3DMark06 - low score? and no lvl2 cache listed!

[Jizumonkey]

Hi, I just bought a new system (as some of you may have read on here recently) and although I know my graphics card is currently the weak link (bought with the intention of upgrading to dx10) I'm wondering about two things.

My score is just over 4200 and also there is no lvl 2 cache listed...is this right?

CPU 1/1
Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
Manufacturer Intel
Family Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
Architecture 64-bit
Internal Clock 2.4 GHz
Internal Clock Maximum 2.4 GHz
External Clock 266.0 MHz
Socket Designation Socket 775
Upgrade Socket 478
MultiCore 2 Processor Cores
HyperThreadingTechnology N/A
Capabilities MMX, CMov, RDTSC, SSE, SSE2, SSE3, PAE, NX
Version Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
Caches
Level 1 20 KB
Level 2 0 B

Graphics card is only a Palit 7900GS and I have 2x1GB OCZ Gold at 800Mhz DDR2.

Thanks.

 

[Jizumonkey]

OK I can not worry about my score too much but has anyone else seen their cache not displaying like this?

 

[Heyyou27]

Ignore the L2 cache reading; your score is normal.

 

[Mousemonkey]

Looks a bit low to me mate, with a 939 3800x2 @ 2.4Ghz and a 7900gt @ 540/780 I get 6199 in 3D mark ’06 and I’m using ‘cheapo’ kingston value RAM DDR400

 

[Heyyou27]

His is a 7900GS, and chances are the clockspeed is also lower than that of your 7900GT.

 

[Assman]

and it looks like he's not running @ 1:1 ratio, which means he's not using his 800mhz ram to its fullest potential. (correct me if i'm wrong)

 

[rquinn19]

my scores are in my sig and i only have 1 gig of 667 ram

 

[dasickninja]

Did you remember to kill any P2P programs, anti-virus software (such as Norton or McAfee) and any other process hogging programs before you started the test? That usually has an adverse affect on how 3DMark scores..

 

[darkstar782]

Yes, thats a significantly overclocked 7900GT from stock speeds (450/660).

His card has 1/6th less pipes (20 vs 24) and probably runs at 450/660.

4200 is fine for a 7900GS. I dont have a Core2 myself, but it wouldnt surprise me if 3Dmark06 cant read the L2 cache as it doesnt understand the unified nature of the C2D's L2 cache.

 

[Mousemonkey]

True my friend ,true, however I still think the OP has some tweaking to do, a score of 5000+ or so should be well within the realms of possibility.



Edit Addum: 5153 with no OC

 

[Jizumonkey]

Hi all and thanks for your replies as usual.

I have updated my bios this morning as it was an old version, new features include voltage optimisations. I have oc'd the cpu to around 2.8GHz and also the graphics cards a little and am running 3dmark06 now.

By the way, this is a clean install of XP...so is very uncluttered!

 

[Jizumonkey]

Can I ask at what resolution you lot run 3DMark? I've found in the past that people can inflate scores by running low resolutions. My desktop and hence tests are run at 1280x1024 with anti-aliasing off and filtering on optimal (which is default).

Anyway, the slight oc i've done has resulted only in a score now of 4550.

Did I do the wrong this by buying 800MHz Ram? My FSB is at about 300 atm (cant remember the exact figure) with a multiplier of 9 for the cpu and 2.5 for the ram. Should I be aiming for a whole number on the ram like 3, with maybe a sacrifice to the multiplier on the cpu?

 

[Mousemonkey]

At 1024x768 CPU @2.4 GPU@540/780 = 3D Mark 6199 CPU Score = 1787

At 1280x1024 CPU @2.4 GPU@540/780 = 3D Mark 5238 CPU Score = 1792

Bit of a difference when the resolutions change, just goes to show how much a 3D Mark actually 'means'

Edit: CPU Scores added

 

[darkstar782]

The default settings on 06 are 1280x1024 I think, and thats what most people use for comparison.

Its not based on your desktop res, I run in 1600x1200 on a HP LP2065 20.1" IPS LCD, and it still defaults to 1280x1024.

I suppose with a Core2Duo giving him a massive CPU score, 5k should be doable with a 7900GS.

With my D805@3.8GHz and my 7900GT at its stock 550/1400 speeds I get 5200 or so.

Thats an old result before I added more RAM (that was with 1gb) and voltmodded my GPU, raising it to 725/1400, but I should think he could come somewhere near, as my CPU score was shite. (and still is, waiting for the Core 2 Quad QX6700 or the Xeon 3070 before I upgrade CPUs)

 

[Jizumonkey]

Well I wont turn this into a 3dmark thread (since we are in the CPU discussion) but I did manage to get higher to 4750 last night with further overclocking of the graphics card. CPU score is 2500.

I do have a CPU related question however. I posted this in the overclock section last night but have had no reply. Can someone please shed some light on this for me? Pasted below:

I got myself a nice new setup yesterday as added in my sig, one quick question. I've followed the guide for overclocking after dabbling myself (not a complete newbie) but there's one thing I've never bothered to ask...until now!

What is the deal with making the FSB and memory speed match with an even multiplier? For example, before I had my FSB at 266 as recommended, and something like a 3.5 multiplier on my memory to 800MHz DDR and a 9x multiplier on my CPU to make 2.4GHz.

After reading I did what I thought was sensible and set my FSB to 400MHz, my memory multiplier to 2x (to match the memory at stock speed, not my main area to oc at the moment) and my CPU multiplier to 7x for a safe oc of 2.8GHz. All is well. which is nice...but why the match up with the ram? Have I done the right thing in dropping the multiplier for the sake of 'matching' the ram like this?


I mean, now my next jump is to put an 8x multiplier on the CPU (the only overclock i have left really unless I want to mess with the RAM) and this will take me up to 3.2GHz. Now, in my mind this is starting to get a little much. I dont want to risk too much, and am not bothered about bragging right, but a nice 'free' gain makes me feel like I'm getting one up on Intel hahaha

Thanks.

 

[mitch074]

Aaah... This is an old problem with Intel CPUs: matching FSB and RAM clock speeds, due to northbridge-based memory controller. It is indeed offset in a large measure by a big unified L2 cache, but still.

The first thing to do: see how high your CPU can go. Then check the highest FSB speed it can support. Once there, see the highest clock rate your RAM can take without shooting timings to hell (if you have to take it more than 1 tick lower - meaning over 4-4-4-8 instead of 3-3-3-7 - than its default speed when overclocking, then it's running too fast and may fry). Now then, look for the closest integer multipliers you can get to match those clock; if RAM and FSB are 1:1, it's GOOD.

Now granted, this had a much bigger impact on AMD K7, but it still can get you some more speed and a lot of extra stability.

 

[Jizumonkey]

Ah OK thanks thats helpful...however, I do not have anywhere in my BIOS (as far as I can see) that allows manual setting of the memory timings on this motherboard. Strange for one of this cost and apparent ability.

Do you think I've done a bad thing by simply setting the FSB to 400MHz with this in mind? Should I take it back a bit and just do things how I was before by setting an odd memory multiplier?

The OCZ i have is 5-5-5-12, i should add that to the sig.

 

[mitch074]

some motherboards have 'advanced' settings only available if you press a key combination. Read the manual.
Moreover, it can be hidden due to some parameters set to 'auto' - on my mobo for example, the timings are completely hidden if I don't first Shift-F1, then go to Northbridge, then to RAM, select 'Manual' for RAM settings, then switch 'RAM timings' from 'SPD' to 'manual' - quite a quest.

 

[Mousemonkey]

The thing I was trying to avoid was any kind of AMD Vs Intel slagging off which is why I posted my CPU scores, and whilst I know GS is no rocketship with the C2D backing it up I felt a better score was in there (as did yourself) and as this thread has progressed so the OP is reporting higher numbers (c'mon 5K, c'mon 5K) and yes you are correct 1280x1024 is the default for 3D '06 but I have noticed that when some folks have posted screen shots of their scores the settings box shows that the test was run at 1024x768 which is default res for 3D '05