• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

processor speed

qwo

New Member
Joined
Aug 22, 2006
Messages
48 (0.01/day)
i wonder how can processors' performance to be measured. Are they measured in FLoating point Operations Per Second?

i read the statistics on folding@home, i could not understand the performance measurement between CPU,GPU and the Cell processors
 

qwo

New Member
Joined
Aug 22, 2006
Messages
48 (0.01/day)
i heard comparing Cell(3.2Ghz) i7 965(3.2Ghz) and GeForce 280 (602Mhz)
GeForce will have the higest FOPS, cell ranks 2nd and i7 ranks third,
is this correct?
does that mean i7 is the slowest?
thanks
 

qamulek

New Member
Joined
Feb 7, 2008
Messages
184 (0.03/day)
its all about the ppd(points per day)

i heard comparing Cell(3.2Ghz) i7 965(3.2Ghz) and GeForce 280 (602Mhz)
GeForce will have the higest FOPS, cell ranks 2nd and i7 ranks third,
is this correct?
does that mean i7 is the slowest?
thanks

i wonder how can processors' performance to be measured. Are they measured in FLoating point Operations Per Second?

In a nutshell yes, however the definition of a FLOP is a bit shady. Reading on their website faqs it seems that if a processor(gpu/ps3) can do a math expression such as exp(x) in one operation they say it is one(sometimes two?) FLOPS, but if the exp(x) needs to be calculated in 10 or so operations then they say it is 10 FLOPS. FLOPS are a bit sketchy here, which is why I'm guessing people look at real world performance and not FLOPS. Not only are the definition of FLOPS a bit shady between the processor types, but the FLOPS, as mentioned in their faq, don't represent pure performance since some[many?] people pause their WU's to do other stuff before the WU is completed.

You also have to remember that the there are differences between each processing type which is represented in the number of points given to each processing type for a particular WU. I have scoured their website and it seems they take a benchmark machine of each processing type, then give it a WU and see how long it takes to execute that WU. If a WU takes one day to execute on the benchmark machine then they give it K points for that WU where K is different for each processing type. For each processing type K is:
gpu===> K=1500
ps3===> K=900
cpu===> K=110

These values of K take into account how flexible each processing type is compared to the others. For instance the FAQ for the ps3 says the ps3 is about 10x as fast as the [benchmark?] cpu, but the points-per-day[ppd] the ps3 recieves in comparison is less then 10x the ppd of the cpu which represents the reduced flexibility in the types of WU's that the ps3 can execute. The same reduction in ppd is no doubt also reflected in the ppd of the gpu.

The benchmark machine for each processing type is:

gpu===> ati-3850 using an athlonx2 4000+
ps3===> ps3
cpu===> pentium4@2.8Ghz

For folding@home I believe all current cpu's are slower then the ps3, and in general the ps3 is slower then most gpu's(is there a gpu that is slower then the ps3? I don't know...).
 
  • Like
Reactions: xvi
Joined
Nov 4, 2005
Messages
11,687 (1.73/day)
System Name Compy 386
Processor 7800X3D
Motherboard Asus
Cooling Air for now.....
Memory 64 GB DDR5 6400Mhz
Video Card(s) 7900XTX 310 Merc
Storage Samsung 990 2TB, 2 SP 2TB SSDs and over 10TB spinning
Display(s) 56" Samsung 4K HDR
Audio Device(s) ATI HDMI
Mouse Logitech MX518
Keyboard Razer
Software A lot.
Benchmark Scores Its fast. Enough.
Processor speed is no longer the issue, performance is. Performance per watt, performance at clock speed and performance at price.


AMD/ATI has the market cornered in performance per dollar, and the midrange. Intel has a good midrange offereing, holds the performance crown by a small margin, and has a very mediocre low end offereing. nvidia is still overpriced and believe their shit doesnt stink.


Imagine it like this.


Intel i7 is like a Porsche, it costs as much as a porsche, the accessories cost as much as a porsche accessory would and it goes 150MPH but only carries two people.

AMD Pehnom is like a Covette, it only costs as much as a corvette, the accessories are available for cheap and it runs 145MPH

The cell processor is like a bus on crack, it costs as much as a crackhead to support, only wants what it wants and runs 200MPH


ATI/Nvidia GPU's are like the bus strapped on the space shuttle, they are way more expensive, and do way more than just graphics these days.






So, in short. CPU's are slower than either, the cell processor is a costly oddball that no one wants at the family dinner party, and if you believe a graphics card is only for that you need to read up on what has happened int he last three years.

I watch shows on the internet and my ATI card accelerates it, cleans it up, and only uses 3% of its total processing power.
 
  • Like
Reactions: xvi
Joined
Nov 10, 2006
Messages
4,665 (0.73/day)
Location
Washington, US
System Name Rainbow
Processor Intel Core i7 8700k
Motherboard MSI MPG Z390M GAMING EDGE AC
Cooling Corsair H115i, 2x Noctua NF-A14 industrialPPC-3000 PWM
Memory G. Skill TridentZ RGB 4x8GB (F4-3600C16Q-32GTZR)
Video Card(s) ZOTAC GeForce RTX 3090 Trinity
Storage 2x Samsung 950 Pro 256GB | 2xHGST Deskstar 4TB 7.2K
Display(s) Samsung C27HG70
Case Xigmatek Aquila
Power Supply Seasonic 760W SS-760XP
Mouse Razer Deathadder 2013
Keyboard Corsair Vengeance K95
Software Windows 10 Pro
Benchmark Scores 4 trillion points in GmailMark, over 144 FPS 2K Facebook Scrolling (Extreme Quality preset)
In a nutshell yes, however the definition of a FLOP is a bit shady. Reading on their website faqs it seems that if a processor(gpu/ps3) can do a math expression such as exp(x) in one operation they say it is one(sometimes two?) FLOPS, but if the exp(x) needs to be calculated in 10 or so operations then they say it is 10 FLOPS. FLOPS are a bit sketchy here, which is why I'm guessing people look at real world performance and not FLOPS. Not only are the definition of FLOPS a bit shady between the processor types, but the FLOPS, as mentioned in their faq, don't represent pure performance since some[many?] people pause their WU's to do other stuff before the WU is completed.

You also have to remember that the there are differences between each processing type which is represented in the number of points given to each processing type for a particular WU. I have scoured their website and it seems they take a benchmark machine of each processing type, then give it a WU and see how long it takes to execute that WU. If a WU takes one day to execute on the benchmark machine then they give it K points for that WU where K is different for each processing type. For each processing type K is:
gpu===> K=1500
ps3===> K=900
cpu===> K=110

These values of K take into account how flexible each processing type is compared to the others. For instance the FAQ for the ps3 says the ps3 is about 10x as fast as the [benchmark?] cpu, but the points-per-day[ppd] the ps3 recieves in comparison is less then 10x the ppd of the cpu which represents the reduced flexibility in the types of WU's that the ps3 can execute. The same reduction in ppd is no doubt also reflected in the ppd of the gpu.

The benchmark machine for each processing type is:

gpu===> ati-3850 using an athlonx2 4000+
ps3===> ps3
cpu===> pentium4@2.8Ghz

For folding@home I believe all current cpu's are slower then the ps3, and in general the ps3 is slower then most gpu's(is there a gpu that is slower then the ps3? I don't know...).

How on Earth can you write such good replies as through as that and only ever have one "thanks"? I tip my hat to you, sir.


If I may add my two cents to a slightly old thread, your average processor is a bit like a jack of all trades. Good at all and master at none. A processor can be programed to do nearly anything, but it will not do it quickly. Instruction sets can help, but they only add slightly to efficiency (relative to the two below).

Stream processors, such as the ones found in today's video cards, can not be used to do anything and everything, but what they can do, they do extremely quickly. They're much more specialized. They usually get their advantage from parallel computing, something that a lot of the @home projects like.

Cell processors I am less familiar with, but if I understand them correctly, they're good with Floating Point Operations Per Second(FLOPS). Any program properly coded to take advantage of the hardware (if available) should benefit greatly.

Stream and Cell processors, being the specialized hardware it is, will naturally be faster than a CPU processor.

I may be wrong in the details, but I should have the basic principles correct.

If you'd like to look in to it more, check the Wikipedia articles.
Stream Processing
Cell Microprocessor

..and reading up on RISC versus EPIC wouldn't hurt either. RISC being today's CPUs in a nutshell and EPIC being somewhat relevant and/or similar to Stream processing. (..sort of)
RISC (Reduced instruction set computing)
EPIC (Explicitly parallel instruction computing)
 
Last edited:
Top