• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Stressing FAB process over actual performance?

X

xBanzai89

Supreme [H]ardness
2FA
Joined
Aug 24, 2007
Messages
4,812
I ran across a thread on another gaming forum. The OP is looking for what cards would be better for TF2. One posters in the thread is stressing FAB process over the actual performance of the card.

Here is the post:

"Only people that stress fab process know what theyre talking about
GTS 250 is built on 55nm
GTS 240 is built on 40 nm
GTS 250 is faster, but gives only about 80 percent of performance per watt.
So only me, and xyz are the only ones giving good advice.
I dont own either 5770 or 4770, i can guarentee you that 5770 isnt anywhere close to 50 percent better than 4770. If it was it would break the laws of physics (not physx)."

I have never seen anyone stress this for normal games. Sure I see this argument used for CPUs in servers, but for one GPU it doesn't apply as far I know that is what the trend follows.

I mean this guy is recommending a GT240 (pos gpu) over a 4770. The OP has the budget for that as well. Can anyone back up what this guy is saying?
 
Performance/watt is a number that depends not-only on on fab process used, but also on the configuration and voltages used on the card itself.

Does the card have an efficient VRM?

Does the card use GDDR3 or GDDR5 (more power consumed)?

Does the card use 4, 8, 14 or 16 DRAM devices (more devices = more power).

What voltage and frequency does the GPU use (higher voltage/frequency = more power used).

What type of process was used (high-performance or low-power)?

FAMOUS EXAMPLE:

The 6600GT could offer the same performance as a 6800 vanilla, but it used 48w peak versus just 38w peak for the 6800 vanilla.

http://www.xbitlabs.com/articles/video/display/geforce6600gt-oc_3.html

This despite the fact that the 6800 was built on an older .13 micron process, and the 6600GT was using the new .11 micron process. This happened because the .11 micron process was cutting-edge and leakier, and the fact that they cranked-up the voltage on the 6600GT to get the core speed up to a whopping 500 MHz (6800 had a core speed of 325 MHz). Also, the 6800 used lower-power DDR1 memory, while the 6600GT used newer power-hungry GDDR3.

There's always more to choosing a card than the fab process.
 
Last edited:
My my thank you! The 2900 and 8800 is a example of that as well correct?
 
xBanzai89 said:
My my thank you! The 2900 and 8800 is a example of that as well correct?
Click to expand...

Yes it is. that was mostly because the 2900 was on an extremely leaky 80nm process (so leaky most other builders skipped it), while Nvidia was using a mature 90nm process. Add in the 4 extra memory chips required by the 512-bit bus, and the 2900 was an energy hog.

Of course, there's more to it than that: you can tell from the chip design that ATI was in a hurry, and skipped some obvious improvements. The first of these would have been increasing the number of texture units. 16 texture units wasn't enough on the x1900 series, and it certainly wasn't enough to power the 2900XT. The second was the exclusion of the FSAA hardware: somehow they couldn't get the hardware working in-time, so they shipped the card as-is and used shaders to perform the AA.
 
Last edited:
Performance per Watt is a useless metric unless you only care about how much power a card uses. If you are a gamer, it really only factors into deciding between two cards. It doesn't matter if the GTS 240 is more efficient than the GTS 250, at the end of the day it is *much* slower, so who cares?

EDIT: Oh, and here is the forum thread in question: http://forums.steampowered.com/forums/showthread.php?t=1073718&page=4 Looks like the thread is mostly filled with useless fanboy drivel and people who judge a card based off of retarded metrics like fab process and *TRANSISTOR COUNT* - I shit you not, the guy the op is talking about claimed a 4770 and 5770 performed about the same.

"One last thing, the performance between the 9800 gtx, 9800 gtx + 4770 and 5770 are all about the same"

Hahaha, epic fail thread
 
Last edited:
You guys found the thread :O. Oh well I don't care lol.

Yeah it drove me nuts when I saw that guy talking about that on top of the thread being filled with fan boys.
 
You must log in or register to reply here.
Back
Top