View Full Version : How to calculate RDRAM and DDR SDRAM performance
Ghetto_Child
Apr 9th, 2006, 03:04 AM
Ok I'm getting confused lately so I need some help to be sure of the specifics.
PC800 RDRAM does 3200MB/s because it is running at 400MHz DDR which equals 800MHz actual and each module is 16bits or 2bytes and because it's dual channel then the 2 pairs are sending a total of 32bits or 4 bytes right?
So how does DDR SDRAM work? say PC3200, does that run at 266MHz or 333MHz? Then does it send 32bits twice because of Dual Channel meaning 64bits actual? Does the speed double because of Double Data Rate? I'm asking because a friend has PC3200 ram for a board that is supposed to handle PC3200 ram but for some reason the only bus speeds selectable is 266MHz is this normal and the actual performance is 533MHz. How does one calculate the details please?
Thanks
ProfessorChaos
Apr 9th, 2006, 03:29 AM
Ok I'm getting confused lately so I need some help to be sure of the specifics.
PC800 RDRAM does 3200MB/s because it is running at 400MHz DDR which equals 800MHz actual and each module is 16bits or 2bytes and because it's dual channel then the 2 pairs are sending a total of 32bits or 4 bytes right?
So how does DDR SDRAM work? say PC3200, does that run at 266MHz or 333MHz? Then does it send 32bits twice because of Dual Channel meaning 64bits actual? Does the speed double because of Double Data Rate? I'm asking because a friend has PC3200 ram for a board that is supposed to handle PC3200 ram but for some reason the only bus speeds selectable is 266MHz is this normal and the actual performance is 533MHz. How does one calculate the details please?
Thanks
DDR PC3200 runs at 200MHz and is dual data rate so it is essentially 400MHz aka DDR400...
the bus speed of 266 is probably because he is either mixing the ram modules with slower modules...or his ram cannot run DDR400 speed on that particular motherboard.
tdotcbc84
Apr 9th, 2006, 03:36 AM
hehehhee
i remember i had a simliar post like this awhile back !
umm wat the guy said above basically :)
Ghetto_Child
Apr 9th, 2006, 04:52 AM
DDR PC3200 runs at 200MHz and is dual data rate so it is essentially 400MHz aka DDR400...
the bus speed of 266 is probably because he is either mixing the ram modules with slower modules...or his ram cannot run DDR400 speed on that particular motherboard.
Alright so DDR PC3200 is actually running at 200MHz or 400MHz effective and it sends data in 64bit chunks or 8Byte chunks? So a single DDR module is 64bits and a single RDRAM module is 16bits?. Wouldn't that make the effective latency of RDRAM better/small/faster than DDR SDRAM at equal data rate :confused: . If that's true then DDR has been playing catch up for 5yrs :lol:
Should my friend's board say 200Mhz or is it really supposed to say 400MHz for these particular modules? Does it make sense that the speed settings listed (either in the bios options or at POST boot up) will be the actual speed or the effective speed?
willy
Apr 9th, 2006, 07:07 AM
Should my friend's board say 200Mhz or is it really supposed to say 400MHz for these particular modules? Does it make sense that the speed settings listed (either in the bios options or at POST boot up) will be the actual speed or the effective speed?
Download and run (no installation required) CPU-Z. Post the info on the "CPU" and "Memory" tabs please.
http://www.cpuid.com/download/cpu-z-133.zip
http://forums.ocmodshop.com/attach.aspx/358/25%20250%20cpuz.jpg
JAC
Apr 9th, 2006, 10:19 PM
Heh, I looked at the thread title and thought, "Who bumped this ancient post?"
Then I looked at the date and thought, "Who still gives a **** about RDRAM?"
goofball
Apr 9th, 2006, 11:28 PM
PC3200 = 200mhz, multiplied by 64 bit, divided by 8 bits, multiplied by 2 = 3200 MB/sec.
64 bit = width of the memory bus.
divide by 8 bits, because 1 byte = 8 bits. To convert to bytes.
Multiplied by 2 because DDR sends 1 piece of data on the rising and falling part of a clock cycle. It's not actually making the signal faster, it just sends 2 pieces of data/clock cycle so it's doubling the amount of data sent per clock cycle.
CompWizrd
Apr 10th, 2006, 08:50 AM
Wouldn't that make the effective latency of RDRAM better/small/faster than DDR SDRAM at equal data rate :confused: . If that's true then DDR has been playing catch up for 5yrs :lol:
Yes. And DDR2 has even higher latency.
However, you can buy a motherboard and cpu for the cost of upgrading the ram on a rdram system.. and then there's the ethics of dealing with such a shady company as rambus.
ShadowVlican
Apr 10th, 2006, 10:55 AM
Heh, I looked at the thread title and thought, "Who bumped this ancient post?"
Then I looked at the date and thought, "Who still gives a ***** about RDRAM?"
my thoughts exactly
Ghetto_Child
Apr 11th, 2006, 07:41 AM
Yes. And DDR2 has even higher latency.
However, you can buy a motherboard and cpu for the cost of upgrading the ram on a rdram system.. and then there's the ethics of dealing with such a shady company as rambus.
Why would Rambus be a shady company if they actually made something 5yrs more advanced than anything else?
Heh, I looked at the thread title and thought, "Who bumped this ancient post?"
Then I looked at the date and thought, "Who still gives a **** about RDRAM?"
Because the price you all pay for DDR and DDR2 systems is wasted because of the bandwidth/$ compared to old Rambus boards. Sure you can't get cpu speeds like they have now but that's only because the companies designed it that way. If they wanted to they could push everyone to RDRAM boards with the PCI-Express bus and LGA775 cpus AND since 2002/2003 PC1066(4.2GB/s) and PC1200(4.8GB/s) was already produced with asus motherboards as the pilot. The problem was everyone got scared of paying twice the money for 2-4times the performance of DDR. The loonacy of people trying to save a loon :rolleyes: , go figure :lol: .
In fact running the exact same programs I have. My P4 1.4 PC800 RDRAM system runs the programs and windows faster than my school's 2.4GHz celeron 533MHz bus with DDR 333MHz. Both systems have the same amount of ram.
CompWizrd
Apr 11th, 2006, 08:30 AM
Why would Rambus be a shady company if they actually made something 5yrs more advanced than anything else?
They were a member of JEDEC, and as part of being in that membership, they were supposed to reveal what patents they held on memory.
Instead, they pushed a certain method of building ram, and secretly took a patent on the method. That way, when it was ratified as a standard, they could go after other manufacturers for royalty payments.
Because the price you all pay for DDR and DDR2 systems is wasted because of the bandwidth/$ compared to old Rambus boards. Sure you can't get cpu
The "advantage" of rambus was low latency, not the high transfer rate. The P4 was limited to 100 mhz FSB(400 quad pumped) back then, which is pc1600, or 1.6 gig/second.
In fact running the exact same programs I have. My P4 1.4 PC800 RDRAM system runs the programs and windows faster than my school's 2.4GHz celeron 533MHz bus with DDR 333MHz. Both systems have the same amount of ram.
Same hard drive, same cpu type? Didn't think so.
Rambus got screwed in the market because of their prices, and predatory practices against other memory manufacturers. These other manufacturers retaliated by selling their DDR ram as cheap as they could, effectively pushing RIMM's out of the market.
Ghetto_Child
Apr 11th, 2006, 09:50 AM
They were a member of JEDEC, and as part of being in that membership, they were supposed to reveal what patents they held on memory.
Instead, they pushed a certain method of building ram, and secretly took a patent on the method. That way, when it was ratified as a standard, they could go after other manufacturers for royalty payments.
Oh that is shady sounding. For once the asians actually do something shady to get ahead in the corporate world. They must have learned from Micro$oft's practice of Micro$hit :lol: . However for the performance I get with rambus and the price of rambus ram on ebay and how cheap the boards are too. I'll stick to that unless I require PCI-Express, ATI Crossfire, or nVIDIA SLI. Now want to see the next performance king. Look up the Intel Conroe core, it's apparently smoking even an overclocked Athlon64 X2.
The "advantage" of rambus was low latency, not the high transfer rate. The P4 was limited to 100 mhz FSB(400 quad pumped) back then, which is pc1600, or 1.6 gig/second.
Not sure where you get the 1.6 gig/sec from. PC600 ram was 2.4GB/s and PC800 is 3.2GB/s. If you're saying that the P4 cpu was limited to 1.6GHz that was only when the line first came out (1.2, 1.3, 1.4, 1.5, 1.7). Cpu clock cycles doesn't really have anything to do with the bandwidth it can handle. I mean 64bits multiplied by 1.4GHz seems a little out worldly. CPUs use shorthand code for repeated data so if it has to send 1GB of data repeatedly it will spend time sending the first 1GB and use a repeat instruction for the sequential identical 1GB of data instead of sending 1GB over and over. Also cpu performance is not based on clock cycles. It's the product of the number of clock cycles per second multiplied by the maximum number of instructions it can complete in either one clock cycle or in one second.
I think I see where you get the 1.6gig/sec. You mean the bus bandwidth from cpu to ram. True it probably is just 1.6GB/s but the ram kept a simultaneous independant link with the AGP bus also. So at any time both the AGP bus and cpu can receive 1.6GB/s each.
Same hard drive, same cpu type? Didn't think so.
Rambus got screwed in the market because of their prices, and predatory practices against other memory manufacturers. These other manufacturers retaliated by selling their DDR ram as cheap as they could, effectively pushing RIMM's out of the market.
Yes same hard drives too but as I stated before the cpus were different. My IBM P4 1.4GHz 400MHz bus Willamette & PC800 RDRAM vs. Celeron 2.4GHz Presscott 533MHz bus DDR 333MHz Dell. My 1.4GHz RDRAM system is faster. See I have a multiboot on the same hard drive so I just carry my boot drive from system to system and on my P4 1.4GHz everything loads faster more programs open at the same time upon startup and the mouse clicks are responded to faster on my RDRAM system. Running the same things on the Dell (and I have several identical dell systems at my school) takes way longer and the lag between mouse click and load while windows is still booting up is terrible.
matkun
Apr 11th, 2006, 03:59 PM
As a note, Sisoft Sandra has just clocked my DDR400 Ram, (running at 230 Mhz, so in reality running at 460) at ~5,200 MB/sec.
Theoretical limit on DDR400 Ram is actualy 6.4 GB /sec. Each DDR Channel is 64 bits wide, and the dual channel is a total of 128 bits wide.
Thus your numbers for DDR Ram are half of what they should be.
So no, DDR has not been playing catch up, and you would need 'PC1600' RDRAM to have equal bandwidth.
I do not know about the latency calculations, but the Bandwidth calculations are correct.
If you doubt the 64 bit and 128 bit facts, then this Intel Whitepaper will set you straight:
http://www.kingston.com/newtech/MKF_520DDRwhitepaper.pdf
As for my numbers, yes, it's about 71% efficiency (DDR 460 max should be ~7360 MB/sec), but it is running four 512 MB modules, with 2-3-3-7 timings and 2T command rate.
Also, to add, in your PC Comparison... your comparing a P4 to a Celeron.. I would actually believe that a 1.4 GHz P4 is faster then a 2.4 GHz Celeron, since the Celerons are POS ultimate budget chips.
CompWizrd
Apr 11th, 2006, 06:44 PM
Also, to add, in your PC Comparison... your comparing a P4 to a Celeron.. I would actually believe that a 1.4 GHz P4 is faster then a 2.4 GHz Celeron, since the Celerons are POS ultimate budget chips.I suspect the video card drivers are the default that shipped with Windows as well, which can make a HUGE difference.
vBulletin® v3.8.4, Copyright ©2000-2009, Jelsoft Enterprises Ltd.