64-bit CPU question

Open discussion about any topic, as long as you abide by the rules of course!
Psyche911
Posts: 1742
Joined: Wed Jan 02, 2002 8:00 am

Post by Psyche911 »

I'm not trying to start any more debates, I just wanted to share an article I found interesting:

http://www.driverheaven.net/articles/efficiency/

They drop the Radeon R520's clocks down to the level of the 7800 GTX's and disabled 8 of the GTX's pipelines to make the stats of both cards equal, then ran some tests.

Interesting test, I thought.
shadd_
Posts: 2512
Joined: Tue Jun 28, 2005 4:02 pm

Post by shadd_ »

well the fact g70 doesn't use the same internal clocks for vertex and pixel shaders it's not a good comparison. maybe if they allowed the 520 to use the same clocks as nv's vertex shaders would have been better.

the hardware is so different "pipes" don't really define anything anymore really or a "16" pipe 520 in the old sense would be utterly tromped by "24" pipe g70. wich it's not.
[size=75]i never meant to give you mushrooms girl[/size]
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

Looking at the last screen on the first page there, isn't the ATI memory clock set to 500 - half of the 7800's memory clock speed? Is this offset because of the ATI's new architecture?
Psyche911
Posts: 1742
Joined: Wed Jan 02, 2002 8:00 am

Post by Psyche911 »

It's 500MHz times 2 for DDR = 1000MHz.
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

Gotcha. Is DDR really perfectly comparable to double the clock speed?
[FTF]Pyro
Posts: 1233
Joined: Mon Oct 23, 2000 7:00 am

Post by [FTF]Pyro »

R00k wrote:Gotcha. Is DDR really perfectly comparable to double the clock speed?
Double data rate

ddr
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

No shit.

The operative word being 'perfectly.'
Psyche911
Posts: 1742
Joined: Wed Jan 02, 2002 8:00 am

Post by Psyche911 »

I honestly don't know how it compares to SDRAM.

But you do double the actual clock rate to get the effective clock rate.
shadd_
Posts: 2512
Joined: Tue Jun 28, 2005 4:02 pm

Post by shadd_ »

i'll have to admit from the looks of 7800gt or the 512 mb 1800xl is a tough choice.

though ati's series is new and would not be far fetched too see some good improvements driverwise that may change that.

i was looking at some benches when 6800 and x800 series first came out. the 12pipe x800 pro was matching the 6800ultra pretty much across the board, even high res.

now look, nvidia just needed some time to get their drivers up too snuff and currently the 6800 ultra smokes the x800 pro.
[size=75]i never meant to give you mushrooms girl[/size]
[FTF]Pyro
Posts: 1233
Joined: Mon Oct 23, 2000 7:00 am

Post by [FTF]Pyro »

DDR RAM is Twice as Fast as SDRAM
This is because DDR RAM transfers data on both the rising and falling edge of the clock whereas SDRAM only transfers data on the rising of the clock. So, DDR RAM transfers twice the amount of data per clock cycle and therefore is referred to as DDR (Double Data Rate).

The RAM industry refers to DDR memory as either "DDR" followed by a clock frequency (ie. DDR333) or "PC" followed by a data transfer rate in MB/sec (ie. PC2700). Also, DDR memory consumes less power than SDRAM at 2.5V instead of 3.3V, which helps to reduce the amount of heat that is produced. This helps extend the battery life in portable devices such as laptop computers.

PC2100 - DDR266 - 266 MHz Bus
Speed is the name of the game in the world of technology. Industry performance standards are always advancing rapidly. Lately, you may have been told that PC100 or PC133 memory is the fastest memory available for your home computer systems, but not surprisingly that has already changed. PC100 and PC133 are ratings for DIMMs (SDRAM). The latest generation of SDRAM is called DDR RAM (Double Data Rate RAM). DDR RAM has twice the bandwidth of standard SDRAM. This boost in performance allows personal computers to move data a whole lot faster. The result is more realistic video rendering, both in the professional graphics design and gaming worlds.

PC2700 - DDR333 - 333 MHz Bus
PC2700 (also know as DDR333) memory is now mainstream and a official JEDEC standard. PC2700 DDR memory modules are available in 256MB and 128MB sizes. This DDR memory is able to run at speeds up to 333 MHz. Only a handful of the latest motherboards support this fast memory. The good thing is that DDR RAM is backwards compatible, so if you are thinking about upgrading in the future, you should buy PC2700 now, because you can run it at lower clock speeds such as 266 Mhz like PC2100 RAM or 200 MHz like PC1600 RAM on your older DDR motherboard, then if you ever upgrade to a PC2700 motherboard you can still use your RAM.

PC3200 - DDR400 - 400 MHz Bus
While not quite an official JEDEC standard, DDR400 motherboards and PC3200 RAM have emerged in the market. These components are marketed towards computer enthuisiasts looking to push the speed limits of their PC's (usually for gaming purposes or bragging rights). However, most complete system PC's being released on the market are still using standard PC2100 DDR RAM which also happens to cost less than the faster DDR333 or DDR400 RAM.

PC3500 - DDR433 - 433 MHz Bus
433 MHz DDR RAM has emerged in the market. Like PC3200, these components are marketed towards computer enthuisiasts looking to push the speed limits of their PC's.
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

[FTF]Pyro wrote:
DDR RAM is Twice as Fast as SDRAM
This is because DDR RAM transfers data on both the rising and falling edge of the clock whereas SDRAM only transfers data on the rising of the clock. So, DDR RAM transfers twice the amount of data per clock cycle and therefore is referred to as DDR (Double Data Rate).
So a machine with 2-2-2-5 Corsair SDRAM running at 400mHz will perform exactly, precisely the same as a machine running 2-2-2-5 DDR at 200mHz, given that all the other hardware is identical?
[FTF]Pyro
Posts: 1233
Joined: Mon Oct 23, 2000 7:00 am

Post by [FTF]Pyro »

R00k wrote:
[FTF]Pyro wrote:
DDR RAM is Twice as Fast as SDRAM
This is because DDR RAM transfers data on both the rising and falling edge of the clock whereas SDRAM only transfers data on the rising of the clock. So, DDR RAM transfers twice the amount of data per clock cycle and therefore is referred to as DDR (Double Data Rate).
So a machine with 2-2-2-5 Corsair SDRAM running at 400mHz will perform exactly, precisely the same as a machine running 2-2-2-5 DDR at 200mHz, given that all the other hardware is identical?
Unless you had signal degridation or check errors but usually ram particularly high performance ram is EEC checked - Error Correction Code modules are an advanced form of Parity detection often used in servers and critical data applications. ECC modules use multiple Parity bits per byte (usually 3) to detect double-bit errors. They also will correct single-bit errors without creating an error message. Some systems which support ECC can use a regular Parity module by using the Parity bits to make up the ECC code. However, a Parity system cannot use a true ECC module.


But even if they arent the average failure on ram is once in 6months causing a reboot. No big deal.


But in short. the way the waves move is that you have a 1 bit on +1 and 1 bit on - 1 as far as I am led to believe
shadd_
Posts: 2512
Joined: Tue Jun 28, 2005 4:02 pm

Post by shadd_ »

this seems to be the best price/performer deal i've seen.

http://www.us.ncix.com/products/index.p ... omoid=1001
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

[FTF]Pyro wrote:Unless you had signal degridation or check errors but usually ram particularly high performance ram is EEC checked - Error Correction Code modules are an advanced form of Parity detection often used in servers and critical data applications. ECC modules use multiple Parity bits per byte (usually 3) to detect double-bit errors. They also will correct single-bit errors without creating an error message. Some systems which support ECC can use a regular Parity module by using the Parity bits to make up the ECC code. However, a Parity system cannot use a true ECC module.


But even if they arent the average failure on ram is once in 6months causing a reboot. No big deal.


But in short. the way the waves move is that you have a 1 bit on +1 and 1 bit on - 1 as far as I am led to believe
What I'm saying is that, unless there have been tests proving that DDR performs precisely the same as SDRAM at double the speed, then the benches in this study are not valid.

I don't know if that has been proven or not, I was hoping someone else might.
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

shadd_ wrote:this seems to be the best price/performer deal i've seen.

http://www.us.ncix.com/products/index.p ... omoid=1001
Thanks man, that's exactly what I've been shopping for. I've been told that OCZ has better deals with comparable performance, but so far I wasn't able to find the price differences everyone was talking about. :icon14:

edit: I even found it 7 bucks cheaper at the place I was planning on buying the rest of my gear, since I had the part number:
http://www.monarchcomputer.com/Merchant ... ode=140226
[FTF]Pyro
Posts: 1233
Joined: Mon Oct 23, 2000 7:00 am

Post by [FTF]Pyro »

R00k wrote:
[FTF]Pyro wrote:Unless you had signal degridation or check errors but usually ram particularly high performance ram is EEC checked - Error Correction Code modules are an advanced form of Parity detection often used in servers and critical data applications. ECC modules use multiple Parity bits per byte (usually 3) to detect double-bit errors. They also will correct single-bit errors without creating an error message. Some systems which support ECC can use a regular Parity module by using the Parity bits to make up the ECC code. However, a Parity system cannot use a true ECC module.


But even if they arent the average failure on ram is once in 6months causing a reboot. No big deal.


But in short. the way the waves move is that you have a 1 bit on +1 and 1 bit on - 1 as far as I am led to believe
What I'm saying is that, unless there have been tests proving that DDR performs precisely the same as SDRAM at double the speed, then the benches in this study are not valid.

I don't know if that has been proven or not, I was hoping someone else might.


what you are saying is AMD running at 2.2 ghz but marked 3600+ shouldnt be compared to a PIV3600 as the "actual" clock speed is 2.2GHz
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

No, that's not what I'm saying at all. We already know that AMD processors perform differently than similarly clocked Intels, in several different ways. So they don't run benchmarks using one Intel and one AMD processor, unless the processor itself is what they are comparing.
[FTF]Pyro
Posts: 1233
Joined: Mon Oct 23, 2000 7:00 am

Post by [FTF]Pyro »

R00k wrote:No, that's not what I'm saying at all. We already know that AMD processors perform differently than similarly clocked Intels, in several different ways.
Bah I dont have a clue what you are looking for here...... I mean DDR double data rate.


1000mhz = 500 mhz ddr


If it doesnt perform exactly the same as SDRAM at 1000 mhz then its got to be 99.999% as good as. If you ask me. your being slightly pedantic unless of course this is for some sort of project that will prove something.
AmIdYfReAk
Posts: 6926
Joined: Thu Feb 10, 2000 8:00 am

Post by AmIdYfReAk »

R00k wrote: unless the processor itself is what they are comparing.
bingo..

what are they going to do? call it:

AMD 64 3600+WHENRUNNINGDDR2WITHNFORCE5BASEDMOBO

Cmon :)
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

Are you agreeing with me or trying to tell me something? :paranoid:
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

[FTF]Pyro wrote:
R00k wrote:No, that's not what I'm saying at all. We already know that AMD processors perform differently than similarly clocked Intels, in several different ways.
Bah I dont have a clue what you are looking for here...... I mean DDR double data rate.


1000mhz = 500 mhz ddr


If it doesnt perform exactly the same as SDRAM at 1000 mhz then its got to be 99.999% as good as. If you ask me. your being slightly pedantic unless of course this is for some sort of project that will prove something.
Well yea, it's slightly pedantic, but this is a benchmark we're talking about here. :)

In fact, if no one has proved that they are exactly the same, then that should be one of the caveats in their review/comparison, along with the reduced pipes and such.
AmIdYfReAk
Posts: 6926
Joined: Thu Feb 10, 2000 8:00 am

Post by AmIdYfReAk »

I'm saying that they compare the CPU itsself Vs eachtother, Rather then There setups ( I.e. What Ram they are using and etc )
R00k
Posts: 15188
Joined: Mon Dec 18, 2000 8:00 am

Post by R00k »

Yes - as long as they are each using exactly the same ram. :)
shadd_
Posts: 2512
Joined: Tue Jun 28, 2005 4:02 pm

Post by shadd_ »

if thats the exact same stuff as the ncix link i would make sure i bought it before the 17th it says.
[size=75]i never meant to give you mushrooms girl[/size]
[FTF]Pyro
Posts: 1233
Joined: Mon Oct 23, 2000 7:00 am

Post by [FTF]Pyro »

R00k wrote: Well yea, it's slightly pedantic, but this is a benchmark we're talking about here. :)
.
....... I stopped listening after this point cause in reality you'll never have the same setup regards of exact parts or specifications.


Reason. A whole multitude of reasons


Power Fluctuation
Windows memory cycles
Power Drains
Hard disk caching
Analogue digital convertors
hard disk warm up speed
system tempretures
act of god
etc
Post Reply