DDR2 800 vs DDR2 1066


Recommended Posts

Ok my friend wants to spend around $3000 for a new computer. I was wondering how much of an advantage getting DDR2 1066 and running it in a 1:1 ratio with the FSB would make over having DDR2 800? We will be using an E6600.

Also, would getting 2 7900GTs instead of 1 7950GX2 be the way to go now with the driver issues and such? Im sure a little OCing to the 7900s and it would beat the 7950GX2.

I will post pics and benchmarks whenever we do start on this project though! Thanks for any info!

Link to comment
https://www.neowin.net/forum/topic/486200-ddr2-800-vs-ddr2-1066/
Share on other sites

  bigdawgfoxx said:

Ok my friend wants to spend around $3000 for a new computer. I was wondering how much of an advantage getting DDR2 1066 and running it in a 1:1 ratio with the FSB would make over having DDR2 800? We will be using an E6600.

Also, would getting 2 7900GTs instead of 1 7950GX2 be the way to go now with the driver issues and such? Im sure a little OCing to the 7900s and it would beat the 7950GX2.

I will post pics and benchmarks whenever we do start on this project though! Thanks for any info!

to run DDR2 1066 at a 1:1 ratio with the E6600, you would have to have the CPU clocked at 4.97Ghz (which unless your gonna be using LN2, doesn't seem likely to achieve).

Core 2 Duo uses a real FSB of 266Mhz. The 1066 figure is the effective speed of the FSB as it is quad pumped (can send data 4 times each clock cycle).

Since it only uses an FSB of 266Mhz, to run at stock speeds, you only need DDR2-533 (2x266). However, if your overclocking, go for DDR2 667, or DDR2 800. If you do end up ruinning an FSB of over 400Mhz (3.6Ghz for E6600), then you could just overclock the RAM.

  • 1 year later...
  Jdawg683 said:
when are the next-gen cards released? supposedly by the end of the year, but we'll see.

Ok guys, you're scarin' me :) I'm an old guy that has always built my own pc. It's been awhile since I've rebuilt. To give you an indication of how long, I have the 6800 Ultra and a SCSI 320 drive force fed into the box with an adapter which knocks it down to 160. Anyway, I have ordered parts to rebuild again and wanted to bounce it off you guys since my last search brought me here. I was trying to figure out if I would do just as well with 4G of DDR2-800 as 2G of DDR2-1066. I need to read more because I haven't nailed down what I keep seeing about the 1066 vs 800 vs the FSB of the mobo vs the memory on the 8800GTX etc. etc. If anyone can spit me out a 101 on that subject, it would be great. Thanks

This is my intended setup:

Asus P5K

E6850 3.0Ghz

2 GB DDR2-1066 CL5 Crucial Ballistix (2 1024)

150gb 10k Raptor

8800GTX 768mb

WelshBluebird explained it perfectly and all the numbers even apply to your situation.

However, since you asked, I'll explain it again...I'll try to do it differently.

E6850 has a stock speed of 3.0 GHz, which means it's 333x9 = 2997 MHz. With DDR2-800 you can get to 400x9 = 3600 MHz without overclocking the RAM, with DDR2-1066 you can get to 533x9 = 4797 MHz without overclocking the RAM.

However, 4.8 GHz will require a hefty amount of cooling, something in the range of high end water, peltier, or phase...and if you had those you probably wouldn't be posting :p

If you're confused about the Frontside buses ( 1066 MHz and 1333 MHz ) they're all related to the other FSB ( FSB x Multiplier = clock speed ). Intel systems have quad pumped frontside buses, so 1066 MHz chips have FSB of 266 MHz and 1333 MHz chips have FSB of 333 MHz. These are all stock speeds.

None of it is related to the video card.

Also, Nvidia is supposed to be getting ready to release their 2nd generation DX10 cards in December ( core name is G90/G92 ).

  Wavelength1550 said:
Ok guys, you're scarin' me :) I'm an old guy that has always built my own pc. It's been awhile since I've rebuilt. To give you an indication of how long, I have the 6800 Ultra and a SCSI 320 drive force fed into the box with an adapter which knocks it down to 160. Anyway, I have ordered parts to rebuild again and wanted to bounce it off you guys since my last search brought me here. I was trying to figure out if I would do just as well with 4G of DDR2-800 as 2G of DDR2-1066. I need to read more because I haven't nailed down what I keep seeing about the 1066 vs 800 vs the FSB of the mobo vs the memory on the 8800GTX etc. etc. If anyone can spit me out a 101 on that subject, it would be great. Thanks

This is my intended setup:

Asus P5K

E6850 3.0Ghz

2 GB DDR2-1066 CL5 Crucial Ballistix (2 1024)

150gb 10k Raptor

8800GTX 768mb

That's fairly similar to what I have now, and it works pretty good. Having 1:1 memory isn't that important though, and the speed doesn't seem to matter too much either. I borrowed a stick of DDR2-800 1 GB from my friend yesterday (dropping all 3 sticks to that speed) and I don't notice any difference in FPS when playing CSS. Oh and I have 3x1 GB because I don't feel like dealing with Windows x64 to get 4 GB to work.

  Cephas said:
That's fairly similar to what I have now, and it works pretty good. Having 1:1 memory isn't that important though, and the speed doesn't seem to matter too much either. I borrowed a stick of DDR2-800 1 GB from my friend yesterday (dropping all 3 sticks to that speed) and I don't notice any difference in FPS when playing CSS.

RAM downclocks by itself perfectly fine, I don't see why you'd even see a performance difference because your RAM is underclocked. The only thing you could worry about is possibly having to run a divider because your RAM can't match your FSB speed, but even then dividers hardly have a performance impact.

  bigdawgfoxx said:
Also, would getting 2 7900GTs

If you're considering two 7900s, and unless you got a good deal on them, why not just get a single 8800? You don't have to use it only for DX10. The video playback quality is a little better as well.

In general, multiple cards will always cause more problems than a single card, so you should avoid that kind of situation if possible.

  gwai lo said:
WelshBluebird explained it perfectly and all the numbers even apply to your situation.

However, since you asked, I'll explain it again...I'll try to do it differently.

E6850 has a stock speed of 3.0 GHz, which means it's 333x9 = 2997 MHz. With DDR2-800 you can get to 400x9 = 3600 MHz without overclocking the RAM, with DDR2-1066 you can get to 533x9 = 4797 MHz without overclocking the RAM.

However, 4.8 GHz will require a hefty amount of cooling, something in the range of high end water, peltier, or phase...and if you had those you probably wouldn't be posting :p

If you're confused about the Frontside buses ( 1066 MHz and 1333 MHz ) they're all related to the other FSB ( FSB x Multiplier = clock speed ). Intel systems have quad pumped frontside buses, so 1066 MHz chips have FSB of 266 MHz and 1333 MHz chips have FSB of 333 MHz. These are all stock speeds.

None of it is related to the video card.

Also, Nvidia is supposed to be getting ready to release their 2nd generation DX10 cards in December ( core name is G90/G92 ).

Thank you gwai lo and cephas for your responses. I went back to Welshbluebird's post and read it and I think I understand what you guys are telling me. I apologize for asking you to cover ground again that you've already covered many times I'm sure, I am just an extremely busy fellow and this is very very helpful to me in the interest of time. I'm busy out there laying down fiber so I can ask you guys these questions :)

What I've gathered from your post and Welshbluebirds post is that my e6850 runs at 333 quad pumped for the 1333 as opposed to 266 on the 6600 for 1066 welsh spoke to in his earlier post. That said, I would only need ddr2-667 to run at stock speeds (333x2). With ddr2-800 I would have a little room to oc the processor to 400 yielding 3.6 GHz. Would that require additional cooling?

All that said, it sounds like the 1066 chips are overkill and I don't need them unless I plan on going nuts with cooling and oc'ing. Would there be any future benefit to having the 1066 chips now?

I appreciate your statement Cephas about the 3 each 1G chips vs. 4G and 64 bit windows. I'll keep that in mind.

Yep, got it perfectly. You might be able to survive on stock, but it's not something I'd try. You can, just watch your temps.

As for benefits of DDR2-1066, hardly anything beyond overclocking headroom. You can run a divider so that your RAM is two times faster than your FSB ( it's a 2:1 divider ), but that hasn't been demonstrated to net any real performance.

  gwai lo said:
Yep, got it perfectly. You might be able to survive on stock, but it's not something I'd try. You can, just watch your temps.

As for benefits of DDR2-1066, hardly anything beyond overclocking headroom. You can run a divider so that your RAM is two times faster than your FSB ( it's a 2:1 divider ), but that hasn't been demonstrated to net any real performance.

again, thank you for your help, you are most kind.

  • 1 year later...
This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Posts

    • inception horn at max volume every time you copy.
    • Doesn't Windows have I build clipboard history? Win+V
    • Quantum computer does something for first time, creates "certified truly random" numbers by Sayan Sen Image by Ron Lach via Pexels A group of researchers from JPMorganChase, Quantinuum, Argonne National Laboratory, Oak Ridge National Laboratory, and The University of Texas at Austin has reached a major milestone in quantum computing. In a new paper published in Nature, they describe how they used a 56-qubit quantum computer to generate random numbers and then proved those numbers were truly random using powerful classical supercomputers. This achievement, called "certified randomness," could be useful in areas like cryptography, privacy, and fairness. Certified randomness means the numbers are not just unpredictable, but also freshly created and mathematically verified. Classical computers can’t do this on their own. They usually rely on hardware random-number generators, which can be tampered with. But with this new method, even if someone tried to interfere with the quantum computer, they wouldn’t be able to fake the randomness and still pass the certification. The idea behind the protocol was first proposed by Scott Aaronson, a computer science professor at UT Austin. He worked with his former postdoctoral researcher, Shih-Han Hung, to support the experimental team. “When I first proposed my certified randomness protocol in 2018, I had no idea how long I’d need to wait to see an experimental demonstration of it,” Aaronson said. “Building upon the original protocol and realizing it is a first step toward using quantum computers to generate certified random bits for actual cryptographic applications.” To carry out the experiment, the team accessed Quantinuum’s System Model H2-1 quantum computer over the internet. They used a method called random circuit sampling (RCS), which is known to be extremely hard for classical computers to simulate. The process had two main steps. First, the researchers sent the quantum computer a series of challenge circuits created from a small seed of randomness. The quantum computer had to solve these challenges by picking one of many possible answers at random. Then, in the second step, classical supercomputers checked the results to confirm the randomness was real. The team used several supercomputers with a combined performance of 1.1 × 10¹⁸ floating-point operations per second (1.1 ExaFLOPS) to certify 71,313 bits of entropy. This means they proved that the random bits couldn’t have been generated by classical methods, at least not under realistic conditions and assumptions. “This work marks a major milestone in quantum computing, demonstrating a solution to a real-world challenge using a quantum computer beyond the capabilities of classical supercomputers today,” said Marco Pistoia, Head of Global Technology Applied Research at JPMorganChase. Quantinuum had upgraded its H2 system to 56 qubits in June 2024. Thanks to its high accuracy and ability for any qubit to connect with any other, the system was able to perform RCS far better than previous machines. This upgrade, combined with Aaronson’s protocol, made the breakthrough possible. “Today, we celebrate a pivotal milestone that brings quantum computing firmly into the realm of practical, real-world applications,” said Dr. Rajeeb Hazra, President and CEO of Quantinuum. “These results in quantum computing were enabled by the world-leading U.S. Department of Energy computing facilities at Oak Ridge National Laboratory, Argonne National Laboratory and Lawrence Berkeley National Laboratory,” said Travis Humble, director of the Quantum Computing User Program at ORNL. While quantum computers have shown they can outperform classical ones in theory, turning that power into something useful has been a challenge. This experiment shows that quantum computers can now do something practical that classical computers simply can’t match. Source: University of Texas, Nature This article was generated with some help from AI and reviewed by an editor. Under Section 107 of the Copyright Act 1976, this material is used for the purpose of news reporting. Fair use is a use permitted by copyright statute that might otherwise be infringing.
  • Recent Achievements

    • First Post
      EzraNougat earned a badge
      First Post
    • One Month Later
      westDvina earned a badge
      One Month Later
    • Community Regular
      Bern@rd went up a rank
      Community Regular
    • Week One Done
      Joey Solo earned a badge
      Week One Done
    • Week One Done
      Vicente C Alves earned a badge
      Week One Done
  • Popular Contributors

    1. 1
      +primortal
      711
    2. 2
      +FloatingFatMan
      201
    3. 3
      ATLien_0
      163
    4. 4
      Xenon
      130
    5. 5
      wakjak
      110
  • Tell a friend

    Love Neowin? Tell a friend!