AMD's smoothness factor put to the test by AMD & HardOCP...

[kl0012]

Blind experiment

...In some disciplines, such as drug testing, blind experiments are considered essential. In other disciplines, blind experiments would be very useful, but they are totally impractical or unethical...

A classic example of a single-blind test is the "Pepsi challenge". A marketing person prepares several cups of cola labeled "A" and "B". One set of cups has Pepsi, the others have Coca-Cola. The marketing person knows which soda is in which cup but is not supposed to reveal that information to the subjects. Volunteer subjects are encouraged to try the two cups of soda and polled for which ones they prefer. The problem with a single-blind test like this is the marketing person can give subconscious cues which bias the volunteer, whether or not these were intended. In addition it is possible the marketing person could prepare the separate sodas differently (more ice in one cup, push one cup in front of the volunteer, etc.) which can cause a bias. If the marketing person is employed by the company which is producing the challenge there's always the possibility of a conflict of interests where the marketing person is aware that future income will be based on the results of the test.

[Manicdan]

You can learn nothing from this benchmark. First of all this "benchmark" is not reproducible because it is built on "human opinion" and therefore can be easily rigged w/o any danger to be caught. Second, "human opinion" is not good measure metric because "human opinion" can be easily manipulated - there are a lot of techniques how to do it.

human opinion is the final result of all graphics rendering. if somehow amd found a way to get 30fps to feel better than 45, we should actually applaud them (although its probably not that big of a difference between the systems)

granted we have no idea how the test was set up, someone a few pages back says it was set up by hardocp i think (point being NOT amd), we cant be sure how things were manipulated.

but to say human opinion means nothing is like calling rasterization a hoax and all games that dont raytrace as con-artists.

[kl0012]

but to say human opinion means nothing is like calling rasterization a hoax and all games that dont raytrace as con-artists.

Please dont try to pervert what I said. I didn't say that human opinion means nothing. I said that the result of such "blind" test based on human opinion means nothing because human opinion can be easily manipulated (read my prev. post).

[SMTB1963]

Testing party does not have to be neutral, only must have a truthful setup of the equipment (no software tweaks etc) and provide full disclosure of testing data and methodology.

Sorry, it's more than that. I can setup the equipment "truthfully", but if I'm not controlling the testing environment, then the results will be suspect. If a booth babe rubs her tits against testers playing one system but not the other during testing, do you think the results might be skewed?

Or there could have been more subtle environmental influences:



I briefly stuck my head into Eddie Deen's Ranch on Saturday. Believe me, what was happening in there should not be confused with a controlled testing environment.

[naokaji]

both cpus are so fast I'm super skeptical you can tell a difference. games are held back by the graphics card, and the cpu difference is always small...

And once you reach 60 FPS the difference is null and void due to the Monitor not displaying more than 60 FPS anyway.

[BababooeyHTJ]

And once you reach 60 FPS the difference is null and void due to the Monitor not displaying more than 60 FPS anyway.

Unless you have a 120hz display.

[Mats]

A KVM switch would have made sense here.

[STEvil]

You can learn nothing from this benchmark. First of all this "benchmark" is not reproducible because it is built on "human opinion" and therefore can be easily rigged w/o any danger to be caught. Second, "human opinion" is not good measure metric because "human opinion" can be easily manipulated - there are a lot of techniques how to do it.

Plenty to learn from it, unless you're a fanboy.

You learn: There is a possibility that your rival is doing something you arent.
You learn: We should test this.
You learn: We should look into using a more correct testing methodology
You learn: Fanboys cry and scream at night when you bother to learn things to make life hard for them.


SMTB: Environment is something that should be disclosed in the testing methodology.

[kl0012]

Plenty to learn from it, unless you're a fanboy.

You learn: There is a possibility that your rival is doing something you arent.
You learn: We should test this.
You learn: We should look into using a more correct testing methodology
You learn: Fanboys cry and scream at night when you bother to learn things to make life hard for them.


SMTB: Environment is something that should be disclosed in the testing methodology.

Oh, please... What is the point of all this BS?
Who cares about old, cheap Pepsi-Co marketing tricks? May be only some really hard fanboys.
Sorry AMD, but this is a new low...

[kl0012]

I don't understand how people can be so naive (or perhaps fanbois?). There is no such thing - unmeasurable smoothness. Everything is measurable, even if it has hungered of parameters (ask Apple about that). And AMD has all the resources to measure it and to show the numbers. But instead they went with the cheap marketing action.
BTW, it's not that hard to rig the test - just decrease mouse dpi in drivers - and you're done even if your system has much higher fps.

[STEvil]

I dont understand how you can be so naive. I am not saying parlor tricks are ok use, I am only saying that anyone can do any test they please as long as they tell us how it was done so we can tell if they used parlor tricks or not.

YES EVERYTHING IS MEASURABLE, WE WANT TO MEASURE THIS.

I think you need to re-read my posts. I am not defending the specific methodology used by H or AMD, only saying that what was done should be learned from. Is that hard to understand, are you so blind you cant see that?

[ice_chill]

I dont understand how you can be so naive. I am not saying parlor tricks are ok use, I am only saying that anyone can do any test they please as long as they tell us how it was done so we can tell if they used parlor tricks or not.

YES EVERYTHING IS MEASURABLE, WE WANT TO MEASURE THIS.

I think you need to re-read my posts. I am not defending the specific methodology used by H or AMD, only saying that what was done should be learned from. Is that hard to understand, are you so blind you cant see that?

The only thing to learn here is how to BS.

[cold2010]

I have a Surprise

My 8120 @4.0 run skyrim smoother than 2500k @4.5
Both are my pc s

[xlink]

http://www.anandtech.com/show/2715

give it a read...

I noticed that things felt slower when I upgraded from my 2.6GHz Opteron 165 to a 3.2 - 3.6 Ghz C2d e6400.
I also noticed an increase in snappiness when going to my PII x4 720be.

The Intel systems definitely install things more quickly though.

I'm going to say it's probably something along the lines of this - the intel systems have more raw horse power, the AMD systems have more torque.
To some extent, it doesn't matter a ton though, most people, even enthusiasts have more CPU power than they need. I'm primarily limited by my internet speed, or my own inability to type more quickly.

[Tha Last Meal]

My 8120 @4.0 run skyrim smoother than 2500k @4.5
Both are my pc s

lol .....You are dreaming, man
intel 2500K processor can kick your amd 8120 processor ass

you hear me
you hear me
you hear me
you hear me

[tiro_uspsss]

lol .....You are dreaming, man
intel 2500K processor can kick your amd duron 8120 processor ass

you hear me
you hear me
you hear me
you hear me

ftfy!

[Mechromancer]

In light of recent events I felt it was necessary to necro this post.

Reviewers are using a new method to review GPUs. Nvidia has created a tool called FCAT that takes a look at how frames are displayed, rather than just what the game engine reports the FPS is. Links describing this in complete detail are below (PCPer has a nice video if you want to be quick about it).

Techreport: http://techreport.com/review/24553/i...-capture-tools
Anandtech: http://www.anandtech.com/show/6862/f...marking-part-1
PCPer: http://www.pcper.com/reviews/Graphic...ormance-Testin

I feel the topic of the OP relates to all of this new frame time testing directly. AMD systems may in fact be SMOOTHER than Intel system. I have a Core i7 2600k/Z77 system running crossfire. I can play Battlefield 3 on High at 60 to 120fps....albeit with a ton of stutter/dropped frames/runt frames. My coworker has a measly AMD FX-4100, with the same HD 6850 crossfire on an AMD 970 chipset. His system allows for CrossfireX to be enabled (Crossfire through the chipset/PCIe AND Crossfire cable simultaneously). His system ran only between 30 and 65fps during gameplay but clearly had no stutter/dropped frames/runt frames. At a reported 35fps his system played smoother than mine at 75fps. His 970 chipset was also pushing 1 card at PCIe 2.0 16x and the other at 4x, yet he was still 'smoother'.

This irked me...it irked me enough to necro this thread. AMD systems may very well be smoother in Crossfire configurations given the added features that support crossfire on their chipsets. I urge XS members to please write Techreport, Anandtech, and PCPer to do more testing with AMD systems vs. Intel systems. Reviewers tend to only use Intel systems when doing all testing, but this may not be showing the entire picture (literally). Also let's continue this discussion given this new frame time point of reference and get to the bottom of this on XS.

Thanks folks.

[Greg83]

Size doesn't matter in processors either, nor how fast the performance, its all about how you use it and the hands on approach, but only developers get their hands inside that processor .

[zalbard]

I can play Battlefield 3 on High at 60 to 120fps....albeit with a ton of stutter/dropped frames/runt frames. My coworker has a measly AMD FX-4100, with the same HD 6850 crossfire on an AMD 970 chipset. His system allows for CrossfireX to be enabled (Crossfire through the chipset/PCIe AND Crossfire cable simultaneously). His system ran only between 30 and 65fps during gameplay but clearly had no stutter/dropped frames/runt frames. At a reported 35fps his system played smoother than mine at 75fps. His 970 chipset was also pushing 1 card at PCIe 2.0 16x and the other at 4x, yet he was still 'smoother'.

Umm, what?

First of all, CrossfireX is just a name for the third generation of Crossfire, signed by switch to ribbon interconnect cables.
Secondly, both Crossfire and SLI heavily rely on PCI-E bus for load balancing.
Pretty sure it works this way (people in the know will correct me if I am wrong):
PCI-E is used to feed cards the data, and for load balancing;
SLI/CF bridge is used for minor stuff, such as synch. purposes and frame buffer swaps.
And cards don't transmit a lot of data between each other directly. Talking to the CPU and RAM is a lot more important and bandwidth intensive.

[Greg83]

plus i remember back in the 4870x2 days, that it only had the bandwidth of like a pci-e 1x slot
not much can be done with that.
wouldn't want to use it for the bandwidth, it would compromise latency. best to keep that high way empty.

[Pantsu]

The CPU bottleneck will separate the AFR frames, while in a GPU bound scenario they'll appear around the same time, which causes the runt frames and makes CF useless in that circumstance. It's no wonder games felt smoother with FX since it was inadvertently fixing the microstutter.