pull it to 21 and see what happens. Thanks for the info about speed gain and TRFC. The higher - more stable, lower - faster. I'll also try 99 memory refresh at 415/435. Just to see what happens.

Yes, your marks are great! 410/410 gives me around 3000 3dmark05.

I also have problems with 3dmark03. It works OK when I install it and then it starts to give me errors. Today it is the second time it is full of corrupted files and that it cannot be started. 3dmark05 and ALL OTHER things on my comp are fine! So I unninstalled it. Sensitive app. . Now I remembered that I used to have this issue years ago on my old Asus notebook. I guess that rules out hardware. Maybe I missed some patch?

 

Chaz, my man!

According to your suggestion I lowered refresh to 99 and put TRFC to 21, and guess what, I got 3108 3dm05 at 415/435! It is almost 40 points more than before! I should edit my article on www.thegamebooks.com to cover this. This is also my personal record (for now). Phew. from 2270 to 3108. Not bad.

Now that memory refresh sounds like a new area to explore.

Thanks man. As we can see - two heads are much better than one.

 

Also remember guys that all these POINTS in 3dmark are very minimal in real game tests.

A gain in 1fps in some tests gives many points. Do real tests in games like doom3 and such instead of 3dmark.

Sometimes running 3dmark over and over results in changes of up to 300points in my experience.

3Dmark synthetic tests are not a good way to test out performance gains, in my opinion.

Run 3dmark03 or 05 with whatever settings you have then run it again back to back without changing any settings. You will see that your scores keep changing. Sometimes not, but it's not consistent.

I didn't notice any changes when messing with latencies and I also couldn't set my cas to 1. Stinks. Oh well, that's maybe why I don't notice any noticeable changes.

Cheers,
Mike

 

Yes, two heads are definitely better than one. Awesome score, and you have a 128MB card. Exceptional!

Yeah, I think the memory refresh bus is a big part of performance. I tested everything from a value of 39 up to 135 in ~10MHz intervals, and 99 was the sweet spot. 75 produced similar results, but once you get to the higher clockspeeds, it limits performance. The clockspeed of the core:mem play a big role in the effectiveness of the memory refresh bus.

Mike - very true, the 3DMark scores aren't that close to real-life gaming, but if you can get the right settings to improve performance in sythetic benchmarks, you should see an increase in games as well, to some degree at least. I know I have, especially in the very demanding FEAR and Quake 4.

Looks like you have different memory than Ivan, and he has a similar card to you. What are your default clocks?

 

My default clocks are the same as his at 350/300, but I think he has different vram modules on his as he can actually overclock higher and set cas down to 1.

Oh well, I'm not upset...kinda haha

Whelps....happy new year~!
Mike

 

Hi Mike,

This is what my BIOS holds:

-- Additional hardware info --
SDRAM Mode Register: 0x1023
MCLK source select: MPLLCLK/2
SCLK source select: SPLLCLK/2
Chipset use A,B memory channels
SDRAM specific: 2**12 rows, 512 columns
SDRAM dynamic CKE is Enabled

Is your hardware info like this? Maybe that would help in determining what is the difference in memory?

Cheers,

 

What is your score in 3dMark05 with default clocks speeds with that card ChazMan? The reason why I'm asking is that I'm interested in what the performance gain from 128mb to 256mb vram is. Thanks to the tip about the NHC from ikovac, I'm now running happily with 358/345 which gave me a nice 200 points boost in 3dMark05 from ~2250 to ~2500. Oh, and Everest tells me the card has DDR memory while ATT tells me it has GDDR3(RaBiT seems to do it as well). Which one of them is correct? And which program do you guys use to change your latencies? ATITools doesn't seem to work for me and in ATT I can't find an option to change them.

 

I get 2,800 with the stock 351:331 clocks and the Mobility Catalyst 5.13 drivers.

There is definitely a difference between the 128MB X700 and the 256MB X700. The X700 doesn't have the fill rate to use all 256MB, but it can certainly use more than 128MB. The difference will be especially aparant in games such as FEAR, where it is recommended that you have a 256MB card in order to run at higher textures/settings in general. Having 256MB of dedicated video RAM helps 'future-proof' the notebook as well.

Sager and every other site that has the NP-5320 say that it is GDDR2, and that is correct - you can tell from the default clocks, and how it overclocks, temperatures, latencies, etc.

In ATI Tray, right click the system tray icon, go up to Hardware, Overclocking Settings, and then you can change the latencies in there.

 

Checked Everest again and noticed that it doesn't show the type of the memory at all(sorry for that), so I guess it's using GDDR3 then(I'm talking about my own machine now). My model of the 5024WLMi is LX.A4605.165 which does only seem to appear in Finland and Sweden(probably because of the localized keyboard).

Sorry to dissappoint you, but I have no setting to change the latencies in ATT. I don't know how many times I've checked the overclocking settings but it's not there.

 

Hi Caze,

Do you use ATi Tray tools that come with Omega drivers? There is no latencies tab in it. You can download one from here: http://downloads.guru3d.com/download.php?det=733

I'm glad it works now. I was sure it would work, and somehow felt dissapointed when you told me it wasn't. Anyway, glad to help.

Actually Everest under Display/GPU/Memory Bus properties/Bus type says what type of memory you have. I have DDR.

Cheers,

 

I'm using ATT that came with the Omega drivers, so that's why I didn't have the Latencies tab.
Everest doesn't show for me, even when I'm using the Ultimate Edition, I've checked several times.


UPDATE: After playing around with the latencies, I've managed to find the most optimal for my system at the current clock speeds. I got a whooping 600 points increase in ATT from ~2600 to ~3200 and in 3dMark05 a 200 points increase from ~2500 to ~2700. I also run CS:Source Stress Test and got a 6fps increase.
Here's my original and tweaked latencies:
Original


Tweaked


TR2R seemed to be the only one where I got more performance actually setting it higher. Those latencies seem to be too aggressive though to allow any further overclocking of the memory. I could only get it to 351mhz before crashing. Am I right in that tweaking the latencies doesn't really produce any noticable amount of more heat?

 

Yes, you should not get any more heat from tweaking the latencies. However, with the memory refresh bus, since it is being refreshed more often, will produce a bit more heat, although I don't think as much as from overclocking the core.

Nice job.