Author Topic: Opt test discussion (Read 6071 times)

Skuzzy · « **Reply #75 on:** August 26, 2005, 07:23:25 AM »

Not quite right streakeagle. The CPU is always important, but you will find those with high end video cards using higher resolutions with AA and AF, which mask the overall potential gains from this test.
Without knowing what those settings are, you cannot really get an idea of how much more the video card is important in this test.

Take a look at this example.

1280x1024 = 1,310,720 pixel per frame
1024x768 = 786,432 pixels per frame

Now, let's take a look at one of the low end video cards numbers, using the ground vis. An FX5200 ran it from 23 to 50 frame rate. On the surface it indicates a 54% increase in performance.

But the real performance gains could be turned into a pixel per second number. At 23 FPS and 1024x768 resolution, that would be 18,087,936 pixels per second. At 50 FPS it would be 39,321,600 pixels per second. I happen to know this particular system was running at this resolution.

Now, let's take an FX6600. It ran from 31 to 75 FPS, a 59% gain in performance. It happens to be running at 1280x1024. At 31 FPS that would be 40,632,320 pixels per second. At 75 FPS, it would be 98,304,000 per second.

What do these systems have in common? Both are 3.2Ghz Pentium 4's with 1GB of PC3200 RAM.

FX5200: 23FPS/18,087,936 PPS to 50FPS/39,321,600 PPS.
FX6600: 31FPS/40,632,320 PPS to 75FPS/98,304,000 PPS.

Now if you look strictly at frame rate gains, the FX6600 gained 16% over the FX5200 at the low end. Looking at the PPS gains show a very different number. A 56% gain over the FX5200 at the low end, and a 60% PPS gain at the high end.

Bottomline, you cannot do a proper comparison using only frame rates. Pixel rates are more important. And I have not attemtped to factor in the AA or AF impact either.

AKDogg · « **Reply #76 on:** August 26, 2005, 08:11:39 AM »

My test was with 4xAA temporal at 1280x1024x32 bit color. I probably over the 100 mil mark in pixel rate,lol.

streakeagle · « **Reply #77 on:** August 26, 2005, 09:19:34 AM »

I will post a graph of the above chart and you can draw your own conclusions. I cannot compensate for variations in resolution and quality settings, but most that bothered to post their settings in the mid range and higher computers listed a resolution of 1280x1024... presumably to match their LCDs?

If you group results by video card, the numbers are all over the place, grouped by FPS, it almost looks like they are grouped by CPU clock. I am positive this is not a coincidence. Put the best video card in the worst pc and play AH2... you will get low fps no matter what resolution or quality settings you use. Put the worst video card in the best pc and play AH2... if you drop your resolution and/or quality settings low enough, you will get playable frame rates. If you take a medium video card like the 6600GT, it scales with the CPU clock.

Some games tend to be GPU limited. Others CPU limited. Based on my own experience with 5 computers in my home and 1 laptop from work, AH2 has a very strong lean toward the latter.

Skuzzy · « **Reply #78 on:** August 26, 2005, 09:43:06 AM »

It depends on the combination. Some systems are CPU limited, some will be GPU limited.

The 3.2Ghz P4 will be very GPU limited with an FX5200 card, but this flips when you install,say an ATI850XTPE or NVidia 7800. This is why people with higher end cards put more stress on the video card by upping the AA and AF levels.

A simplistic comparion of just FPS is really not a good indicator of where the performance bottlenecks are.

Note, I am not saying AHII does not depend on a fast CPU. The faster the better. But, overall performance of a good CPU can be strangled by a slow video card.

The only reason I am bring this up is you made a very pointed statement which can be read as the video card does not matter to AHII. Nothing could be further from the truth, especially as we go forward.

NAVCAD · « **Reply #79 on:** August 26, 2005, 09:47:02 AM »

I ran the OpTest and saw pretty dramatic results.

The only gripe I had was when I went into a Field gun, the screen was filled with green. It was as if the feild gun icon was extra large. If there were more than one gun available, you couldn't destinguish which gun was selected. (If that makes any sense at all). I also notice when taking off from an airfield there was a base of a field ack on the runway. It wasn't solid so i didn't hit or crash, but it was a little strange seeing a feild ack in the middle of the runway.

Don't know if anyone else has seen this or not.

Azul32 · « **Reply #80 on:** August 26, 2005, 10:51:21 AM »

Yeah ,I have seen the field gun in middle of runway. It disappears when you get close to it while trying to land.
As for the field gun in hanger you can pick which 1 you want just click around for awhile till you see it turn green. Thats what I did and had no problems.

Mister Fork · « **Reply #81 on:** August 26, 2005, 12:31:45 PM »

And to add to Skuzzy's comments streakeagle, another bottleneck these days are soundcards.

If you're running an old 16bit or an on-board soundcard (soundcard that is built into the motherboard), you will have problems playing Aces High or other hi-end sims.

I found that I had slowdowns that was directly related to explosions, especially when there were multiple sound directions in a heavy battle at an airfield. I was using an SBLive Value soundcard, which is pretty slow for a modern machine (my on-board sound wasn't much better). When I popped in a Audigy 2 ZS (a $70 soundcard), ALL my stutters and micro-freezes went away during these heavy engagements. My machine is only an Athlon 2500+ (~1.8GHz), Ati 9700 pro, with a gig of PC3200 DDR SDRAM. In the end, it was my sound card that was causing most of my grief. FPS didn't go up too much (if at all), but it definately improved my gameplay in Aces High and other games like HalfLife 2.

To add to Skuzzy's comment on CPU/GPU limits, your machine is as fast as your slowest component. If you're using an on-board soundcard, or an on-board video card, it will really have a negative impact on high-performance games.

Something to think about...

On a side note, with the OPTEST excutable, my FPS have trippled vs the regular.

streakeagle · « **Reply #82 on:** August 26, 2005, 02:02:44 PM »

Of course every component matters, but the question is to what degree? Traditionally, gamers solve their problems by buying more RAM and video power. But some games respond better to CPU power. Take a low end PC, a mid range PC, and a high end PC. Test each one with a low end card, a mid range card, and a high end card. Observe how much performance difference you see with each swap. Only on higher end CPUs will you see a significant FPS benefit to using the better GPU, whereas FSP of mid range GPUs will scale with the CPU.

I think based on what I have seen, you will find that if you hold the resolution and image quality settings constant constant, say 1280x1024 with FSAAx4 AFx8, if you start with a mid range pc and mid range graphics card, , that you will get a better return investing in a $300 CPU with a $150 GPU than a $150 CPU with a $300 GPU.

Of course those who get a $800 CPU with a $600 GPU will get optimum performance. Being a married guy, I settled for a $330 CPU and a $280 GPU and can't complain about the all-round performance I am getting from this combo in all my sims. But I could still use that $800/$600 combo to run LOMAC maxed out, presently I have to knock the ingame settings back to Medium to keep smooth FPS with my 1600x1200x32 FSAAx4 AFx16 settings. Everything else I play pretty much locks at 85Hz Vsync when maxed out

Sound is a very annoying subject. With video cards you have at least two major brands that are aggressively competing with steady driver updates. With sound cards, there is one dominant market leader, and they don't seem to care too much whether their hardware and/or drivers work well with all the possible PC setups. It seems no matter whether you go with a motherboard chipset or a discrete card, there is always some issue with some application. My TUV4X (VIA 694X) chipsets would not even boot Windows with an SB card. I ended up with Santa Cruz cards on those machines. After the last driver was released, the only game that still gave the Santa Cruz any trouble was: Aces High! For my new MB, I decided to give the nForce sound a go... so far it works okay. But nothing impressive either.

Skuzzy · « **Reply #83 on:** August 26, 2005, 03:49:18 PM »

A 2.4Ghz Intel CPU would benefit more with a video card upgrade, than a CPU upgrade, if the system has a middle of the road video card to start with.

Streak, you could just ask us about this instead of speculating about it. I really am concerned someone is going to make a bad decision based on your speculation.
No offense nor disrespect intended.

And on the SB problem with VIA. You are putting the blame on the wrong horse. It is a VIA issue, not a Creative one. Just FYI. I do not think it is up to Creative to work around errors in the VIA chipset design. They could, but that really is the wrong place to put the problem, as it effected more than just Creative.

streakeagle · « **Reply #84 on:** August 26, 2005, 07:48:33 PM »

I guess we will have to agree to disagree, because despite the variance caused by the range of resolution and image qualities used during this test, I don't believe the trend of this graph is speculation. I would be equally concerned that someone with a 1.8 to 2.4 GHz P4 didn't notice the fact that ALL of the 70+ FPS scores are by CPUs of 3.0 or higher in P4 GHz ratings (except one AMD Sempron 2800) whereas some of those are achieved with mid-range GPUs:

I am aware of the limitation of the VIA chipsets since I have owned and maintained many of them. I just find it strange that SB products are the only ones that gave me problems. It's not like VIA users are alone either, there are plenty of intel users that have had nothing but trouble with SB products too... some of them have had experiences comparable to mine

humble · « **Reply #85 on:** August 26, 2005, 08:13:19 PM »

Hmmmm....

Looks great but literally no change in FPS I can see?

Downloaded the file...installed it to AHII main directory...says Opt 1.3 or something when I fire it up....was on line so I'll do it on trinity to be sure. Am I doing something wrong??

Schatzi · « **Reply #86 on:** August 27, 2005, 04:33:08 AM »

Quote

Originally posted by humble
Hmmmm....

Looks great but literally no change in FPS I can see?

Downloaded the file...installed it to AHII main directory...says Opt 1.3 or something when I fire it up....was on line so I'll do it on trinity to be sure. Am I doing something wrong??

You have to go to video setup and check OP terr and OP obj. They are off by default.

Skuzzy · « **Reply #87 on:** August 27, 2005, 11:56:52 AM »

Streak, your chart/graph is missing a significant piece of information which renders them inaccurate. You can base your opinion on them, but note this.

We actually know how the game works and where the loads are. Your assessment is incorrect and I am trying to explain why.

While upping CPU speed will always be a benefit to the game, it becomes an exercise in diminishing returns once you get to a certain point, and at that point, the video card is more important.

This optest relies heavily on the video card performance. The results are pretty much (with few exceptions) what we knew they would be. Everyone who uses reasonable configurations which do not starve a video card's performance got appreciably better results. No CPU change needed. And there is not a CPU on the market which could have produced the significant gains in performance as seen in this optest version. Think about it.

streakeagle · « **Reply #88 on:** August 27, 2005, 03:08:11 PM »

I understand FPS is only part of the picture, but it is the obvious and easily measured part that even your own testing is focused on, otherwise you would have specified the resolution and image quality settings for this test.

I understand the purpose of the optest is to reduce the graphics loading while maintaining the same visual quality and CPU. That is does very well! I am not even questioning that this optimization is more useful if the video card has the power to take full advantage of it.

I also understand that building a computer is like matching an engine, transmission, rear-end gear ratio, and tires. If any one of the PC components is too limited be it the CPU, RAM, GPU, or sound card, the performance of all the others is reduced or even clamped no matter what settings are used. But particulars of when and where a bottleneck will appear vary significantly with the application.

I am debating when the law of diminishing returns begins to apply for AH2 (even with the optest) given the above data (which is all I have). The point of diminishing returns seems to be well past 3.0GHz. In fact, to show diminishing returns with cpu speed increases, the pattern should have curved up as FPS increased past your stated optimum of 2.4 GHz. As long as it is linear or curving the other way, the return isn't diminishing. The weak point of the data is that most people with high end GPUs have high end CPUs and vice versa, most with mid range GPUs have mid range CPUs, but the posted data does include serveral cases of mismatched GPU/CPU combinations that help fill in that void and make the above graph at least halfway useful.

I would be interested in learning what you know since all I have is the optest results charted above. I am always eager to learn more about how pcs and the applications that run on them work since I make a living working with them as well as enjoying buiding and testing pcs as a hobby. So, if you don't mind, please continue to go into more details until I get in over my head or you run out of data/info to discuss and expose. I hope my tone in these posts is not being read as being inflamatory or disrespectful in any way. I just don't accept being told my viewpoint is wrong until the facts convince me otherwise

Skuzzy · « **Reply #89 on:** August 27, 2005, 03:34:56 PM »

Your presumption is based on incomplete facts. There are conclusions which can be properly drawn from the data, but you cannot draw an accurate conclusion about the CPU versus the video card as there is not enough data available. Our intent had nothing to do with the conclusion you are trying to present. The results bore out what we expected.

Frame rate is but a small part of the data needed to figure out whether or not the benefit lies in the CPU versus the video card.

To actually figure it out you need a series of controlled systems with various CPU speeds, and then swap various video cards in and out of those systems. You would need to run all combinations at various resolutions.

Then you could draw an accurate conclusion on what is the best bang for the buck when it comes to upgrades. To draw the conclusion you did based on the data available is a *guess* at best.

Understand, my only goal here is to make sure people understand your "guess" is just that and should not be taken as fact. I do not want to deal with support calls from people getting angry when they upgrade thier 2.4Ghz CPU to a 3.2Ghz CPU and get little to no benefit as they have an ATI 9200 video card and that was the real bottleneck.