Anand: PII vs. Q9550 vs. i7 crossfire, Phenom II = smoother

Printable View

Show 100 post(s) from this thread on one page

02-03-2009, 08:13 PM
crazydiamond

maybe something that could calculate the amount of milliseconds (?) between each frame or something?
02-03-2009, 08:21 PM
roofsniper

1 Attachment(s)

go into fraps and tell it to log frame times. this is what you get.
02-03-2009, 08:32 PM
Titan7171

Quote:

Originally Posted by LedHed

The Phenom II can't even manage to beat out the Q6600 in most cases:

http://www.xbitlabs.com/images/cpu/p...ing/crysis.png

http://www.xbitlabs.com/images/cpu/p...ocking/fc2.png

http://www.xbitlabs.com/images/cpu/p...ocking/ut3.png

http://www.xbitlabs.com/articles/cpu...g_8.html#sect0

Wow a whole 1280x1024!.....maybe they should have run the test through s-video on a regular tv.:rolleyes:

How about using a resolution that actually taxes something in the system other than your eyes.
02-03-2009, 08:37 PM
crazydiamond

Quote:

Originally Posted by roofsniper

go into fraps and tell it to log frame times. this is what you get.

im not doin the math on all them for the amd & intel tests :p:
02-03-2009, 08:40 PM
Sparky

I'm not a pro with excel but I have enough know-how to make it do calculations and some charts and stuff, if someone wanted to run these tests with fraps and send me the data from fraps I'd be glad to do some data analysis of it. Maybe fraps data from Chew's testing?
02-03-2009, 09:03 PM
LedHed

Quote:

Originally Posted by Titan7171

Wow a whole 1280x1024!.....maybe they should have run the test through s-video on a regular tv.:rolleyes:

How about using a resolution that actually taxes something in the system other than your eyes.

If you knew anything about benchmarking then you would know lower resolution tests use the CPU more and the higher resolutions shift to the GPU.

Many CPU reviews use lower resolutions benchmarks to show this.

Meaning when you want to show CPU gains you use lower resolution benches.
02-03-2009, 09:06 PM
roofsniper

Quote:

Originally Posted by LedHed

If you knew anything about benchmarking then you would know lower resolution tests use the CPU more and the higher resolutions shift to the GPU.

Many CPU reviews use lower resolutions benchmarks to show this.

Meaning when you want to show CPU gains you use lower resolution benches.

it can but can you explain why amd tends to do better on higher res than lower res? and when i mean better i mean higher than intel values. nonetheless it has nothing to do with the article or with smoothness.
02-03-2009, 11:03 PM
Aura89

Quote:

Originally Posted by roofsniper

it can but can you explain why amd tends to do better on higher res than lower res? and when i mean better i mean higher than intel values. nonetheless it has nothing to do with the article or with smoothness.

i agree, the guy you're quoting has nothing to do with this forum, this is not about who gets higher frame rates at what resolution, it's about:

Hey, this CPU gets 90fps average in this game, but goes all the way down to 30fps every now and then

While this CPU gets 75fps average, but only goes down to 55fps every now and then

someone wanted the definition of smoothness, there's your smoothness, the difference between lows and highs on many occasions, i mean really, who cares if you get 200fps if every 10 seconds it goes down to 1fps, i wouldn't want to play like that
02-04-2009, 02:41 AM
Final8ty

http://forums.overclockers.co.uk/sho...php?t=17896754

Maybe someone can ask the Author for that program as it will work out the avg deviation for you from fraps.
The DL links are dead.
New link http://www.mediafire.com/download.php?2moyugibhez
02-04-2009, 02:54 AM
Brother Esau

Quote:

Originally Posted by Aura89

i agree, the guy you're quoting has nothing to do with this forum, this is not about who gets higher frame rates at what resolution, it's about:

Hey, this CPU gets 90fps average in this game, but goes all the way down to 30fps every now and then

While this CPU gets 75fps average, but only goes down to 55fps every now and then

someone wanted the definition of smoothness, there's your smoothness, the difference between lows and highs on many occasions, i mean really, who cares if you get 200fps if every 10 seconds it goes down to 1fps, i wouldn't want to play like that

Well thats probably the most intelligent way of putting it by just keeping the logic basic:)
02-04-2009, 05:09 AM
Hornet331

Quote:

Originally Posted by *GameOver*

Intel guys come here and combat the AMD guys asking what is "Smooth or Smoothness" I consider smoothness to mean, While im playing "insert game" i see very little jerky moments where i can tell the frames are dropping below 30fps.. If your frames go from 60 to 9.... You notice... You also notice large fluxes from say 30 to 80.. To be honest the fact that Intel system CAN pull higher FPS sometimes hurts it... Going from 30 to 160 LOOKS different then going from say 30 to 60.. You're eye might not notice the difference in those number in real life situations but Monitors react differently. When your frames go from 60 to say 20.. Its Smoother. Now mind you this is not ALWAYS the case. Sometimes AMD drops just as many frames as the Intel Chips do. But i tend to notice far greater FPS Pits on my Intel systems... In my store we have our "Head to Head" center. We have a Q9550 Running along side a Phenom 9950. 8 times out of 10 people will pick my AMD system running the Same game.. On the SAME monitor. with the Same ram Ect. And the same thing i hear time and time again is "Monitor 2 just looks smoother!" Take it for what you will. But if the shoe fits.. This is coming from your Average user coming in and just Playing a game on 2 similar machines... Most of these users are not even aware that there are 2 different chip makers... All they care about is How many GB's the computer has :P When you take the Pepsi Challenge you cant get pissed when people pick Coke. Its also not about what chip is better on paper... a Corvette will get you to work... But when you hit the Pot holes wouldnt you rather have the Buick? :P

there where already many threads about that topic which came to similar ideas. When in fact the only thing that influences the "smoothness" factor is the difference between max/min fps, then there is a easy fix for the intel system -> limit max. fps or enable vsync.

If that hypothesis would turn out to be true, a intel rig would be sometimes even more smoother cause of the often higher min fps rate then a amd rig or the other way riudn if the amd rig has a higher min fps and the max fps is limited.

Imho its the "smoothness" debate is the same thing as the "they eye cant see more then 30fps" debate. It is based to much on the individual to make a general statment, just like people say they can see the difference between 60fps and 100fps. Maybe some individual can but the majority can't, its the same with audiophiles, play joeavarage a 192kbit/s encoded mp3 and a flac file on a highend headphones, they won't notice the difference, hell maybe they would even say the mp3 sounded better. :p:

Quote:

Originally Posted by Aura89

i agree, the guy you're quoting has nothing to do with this forum, this is not about who gets higher frame rates at what resolution, it's about:

Hey, this CPU gets 90fps average in this game, but goes all the way down to 30fps every now and then

While this CPU gets 75fps average, but only goes down to 55fps every now and then

someone wanted the definition of smoothness, there's your smoothness, the difference between lows and highs on many occasions, i mean really, who cares if you get 200fps if every 10 seconds it goes down to 1fps, i wouldn't want to play like that

well just like said above, there are often cases where the min fps is higher on the intel rig -> enable vsync, limit max. fps and you have the same smoothness as the amd rig, but your artificially limiting your gfx card of showing what power it has.
02-04-2009, 05:13 AM
marten_larsson

Quote:

Originally Posted by Brother Esau

Well thats probably the most intelligent way of putting it by just keeping the logic basic:)

But with same/lower min fps for AMD systems this doesn't clarify anything. It's more like this:

System A has avg fps of 120 but has min fps at 40 which happens every five seconds.

System B has avg fps 100, has min fps of 35 but this happens every 40 seconds.

System B would probably be a more fluid experience than A. The main issue is, as maany has already stated; fps stability. If a system fluctuates between 50 and 150 fps dropping to 35 sometimes, averaging 120, it will probably be considered more stuttering than a system fluctuating between 50 and 70 fps, dropping occasionally to 35, avg fps at 55. Microstuttering should be a similar thing.
02-04-2009, 05:20 AM
Hornet331

Quote:

Originally Posted by marten_larsson

But with same/lower min fps for AMD systems this doesn't clarify anything. It's more like this:

System A has avg fps of 120 but has min fps at 40 which happens every five seconds.

System B has avg fps 100, has min fps of 35 but this happens every 40 seconds.

System B would probably be a more fluid experience than A. The main issue is, as maany has already stated; fps stability. If a system fluctuates between 50 and 150 fps dropping to 35 sometimes, averaging 120, it will probably be considered more stuttering than a system fluctuating between 50 and 70 fps, dropping occasionally to 35, avg fps at 55. Microstuttering should be a similar thing.

afaik JumpingJack already did show framrate over time with a QX9650 and a P1 9850 on quadfire.

edit:
just look at that for example:
Company of Heros
http://forum.xcpus.com/gallery/d/756...ps_compare.JPG

Racedriver Grind (done on a 8800 GTX)
http://forum.xcpus.com/gallery/d/701...15SEC_FRAP.JPG

also this smashes the theory that systems behave differently form each other, if a system gets a drop every 30secs the other system also will get a drop every 30sces, when the same situation is played/displayed.
02-04-2009, 05:31 AM
Titan7171

Quote:

Originally Posted by LedHed

If you knew anything about benchmarking then you would know lower resolution tests use the CPU more and the higher resolutions shift to the GPU.

Many CPU reviews use lower resolutions benchmarks to show this.

Meaning when you want to show CPU gains you use lower resolution benches.

And if you knew anything about benchmarking you would know that using results from a different setup with different settings at a different resolution is totally useless here.

The tests AT did wasnt benchmarking they were actually playing the games using real world resolutions.

You cant say a game felt smoother by watching a benchmark
02-04-2009, 07:30 AM
Sparky

Quote:

Originally Posted by Hornet331

well just like said above, there are often cases where the min fps is higher on the intel rig -> enable vsync, limit max. fps and you have the same smoothness as the amd rig, but your artificially limiting your gfx card of showing what power it has.

Doesn't vsync add input lag? I thought I've read that somewhere on here so I don't use it.
02-04-2009, 07:42 AM
Xello

Quote:

Originally Posted by Hornet331

Imho its the "smoothness" debate is the same thing as the "they eye cant see more then 30fps" debate. It is based to much on the individual to make a general statment

There's no arguing for the point of "eye can't see more than 30fps" though, the statement is completely wrong, it's not even a matter of 'some people can, some people can't'.

Quote:

Originally Posted by SparkyJJO

Doesn't vsync add input lag? I thought I've read that somewhere on here so I don't use it.

Very slight but unless you're playing online it's negligible :up:
02-04-2009, 07:50 AM
iandh

Quote:

Originally Posted by Hornet331

afaik JumpingJack already did show framrate over time with a QX9650 and a P1 9850 on quadfire.

edit:
just look at that for example:
Company of Heros
http://forum.xcpus.com/gallery/d/756...ps_compare.JPG

Racedriver Grind (done on a 8800 GTX)
http://forum.xcpus.com/gallery/d/701...15SEC_FRAP.JPG

also this smashes the theory that systems behave differently form each other, if a system gets a drop every 30secs the other system also will get a drop every 30sces, when the same situation is played/displayed.

Very interesting graphs, and as a general statement the AMD systems do have less framerate variance from a glance.

I wouldn't necessarily say those graphs "smash" anything though; we are only seeing 1/40th of the available data. Each connected data point is avg fps per second, so any hiccups would be included in that average. we would need to see a non-smoothed raw data graph at least, but a histogram of frame render delay would be absolutely ideal.
02-04-2009, 08:17 AM
gosh

Quote:

Originally Posted by Hornet331

also this smashes the theory that systems behave differently form each other, if a system gets a drop every 30secs the other system also will get a drop every 30sces, when the same situation is played/displayed.

You need to calculate on more detailed level to understand how smooth the game is. The graphic card and CPU works asyncronosly. If you calculate based on when picture is drawn from the GPU this isn't the same as when the CPU calculates points for the picture. The GPU is always behind the CPU. The GPU will also smear time spans between when the CPU calculate the picture.
To get a better value you need to time the EndScene API, when they are called if we are talking about Direct3D.
How the game feels depends on when the picture is calculated, not when it is drawn on the screen. Also the mouse is important, reading mouse data needs to be very exact.
02-04-2009, 08:29 AM
Brother Esau

You know Guys Graph & Charts do not prove everything :)
02-04-2009, 08:35 AM
demonkevy666

Quote:

Originally Posted by Hornet331

afaik JumpingJack already did show framrate over time with a QX9650 and a P1 9850 on quadfire.

also this smashes the theory that systems behave differently form each other, if a system gets a drop every 30secs the other system also will get a drop every 30sces, when the same situation is played/displayed.

Q9650 vs phenom 9850.
12MB cache vs 4Mb cache, and this was about DENEB Not Agena. agena will be slow preclock vs deneb even with agena Nb at 2.6ghz and cores a 3.0ghz.
02-04-2009, 08:37 AM
demonkevy666

Quote:

Originally Posted by Brother Esau

You know Guys Graph & Charts do not prove everything :)

here one for every one

"It's an optical illusion"

that your eyes conclude to. lol
02-04-2009, 09:08 AM
Lostfaith

the COH chart from hornet actually 'proves' this thread's point.

I think the word instead of smooth might also be "Stable framerates" as the drops and speedups in fps for the Q9550 are more noticeable while playing and thus not so smooth.

obiously vert. sync will help but then we're doing apples vs oranges in this thread.
02-04-2009, 09:18 AM
keithlm

Quote:

Originally Posted by roofsniper

go into fraps and tell it to log frame times. this is what you get.

I think that the chart you displayed with the frames and the time would be much more useful than the "max/min" idea or any graph of FPS over time.

Using the data you displayed you could make a graph of the time DELTA (change in time) between frames. (So it would show the time spent between each frame.)

Optimally this graph should be as flat as possible. Although realistically it won't be due to the operating system, rendering different things in the game, communication between CPU cores in a multi-threaded games, cache contention in the CPU, and a zillion other factors. (i.e., some factors related to the CPU, some are not. A good tester would attempt to eliminate most factors except for the CPU.)

BUT if you look at a graph of the delta vs the frame generally the system with the flatter graph would probably be the "smoother" system. Or if one system had a major dip in this graph on occasion and the other system did not. Of course this would have to be something that can be reproduced; a good tester would throw out weird anomalies that only happen one time and not use them.

NOTE: I know this image I am showing below is NOT REALLY RELATED. I am throwing it here into this post as an example of something similar to what we might expect to see by creating the type of graph I propose. POP QUIZ: If this WAS a graph of the time delta versus frame... which system would you rather have? (And amusingly the system with the higher min and max in this non-related example is also not the "better" system.)

http://img3.imageshack.us/img3/3779/...examplevj5.jpg
02-04-2009, 09:43 AM
Hornet331

Quote:

Originally Posted by demonkevy666

Q9650 vs phenom 9850.
12MB cache vs 4Mb cache, and this was about DENEB Not Agena. agena will be slow preclock vs deneb even with agena Nb at 2.6ghz and cores a 3.0ghz.

just look back before the release of P2, even there you had the "smoother" argument.

In my eyes the whole "smoother" debate is a non issue, if people have concerns about "smoothness" just tell them to enable vsync/frame rate limiter and everythings fine.
02-04-2009, 09:49 AM
keithlm

Quote:

Originally Posted by Hornet331

In my eyes the whole "smoother" debate is a non issue, if people have concerns about "smoothness" just tell them to enable vsync/frame rate limiter and everythings fine.

<sarcasm mode ON>
We're all just shocked that you, of all people, find this to be a non-issue.
<sarcasm mode OFF>

This is most especially true since we almost have a constructive conversation going on. That can NOT be allowed. You'll need to run to the other forum subsection and gather some of your cohorts.

Actually I'm sure I speak for many people when I say your comment is worthy of a very large yawn.

And changing the vsync won't actually change anything. It might hide the issue from any method that could possibly be used to determine the difference. (But then hiding the issue is probably acceptable to some people.)

Show 100 post(s) from this thread on one page