I just benchmarked my new rig with 3DMark 2003 and 2006 for the first time, and to compare checked out some reviews online. According to this article my scores are WAY below where they should be... Leadtek, EVGA Geforce 8800 GTX tested, even in SLI And according to this one my system rocks ths casba... Nvidia GeForce 8800 GTS Reviews. Graphics Cards Reviews by CNET. What's the deal? I'm using the default settings on both, no tweaking or overclocking yet. I looked around both sites to see how they set up their benchmarks and didn't find much... :-k
I'd like to know as well, when I tested my 7950gt in this system it came up well below many published 3Dmark benchmarks. I tested as soon as it was up and running and again after OC But before I had all the junk "start up" programs and "add ins" had a chance to accumulate.
No doubt. I mean, CNET scored 8890 in 3DMark 2006 with their GTS and I scored something like 8000. Pretty close. But other sites are reporting numbers around 20,000-30,000. Clearly there has to be some variance in the way the test is set up. And take 3DMark 2003. I searched around here and Bubbles and few other folks were getting 40,000+, whereas my rig scored around 15,000! WTF?
Well, first the Inquirer tends to just pull shit out of its ass (much like its namesake). Also, I don't even see the 8800GTS listed in their run article. CNET looks to be about right. You could also check Hardocp and nvnews for further articles and (more importantly) active forum threads dedicated to actual user scores. Not to knock scrubbin, but we have no idea what settings and other hardware he has in his system. Check here for some comparison benchies.
what about the Futuremark online results browser? nobodys mentioned it yet, have they? or is it something obvious that i missed?
Hah! Funny you should mention it because I was browsing it just now. And my numbers are dead in the middle of those posted there. So it begs the question, how in the fuck are people getting such high scores? Hey Bubbles, did you alter any of the default settings when you ran your 3DMark 2003 test? What is your system specs again?
That brings back memories! I gave up on benchmarks a long time ago. Now I just go by whether the games I want to play work well or not... There are so many things that affect how the 3dmark scores come out. I remember how my scores dropped in 3dmark when I got a new soundcard, simply because the new soundcard had testable features that the AC97 sound did not. All my games ran better because of the new card, but my score declined. I often found that settings that gave me a better 3dmark, often gave me choppier game performance, while low scores gave me a better gaming experience overall.
Turbo button, that's awesome. I don't dwell on benchmarks, but when you build a new system they are absolutely necessary to make sure things are installed and configured properly. From the numbers I'm seeing it would appear everything is OK. So any other benchmarks that you would suggest to measure general performance?
A lot of times, if a system is tweaked out just a little bit (ie. bus speed, processor speed, ram timings, video memory timings, etc.) they can HUGELY increase the scores in Futuremark. I guy I used to work with had two 7800GTXs in SLi on a pretty decent Athlon64 and at the time, his score was low to mid, compared to the rest. Well..it turns out that lot of the systems that had the same cpu speed, weren't close to the same cpu. In other words, overclocked to the max. So...who cares what those stupid benchmarks say. As long as it smokes like you want it to, that's all that matters.
Personally I only use FM2006 for a rough gauge of: A) is my sytem performing up to simular systems B) did my little tweak/OC give me any boost. Also for stability testing. Right now I just use ATItool "Scan for Artifacts" while running Orthos. It doesn't give me any compariable numbers, but it assures stability and gives me my max thermal load. If I can run that overnight without issues, then its solid. For benchmark numbers I've been using FEAR.
Hey Lazaredz (and you other Core 2 Duo folks), what voltage are you using for your overclock? My 6400 is rock solid at 2800mhz / 1.36v, idle temp of 37 full load of 42. With a little more juice I know I could easily hit 3000-3200 but I don't want to go hog wild. I may just stop where I am anyway. The truth of the matter is that the system in it's current state has more horsepower than I am likely to need very much. Outside of Oblivion of course... EDIT: And how are you going about the overclock? Just upping the memory speed, or what? We need details!
Your stock voltage overclock sounds about par for the corse. Nice little boost with decent temps. To squeeze the most out of the Core2s though most people are uping the voltages to around 1.5. This article at IGN is acutally a good start and a shorter version of the same guide that nvidia put out themselves. Process is pretty simple, once you have bumped your voltages and locked down/turned off certain BIOS features, push the FSB a bit, test your temps, check for stability while monitoring those temps. If it looks good Wash, Rinse, Repeat. Good quick tests included Super Pi 1.5 over at xtremesystems.org (for the CPU) and ATItool's "Scan for Artifacts" (for the GPU). If you can make it though a few minutes of each, your at least 95% stable. For long term tests I run Orthos (Prime 95 for Duo Cores) while running the ATItool. After a few hours, I consider it case closed. IF you are overclocking your memory (which really isn't necessary *) then try Memtest86+ for stability testing. You'll need to boot to it, but its easy to find a boot CD with it ready to go. *= Tests have shown that not only do memory timings have little to no affect on Core 2 performance, but even what SPEED that memory is negligiable (1-2%). Unless you have some GOOD memory you bought just for overclocking , I would not even mess with it. Just use dividers or whatever your motherboard provides to keep the speed as close to stock as possiable. Also stay with default memory settings (4-4-4-12 or whatever). Its not worth the headache/time. Unfortunatley this is no good "covers it all" temp monitoring program. I have been using Intel's Thermal Analysis Tool (TAT) for CPU, ATItool for my 8800GTX (ironic???), and speedfan for my hard drives. At max thermal output, you would like your CPU to be under 60C, CPU under 90C, and hard drives under 40C. Those are pretty safe maxes that still allow for a little wiggle room. Get another 10C above those and you are looking at some damage. Oh, and use Rivatuner to overclock your GPU. The utility with the latest nvidia driver is not working correctly. Anything else?
Another reason why benchmarking is important! I was finishing up my build today, installing my remaining applications, etc, when I noticed that my video card was running at a whopping 82 degrees C at idle! Ran some benchmarks and found that my GPU fan was spinning at 0 RPMs... My damn IDE cable had gotten jammed up against it. I freed it up and now my card idles at 54 degrees C.
No doubt. The scary thing is that there was no warning at all. My games ran fine, the Nvidia software didn't give any warning, etc. You would think that the Nvidia software should mention a fan RPM of zero...
I THINK there is a setting in the drivers to alert for fan speed. I KNOW there is a temp warning that can be implimented. I think the default for it though is 120C! These chips are designed to be HOT!!!
