Printable View
Eh, scores are all over the map, and they don't seem to match the users hardware very well. I'm not going to pass judgement until it's been aroundfor awhile and analyzed. At least my scores are about what I expected:
overall: 2699
SM2.0: 1155
HDR/SM3.0: 1091
CPU: 886
System in sig, although my CPU is running at 2.5ghz for everyday use.
i got 4453, but apparently my drivers aren't approved.
Its not that hard to figure out, if you think about it. The false illusion that a dual core CPU will give you such a huge boost in real games. At 1920x1200, the CPU score doubles, and gives about a 20%+ boost in the total score. That is not going to happen in a real game.Quote:
Originally Posted by Timman_24
Case in point, I have a A8N-SLI, [email protected], 2xBFG 7800GTX's 460/1300, and 2gigs of ram. Another user who I compared my score to, has a A8N-SLI, 4800+ [email protected], 2x7800 GTX's 430/1200, and 1.5gigs of ram. One highly overclocke system, one stock, who scores higher? He does, by over 700 points, 7k, to 6.3k. Does anyone really think his system is faster than mine? I have 250Mhz more on the CPU, a ton more bandwidth, more ram that is in dual channel mode when his is not, and my video cards are clocked higher as well. Yet 3dmark tells me, my system is not as fast as his.
That is misinformation. Yes I know about a select few dual core games that recieve a boost, yet that is only at low resolutions. Most people who has SLI is not going to be playing at 1024x768, or even 1280x1024. My system is faster in every key area, yet his shows a higher score. Please.
Not to mention they show a "0" for the score of HDR+AA for NV cards, because they cannot do it. So they only give a SM2 and CPU score, yet a comparable ATi card that can do HDR+AA gets a SM3 score, SM2 score, and CPU score. The two cards are not doing the same tests. Comparing them like that is misinforming as well.
There are other ways this version is flawed, thats just two of the bigger ones.
Actually quite a lot of recent things are CPU limited. Especially with SLI and the newer cards. How do you know his ram is not running dual channel? Its going to be 2T but he could easily have 2x512 and 2x256.Quote:
Originally Posted by Ackmed
Eric
My score is: zero, zip, nada
I'm too busy playing games to worry about 3DMark6,999,451.
So you are trying to tell me, his system with 250Mhz less on the CPU, less bandwidth (because mine is heavily overclocked), and slower cards, and less ram running in a lower performance mode, is going to be faster than mine? Sorry, but you would be silly to think that. Yet 3dmark shows it being a lot faster.Quote:
Originally Posted by ewitte
Because I know he has 1xgig, and 1x512.
Since we don't know much about the engine or the specifics of the test, your assumption is erroneous. His DC processor may very well score higher than yours, even if it is 250MHz slower clocked, if the engine was optimised for DC. Recent benchmarks of the FX-60 (2.6GHz DC) have shown it to be faster in some games than the 2.8GHz FX-57.Quote:
Originally Posted by Ackmed
Since I'm not sure why they would limit SM3.0 scoring only to systems that could do HDR+AA, I'm not sure on that. Technically, SM3.0 has nothing to do with the ability to run HDR+AA.
His CPU is not 250Mhz less than yours. His CPUs are 250Mhz less than your CPU. The newer drivers do take a little advantage of DC. Depending on the game, some get a good 10-20% higher framerates. While others don't get a thing. If your taking 100% advantage of only the CPU and the application supported SMP there would have to be closer to 1500Mhz CPU difference for you to get faster.Quote:
Originally Posted by Ackmed
Eric
3dmark06 is a benchmark to show you how your system will do in upcoming games. Atleast that is what I always though 3dmark was for. Therefore, I'm sure they programmed it to utilize DC much more than the games that are out right now. Since, later games will use DC and there will be a benefit to it. Right now, your system may be faster, but as programmers start to code for DC yours will slowly lose ground. Just a thought.Quote:
Originally Posted by Ackmed
98 3dmarks, get at me :cool:
I have to ask, why is it you have 600 some points less then me, when your rig is so much OC'd beyond mine? I mean is not the 5900 and the 9800 the same gen and in the same performance bracket?
Maybe I am missing something obvious, and I am not trying to be a ****.
My guess is that it is in fact the video card. While the 5900 can do DX9, it can't do it very well...nowhere near as fast at DX9 computations as your 9800.Quote:
Originally Posted by Endersothergame
It wouldnt be faster than mine, at any sort of high res. 3Dmark gives the same CPU score, no matter what resolution the tests are ran in. If you have any benchmarks to show that at 1600x1200 a DC CPU gets higher frames than a comparable single core CPU, please let me know. Let alone 1920x1200.Quote:
Originally Posted by Un4given
About the HDR+AA, perhaps I worded it wrong. It gives a SM3 score for NV cards in SM3 tests, but only if AA is not selected. If it is, it gives a big fat 0.
Its still one CPU, just dual core. Dont argue semantics. Please show me one game, where at 1600x1200 a DC CPU system will get 20% more frames than SC system.Quote:
Originally Posted by ewitte
I cant believe some of you are really trying to tell me his system is faster than mine. I have 250Mhz more Mhz on the CPU. I have 2gigs of PC4000, he has 1.5gigs of PC3200. I have 30/100Mhz more on the core/memory on the GTX's, and a lot more memory bandwidth because my system is overclocked, and his is not. It just takes a little common sense to tell me that mine will run games faster at any sort of high res. This is why 3mdark06 is more misinforming that ever before.
Once again, when its not at CPU bound resolutions, the DC CPU wont do a thing for him, or anyone. And considering the extra horse power I have, its laughable to think his would be faster than mine at any sort of high res (or otherwise really), simply because of a DC CPU.Quote:
Originally Posted by Timman_24
3DMark has a CPU only test. Even with your 250MHz of extra clock speed, his DC processor may score sufficient enough points here to have a higher overall score. A64 chips are not bandwidth hindered, which is why S754 Athlons score nearly identical to S939 chips when CPU speed is the same. Also, just because you have 512MB more RAM doesn't mean squat. RAM is one of those things that only increases speed if the program requires more RAM than that which is present in the system. 1.5GB of RAM is more than sufficient and would not require his system to swap data from the hard drive, thus making your extra RAM irrelevant.Quote:
It wouldnt be faster than mine, at any sort of high res. 3Dmark gives the same CPU score, no matter what resolution the tests are ran in. If you have any benchmarks to show that at 1600x1200 a DC CPU gets higher frames than a comparable single core CPU, please let me know. Let alone 1920x1200.
Then I'm not sure what the problem is. NV cards cannot do HDR+AA, so obviously the AA would have to be disabled to prevent the program from trying to perform AA functions where HDR is enabled. In this case he may score higher because he gets extra points for being able to perform this function. The new 3DMark is not meant to look at current game technology, but forward looking where shader power will be more important, and the growing use of HDR.Quote:
About the HDR+AA, perhaps I worded it wrong. It gives a SM3 score for NV cards in SM3 tests, but only if AA is not selected. If it is, it gives a big fat 0.
Quote:
Originally Posted by Un4given
Oh, I had no idea the difference would be so great, but if thats the case then I am surprised nVidia made it through that card cycle at all. Although it does explain the prolific use of radeon based cards from that era. Thanks for the help Un4given.
*edit* I knew that my 5200 was a terrible card but I figured that was because it was the lowest end bracket for the FX class cards. Seems the problems extended all the way through the line?
No problem.Quote:
Originally Posted by Endersothergame
If you recall there was quite a problem with HL2 and Valve was working on separate rendering paths, and NV on drivers to overcome the deficiences in the FX5xxx series cards and their ability to do DX9.
But if you are still using an FX5xxx series card, I would seriously recommend getting a newer card, even if it is something like a 6600GT, which is a little faster than a 9800 Pro, and can be had for < $150, and is even available in AGP.
Quote:
Originally Posted by Un4given
Wow fast responce, and you answered my edited addition before i typed it. ;)
Thankfully no, I got rid of my FX5200 when I got my Ti4200, that was replaced with the 9800P in my sig now.
*edited for spelling mistakes*
I know it has a CPU only test. However, it gives the overall score about a 20% from a DC CPU. That does not translate to real world performance. Show me one game that gets a boost from a DC CPU at 1600x1200. Im well aware DC CPU's can give a boost in a select few games, and a low resolution, and can speed up certain applications, such as video and photo work, but we arent talking about that. We're talking about which will give higher frames in games, a 2.4gig A64 X2, or a 2.75gig single core A64.Quote:
Originally Posted by Un4given
I know the amount of ram doesnt mean higher frames. Futuremark claims 3dmark is a "Gamer's Benchmark", and ram does mean a lot in gaming, as far as smoothness of a game, and the loading time. I pointed out the ram, because in the three main areas (CPU, GPU, Ram) my system is faster, and has more of it in every area.
NV cards can do HDR+AA, just not FP16 HDR with AA, such as is in Farcry. They do it in HL's Loast Coast just fine. The problem with it is, Futuremark gives a score of zero. In FS's review of 3dmark, they showed results with AA applied, and no mention of AA not working for the NV cards. Just the baseline score. That is misleading. It also lumps HDR/SM3 scores together. Since they knew some cards cant do their kind of HDR+AA, they should have seperated the two. And had the ability to give the NV score a SM3 score with AA applied. As AA is generally ran faster on ATi cards in DX9. Either just manually compare the SM2 AA scores (and dont produce an overall score), or produce an overall score using their formulas to get a "pseudo comparison." They are not consistant with their approach.Quote:
Originally Posted by Un4given
Futuremark also made some poor choices. Why penalize ATI for its decision on not supporting HW filtering with a lower score, while not penalize Nvidia for not supporting hardware AA or FP blending? Fetch4 is available on every single channel texture format, while in the test is only used in the PS2.0 tests... why not in the other ones? They didnt use 3Dc because it would have made the download "2x" the size. Who cares how big it is? "Future"mark should look into the future, as 3Dc is, and will be used.
But Im tired of talking about it. Obviously you think its an accurate benchmark that will reflect how a system will play a game. I dont agree. We'll just have to leave it at that.
I don't believe it is an accurate test, and never had. I've run it before, but strictly to compare against identical systems to determine if things are working properly. I have never used, nor indicated in any manner that 3DMark was a determining factor of how a system will perform in games.Quote:
Originally Posted by Ackmed
All I was doing was doing was point out that your assumptions about your increased CPU speed over a DC CPU was wrong, and that having 2GB Vs. 1.5GB wasn't a factor that should have given you a better score.
My point about the video card was offering possible reasons as to why it was the way it was. There just isn't enough known about the way the tests are scored to know for sure.
My assumptions are not wrong. A DC CPU wont give a huge boost in frames. The only time that will ever happen is at lower resolutions, and only a very select few games.Quote:
Originally Posted by Un4given
Here is a prime example of what I am talking about (not who I was talking about earlier); http://service.futuremark.com/compare?3dm06=83941
Mine is on the left, his is on the right. He has a 3800+ X2 @2.45gig. I have a [email protected]. My GTX's were at 475/1.3, his were at 495/1.3 (3dmark scans in 3d his clocks reported are 2d). We both have 2gigs of ram. I tried to find one with the same cards, and clocks. He gets about 100 more points in the SM2&3 tests, probably due to the slightly higher clocked cards. And about 700 points more in the CPU test. Yet his total score is about 1300 points higher. Are you really trying to tell me that his system will run games about 20% faster than mine? There is not one single game out there that will run 20% faster on his at 1280x1024. Keep in mind, he would get the same boost at 1600x1200, which is even more silly.
I also didnt say that more ram should give me a better score. Just as I think a dual core CPU shouldnt. More ram does smooth out gameplay. What I was saying is that my system was faster in the three key areas. If they want to call it a "Gamer's Benchmark", and give a huge boost for a dual core CPU, then the amount of ram should be factored in as well. Im not going to try it, but my bet is that if I put my old 2x256ram in, I would get very close to the same score. While gaming would suffer, big time. Once again showing this is far from a "Gamer's Benchmark".
Simply put, the end score does not equate to real gaming performance. Which is what it is supposed to do. As you can see in this article there is actually a decrease in performace going from single core, to dual core at 1280x0124. As I said, 20% faster in 3dmark, but not in real games. Sure you can argue that they are looking into the future, but they are simply guessing. Which is why I said 3dmark06 is misinforming. Because it simply doesnt equate to real gaming performance.
I rather hate the junk that anandtech spits out.. But...
http://www.anandtech.com/cpuchipsets...spx?i=2410&p=7
Dual core does not offer any benefits to gaming, but it does help quite a bit in multitasking. The single core AMD chips and Intel chips without HT rather suck when handling multitasking...
I notice the lack of performance all the time on my crappy A64 @ 2.75Ghz. That was the only main reason why I went with a DC over a SC solution.
But yes, the main performance boost for a game is video card.
http://www.xbitlabs.com/articles/cpu...-games2_3.html
As you can see, in most cases the CPU didnt matter that much. Some cases it did, but in others.. There were no difference from a Celeron vs FX-57...
2854
sm2.0=1129
hdr=1032
cpu=1679
cpu & vid card @stock settings (until new hsf gets here)
Have to admit I only read the first page of the thread.
One thing is for sure, all of this e-penis fencing leads to two winners - ATi and Nvidia because of all the noobs sucked in by this crappy benchmark and thinking they need to buy uber expensive hardware for leet scores :p
I don't agree with you at all... at least about 1/2. I am not a noob, sucked in and roped into buying new hardware. I bought the hardware first, and was just curious to see where my hardware fell.Quote:
Originally Posted by adslegend
I play games.. not benchmarks. :)
Quote:
Originally Posted by stonez
I agree with above sentement, although I am not convinced that was aimed at my type of user anyway :)