M17x Benchmarks thread !!!

i was refering to GPU scores not overall

 

Looks like you are using the same drivers also. Didn't you say you had something to do with the Dox drivers before? If so will there be a better one coming out anytime soon?

 

What I meant was I was the one recommending the driver from the beginning... no I don't have anything to do with making them

It's a good driver, especially for Crysis and 3dmark06, but I use 186.03. 185.20 is the king for vantage, but sucks you can't use it.

 

you can't compare an intel dual to an amd quad. a core 2 will still smash a quad amd. compare a core 2 duo and a core 2 quad and a core i7

 

Are they at stock clocks? I know 8500 can be pushed so hard and didnt Phenom have a problem with locked cores or something?

Also im pretty sure that AMD set the records before i7.

 

Don look at the scores....

AND

YOU are WAY out of the loop because I'm not comparing CPU brands or even CPU scores

I was comparing GPU scores

 

No, you're out of the loop bruh. Why are you comparing quad to dual, when the quad isn't Intel? Now if it were a Q9650 or whatever, you may have had a point.

The mere fact the E8400 is only 40 GPU points higher than the lowly AMD points to why CPU doesn't mean much of anything to get a high Vantage GPU score. Were talking fractional percentage increase.

 

.....It's pointless ....

Fine you guys can test it yourself ....

The Dual will help the GPU score higher
The Quad will help overall score because it rapes a dual...

Simple as that i don't need to turn this into any thing other then that.

Doesn't matter the brand

 

Moo, you're wrong

 

When i see otherwise i will delete all post and admit it when i see it

 

Is vantage multi-threaded?

If not, this whole argument is pointless......

 

Of course it is, even 3dmark06 is. And that's not limited to the CPU tests.

 

In Vantage's GPU scoring, it's not about cores, it's about clockspeed. An X9100 has a higher clock speed than a QX9300, which allows it to run tasks per core faster, and thus take away some bottlenecking from the GPU.

 

im not arguing what scores higher in the CPU part...

 

Thank you ...i should of been more specific ...Clock speed is where it is @

 

it makes more sense. but seeing you comparing a amd to intel isn't fair and threw me off.

 

I'm moving our "side topics" discussion here since we are talking about memory tweaking.

 

no problem. what are your stock timing.

 

This is with my 1066 mem set at 1200

 

Do you have it set to 1333 in the BIOS? Why is it running 1:2 and not 1:1 if the whole system's bus speed is OCed?

 

1200/4=300 Bus speed
1200/2=600 Mem speed
= 1:2 Bus:Mem

 

drop the trcd to 7 and trp to 7 and the tras leave it and run. then drop it to 20 then 18 then 16.

 

The memory is actually rated up to 5-5-5-15. I can set everything but the CL. so I can run 7-5-5-15. The question really is how do I get the CL to 5?

Getting the FSB: Mem to 1:1 have been impossible so far. Setting either the FSB too high or the mem too low and the system won't boot.

Edit: Bringing the memory down to 800Mhz I can get 6-5-5-15.

 

i'm running to work. i'll be back on what to do...

but try 5-4-4-4-15
5-4-4-4-14
5-4-4-4-12

it should pump your 3dmark06 score at least.

 

Right now I have 6-5-5-15. I'll try running 3dmark with these settings. CPU is @3.54Ghz so I don't know if it will be stable.

 

I gave up a bit. OC'ing the GPU has been less stable with the 1066 mem vs the 1333 mem. 3dmark06 and Vantage are crashing more often. I didn't think it would impact it that much.

I'm sure there are other tweaks I can do to make it more stable, just need to find them.

 

What kind of 1066 memory did you get Mandrake ? .

 

The Kingston HyperX 2x2.

 

Doesn't it have XMP? Well even if it does, you'll need a BIOS that would be able to adjust the XMP profiles. Not sure if the Alienware BIOS allows that level of customization.

 

Yes about XMP but no about the Bios customization. Someone in the other thread suggested Thaiphoon burner. Ever hear of it?

 

sorry I am getting confused with which ram you are using and the timings for each.

 

I'm testing out the Kingston HyperX. Rated up to 5-5-5-15 but can't get that without using a XMP profile (like ichime said), using software to change the timings or flashing the memory with new SPD. The Nvida tools will let me change it to X-5-5-15. The X varies depending on my memory frequency. If I lower my mem freq to 800Mhz I'll get a CL of 6, 1066Mhz then CL will be 7, 1200+ CL will be 8.

I've been investigating programs to reflash the memory but I haven't found anything newer than something that came out in 1997. The programs don't even work with Vista 64.

 

oh ok. i haven't done that before with my ram. can you do 8-7-7-15


what ever your cl(the first one is) drop everything down by one except the last one. you would have to play with the one.

 

I'm currently running the mem freq at 1066 so by default I get 7-7-7-20. Changing the other timings I have no issues doing. I can get to X-5-5-15 so 8-7-7-15 is no problem.

From my research I think I have two problems. First is I cannot get my FSB:Mem to a 1:1 ratio. Lowering the memory freq too low or the FSB too high and I can't boot the system. The second problem is getting my CL to 5. I have not found a way to do it yet although fixing might first problem may end up fixing my second issue.

 

yes i have heard that you can't change the first one unless you have that option in your bios or intel xtu.(which isn't released to the public)

i'm going to see what else you can do.

 

have you tried memset.

ok sorry is i keep on repeating my questions

if you set your ratio equal what do you get.

also if you don't what do you get

i understand that default is 7-7-7-20 i believe. which is correct if you don't have xmp profile to lower it to 5-5-5-15 i can't get nvidia ntune to work so i am using memset to test this out as i go along.

 

There was an issue using memset but I'm not remembering what it was. I'll take a look again.

Edit: It's not compatible with the chipset.

 

Moo, are you going to consolidate the 'benchmarks' to the first post? No one wants to troll through 130+ pages. TOTD is 'RAM Timings' in the 'M17x Benchmarks Thread !!!'.

@Mandrake: Is the Vantage score in your sig with PhysX on or off?

 

i put kceissm in the front i gotta get mandrakes to ....i have Clauss's as well

just gotta get them together

 

I'm trying to improve memory timings to see if it can improve benchmark scores. I understand there is probably a bigger bottleneck with the FSB in this regards than the memory latency but I thought it was worth trying. Also it would prove whether or not the 1066 memory would be better, worse or the same as the 1333Mhz memory. If it's worth creating another thread to continue this I can. I'll let Moo decide. Right now there are a limited few actually displaying benchmark results and I thought this discussion applies.

Regarding the Vantage score it is with Physx off.

 

honestly i could care less @ this point Mandrake can post here and when he is done post his findings in another thread......

This whole thread just died only Mandrake was posting and kceissm and Clauss just about

 

The last time a 1066Mhz vs 1333Mhz thread was created it was thrown into side topics. If it's there I fear there would be less technical eyes looking at it. I thought the discussion would be better out of that area.

 

i didn't know you hit 11,909 on vantage. the 100 points should be easy.

download nhancer and set everything to high performance.

 

I'm following your every post like a hawk

 

Lot's of great advice today. Downloading new drivers and nhancer now.

 

let us know your results.

 

I think I'll go back to the 1333 memory so everything is the same as my previous tests.

 

yeah so far you didn't have better results with the 1066 anyway right?

 

No. Things have been less stable with the benchmark tests.

 

are you debating between the RAM to be used? if just benchmarking.. run with only a single 2GB stick..see if that helps. Ideally you'd want 2 1GB sticks for parity. But whatever.. keep everything static as you said for true comparable results. I'll PM you with some ideas.