lik

Distinguished
Jan 27, 2004
222
0
18,680
Is it odd to have very good performance in games and then low scores in benchmarking? I have 5300 in 3DMark 2k3 and 14.500 in 3DMark 2k1! But I play cs 1.6 @ 250-300Fps, FarCry (Demo) @ 40-80Fps, Halo @ +/- 50Fps. Could this mean benchmarks aren't really a good source for new buyers? Or is just this an isolated case? GreatApe in your wisdom help me please! I really like your posts! Thanks
 

andyb

Distinguished
Dec 9, 2003
119
0
18,680
Im not quite great ape but what card do you have?

My system:
AMD Athlon Barton 3000xp+
ATI Radeon Hercules 9800pro 128mb
WD 120gb 7,200rpm 8mb cache.
512mb twinmos pc3200
ASUS a7v8x-x Mobo.
 

andyb

Distinguished
Dec 9, 2003
119
0
18,680
As for your benchmark being too low 5,300 in 3dmark03 is about right and i get that same result with my r9800pro, so no need to worry there. 14,500 does seem slightly low but what are your system specs eg processor,mobo,ram.

My system:
AMD Athlon Barton 3000xp+
ATI Radeon Hercules 9800pro 128mb
WD 120gb 7,200rpm 8mb cache.
512mb twinmos pc3200
ASUS a7v8x-x Mobo.
 

andyb

Distinguished
Dec 9, 2003
119
0
18,680
Although your 3dmark03 score is fine im puzzled as to why your 01 score is slightly lower but id probably put it down to program activity in the background. Other than that everything seems fine even if it is the same after you rebenchmark i wouldnt worry becuase as you say your gaming performance is top notch and thats what really matters, well it does to me! :)

My system:
AMD Athlon Barton 3000xp+
ATI Radeon Hercules 9800pro 128mb
WD 120gb 7,200rpm 8mb cache.
512mb twinmos pc3200
ASUS a7v8x-x Mobo.
 

lik

Distinguished
Jan 27, 2004
222
0
18,680
Yes I think that too! But what I'm seeing now is that the really important numbers to look at when you want to by a new gf card is the fps in games, no benchmarking scores. Is it suposed to be like this? Or is the purpuse of a benchmark to help future buyers? To be or not to be...
 

andyb

Distinguished
Dec 9, 2003
119
0
18,680
Well benchmarks are desined to give people a ground to compare which is good becuase it enables people (like you) to be able to test out your GC and get the reasurance that your card is performing as well as it should be, but this should be taken with a pinch of salt as it were becuase benchmarks are not based on games (directly)and can be affected by other things such as programs running in the background.

My system:
AMD Athlon Barton 3000xp+
ATI Radeon Hercules 9800pro 128mb
WD 120gb 7,200rpm 8mb cache.
512mb twinmos pc3200
ASUS a7v8x-x Mobo.
 

andyb

Distinguished
Dec 9, 2003
119
0
18,680
Just out of interest what make is your card?

My system:
AMD Athlon Barton 3000xp+
ATI Radeon Hercules 9800pro 128mb
WD 120gb 7,200rpm 8mb cache.
512mb twinmos pc3200
ASUS a7v8x-x Mobo.
 

speeduk

Distinguished
Feb 20, 2003
1,476
0
19,280
3dmark 2001 is still the best benchmarking tool around because it gives a good indication of OVERALL performance. Anything around the 16k mark is a decent score and should play games fine and anything over 20k is really fast and will play all this years games fine (hopefully).

<A HREF="http://service.futuremark.com/compare?2k1=7523662" target="_new"> MY A64 System </A>
 

pauldh

Illustrious
Your 3dmark2001 scores are low because you are running 1 stick of Ram instead of 2 in Dual channel. Limiting you memory bandwidth.

ABIT IS7, P4 2.6C, 512MB Corsair TwinX PC3200LL, Radeon 9500 Pro, Santa Cruz, Antec 1000AMG, TruePower 430watt
 

arussell

Distinguished
Feb 3, 2004
68
0
18,630
The above poster hit the nail on the head :). I was thinking the same thing when I saw that 1x512. Get another stick of 512, preferably the exact same model as the other stick and see if you can dual channel, or, if you want a more sure-fire thing, just buy a dual kit and replace that 1 stick, that way your sure that the sticks have been tested to run in dual channel.

System on the way:
Antec Sonata Case w/380W PSU
Athlon 64 3200+
MSI K8T Neo-FIS2R
1 GB Mushkin PC3500
Sapphire Radeon 9800PRO
WD Raptor 74 GB
 

andyb

Distinguished
Dec 9, 2003
119
0
18,680
Anything around the 16k mark is a decent score and should play games fine and anything over 20k is really fast and will play all this years games fine (hopefully).

-----------------------------------------------------------

So are you saying my new system is not going to be able run this years next years games well?? seeing as i get just under 16,000 3d01se? I dont personally think either is best (03 & 01se), also my card is a DX9 card which runs BEST on the newer games so i think your 20k+ result is a bit far fetched, i would say anything over 14000 is a good score and capable to play games well but you have to remember that it only tests DX8 not DX9 and thats were 03 comes in.

My system:
AMD Athlon Barton 3000xp+
ATI Radeon Hercules 9800pro 128mb
WD 120gb 7,200rpm 8mb cache.
512mb twinmos pc3200
ASUS a7v8x-x Mobo.
 

speeduk

Distinguished
Feb 20, 2003
1,476
0
19,280
It depends on what game ur playing. A new DX9 game (X-2) is quite cpu dependent. Your xp3000 rig will get around 80fps stock @ 1024x768 and a 3200 A64 will get 115fps. If you up the resolution to 1280x1024 then ur 3000 will get around 55fps and 20fps low in some areas but the A64 will get about what ur system got @ 1024x768 (80fps).

3dmark 03 isn't a good indication for all upcoming games. I think unreal 2004 will be VERY cpu dependent, and a 3000xp will just cope on max resolution and detail while a 3200 A64 or similar will fly.

I upgraded from a 2500 @ over 3200 speeds and I can tell the difference big time. Even in halo which isnt very cpu dependent.

<A HREF="http://service.futuremark.com/compare?2k1=7523662" target="_new"> MY A64 System </A>
 

andyb

Distinguished
Dec 9, 2003
119
0
18,680
I do agree with you but its slightly unfair comparing with a 64bit processor, and also what aspect of performance are you talking about FPS? e.g. the difference between a 2600 and 3000 processor would be minimal (4 to 6 fps)

My system:
AMD Athlon Barton 3000xp+
ATI Radeon Hercules 9800pro 128mb
WD 120gb 7,200rpm 8mb cache.
512mb twinmos pc3200
ASUS a7v8x-x Mobo.
 
The main problem with benchmarking versus gameplay is that different games stress different aspects than some benchmarks.

Your benchmarks do seem to be about right, but they definitely don't tell the whole story. BTW, with the right CPU/MOBO/HD and Memory I could match or beat your 3Dmk01 score with my R9600Pro, yet get no where near your gameplay, so I don't believe in any benchie being <b>THE</b> one.

Benchmarks can give you some insight into the power of a piece of hardware/software or the combination of many parts of your hardware/software.

The thing is that there is not single indicator or predictor. 3Dmk01 was a good indicator of gaming potential in the late DX7 early DX8 era, and translated pretty consistently. It looked at the overall system and that was the way many games worked offloading many jobs onto the CPU that are now more VPU/GPU intensive.

Now the games are getting more complex. A smoothly coded game that takes adavanatages of extensions and such will run much smoother than a game that is poorley coded (HALO or Splinter Cell)

The thing is to know HOW the benchmarks relate to the games you are interested in playing. Is the game you want to play affected more by the power of the CPU the VPU or both.
Most reviewers and most of us here, are saying that one should wait until the game you want to play has been benhcmarked by knowledgeable/reputable people, and then you will get an idea of what to expect from your system.

3Dmk01 is very cpu/memory (whole system) dependant, and a score can be greatly affected by a different CPU, memory, etc. This is good for some of the older games because they really don't rely on efficient shaders and such to achieve similar effects, they don't use PS/VS 2.0 and neither does that benchmark. You can see Dungeon seige take more advantage from a CPU increase than a VPU change.

3Dmk03 is very VPU dependant. Changes in the rest of the system have less of an impact. This is true for more modern games that rely heavily on the graphics cards. Look at these benchmarks by Xbit, and tell me if you think this tells the whole picture?

<A HREF="http://www.xbitlabs.com/articles/video/display/graphics-cards-2003_27.html" target="_new">http://www.xbitlabs.com/articles/video/display/graphics-cards-2003_27.html</A>
<A HREF="http://www.xbitlabs.com/articles/video/display/graphics-cards-2003_28.html" target="_new">http://www.xbitlabs.com/articles/video/display/graphics-cards-2003_28.html</A>
<A HREF="http://www.xbitlabs.com/articles/video/display/graphics-cards-2003_29.html" target="_new">http://www.xbitlabs.com/articles/video/display/graphics-cards-2003_29.html</A>

Finally you have something like Command and Conquer where the VPU is almost an afterthought that is on required to draw the most rudementary graphics where the CPU has the most impact;

<A HREF="http://www.xbitlabs.com/articles/video/display/graphics-cards-2003_39.html" target="_new">http://www.xbitlabs.com/articles/video/display/graphics-cards-2003_39.html</A>

Of course if you turn on AA/AF then the story changes, but that is because those features are exclusively dependant on the graphics card.

There are also strange anomalies like X2 which doesn't uses Pixel Shaders, and relies heavily on Vertex Shading. It's design would faovur a card like the FX5700U which has 3 vertex engines versus the R9600 series' 2.

Really the day of a single benchmark being the determinant is gone. A review like that one from Xbit is the best way to get an idea of how well a card will perform in the games you favour.

Also rememebr that benchmarks are quickly optimized for whereas games have to be out and deemed worthy of optimizing for in order to receive the same attention that those benchmarks get. Which is one reason why the run-time compiler scores don't tell the whole story, but then again neither do the raw non-forceware(d) results.

Basicallly it's no longer an easy task chosing the card that best suits you, and is the best use of your money. You can't trust just one reviewer, even one like Xbit, THG or Digit-Life., now you must find out what EVERYONE gets/thinks about the new card. Perhaps only one of them will run the benchmark or game that shows the cards weakness. That was the case with 3Dmk03 and the early FX series. Every other game/benchmark said it was faster and better and there were no indications of potential issues.

Many people point to synthetic versus gaming benchmarks, but really there is no difference. The results in Quake 3 are no more indicative of future performance than 3Dmk01, same with Max Payne 2 and 3Dmk03. Each has their strength and wekanesses and may expose something you hadn't thought about in a specific architecture, but none will give you the overall picture. I think those days are long gone.
You may be able to think of it this way in the future AI=CPU and Graphics = VPU, but really that's just one trend, and there are many others like raytracing that go in a completely opposite direction. At best we can get a 'feel' for one era at a time, and even then there will likely be a bunch of surprises.


- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! - <font color=green>RED </font color=green> <font color=red> GREEN</font color=red> GA to SK :evil:
 

speeduk

Distinguished
Feb 20, 2003
1,476
0
19,280
The thing is 3dmark 03 only went up about 400 marks and the game tests a few fps where as in games the performance is improved at least 20% over my last setup. If you just upgraded yer 3d card then 3dmark 03 is a good test too see how much faster it is. If you just upgraded yer cpu/mem then 2001 is the way to go.

<A HREF="http://service.futuremark.com/compare?2k1=7523662" target="_new"> MY A64 System </A>
 
Exactly. The thing is to know how the benchmarks relate to what you're looking at. Perhaps even some benchmarks stress factors that we won't see in games for a while (like 3Dmk03 or rthdribl at the beginning of the year) so there is little way to correlate it to something other than itself and the hardware it tests.

Considering the number of benchmarks out there (recently added nbench to my repertoire) you can get a very good overall picture of your system from head to toe. From PCmark, Sandra, Htach and nbench to 3Dmark, Aquamark, SPECViewPerf and Codecreatures, etc. there should be something to stress your rig.

It's also interesting to see which programs are and aren't affected by certain tweaks, like raytracing being totally unaffected by VPU overclocks, but dependant on memory and cpu changes.


- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! - <font color=green>RED </font color=green> <font color=red> GREEN</font color=red> GA to SK :evil:
 

lik

Distinguished
Jan 27, 2004
222
0
18,680
Thanks a lot for everything! Is it better to spend lots of money in new sticks of ram? or to by another equal to the same I have? Considering it is not from any known RAM brand. I'm not interested in OC.