Upcoming 533 MHz FSB - Conflicting Information

From what I understand, there will be two versions of "PC1066". One will be called PC1066 and simply be a higher clocked version of what we now have. This upgrade will most likely require a BIOS flash at the most.

The other version is RIMM4200 (unsure of what the release name will actually be), and will be single-channel 32-bit modules providing the same amount of bandwidth.

<font color=blue>If you don't buy Windows, then the terrorists have already won!</font color=blue> - Microsoft

Hmm, any word on latency? One would think RIMM4200 has better latency than PC1066?

:wink: <b><i>"A penny saved is a penny earned!"</i></b> :wink:

Why would you say that? One of the reasons for dual-channel in the first place was that it decreases latency.

<font color=blue>If you don't buy Windows, then the terrorists have already won!</font color=blue> - Microsoft

Hmm, I don't know, but I would still like to see a latency comparison between the two.

:wink: <b><i>"A penny saved is a penny earned!"</i></b> :wink:

Me too

<font color=blue>If you don't buy Windows, then the terrorists have already won!</font color=blue> - Microsoft

Currently only boards that support overclocking will support the 533 bus, OEM boards such as Intel's will not, until the release of the i850E versions.

What's the frequency, Kenneth?

So can you run PC800 RDRAM with 533 MHz FSB CPU ?

Yes, the ABit TH7-II supports that.

:wink: <b><i>"A penny saved is a penny earned!"</i></b> :wink:

My guess is that the PC1066 chip will have a higher latency than a RIMM4200 chip, but dual channels for PC1066 will make up for it and excede it.

Depends on how they engeneered the RIMM4200 though. If it's simply 2 16-Bit RIMMs effectively pasted togeather onto one chip, they it will effectively have the same performance.

The Windows Gods demand money to appease the BSOD! - Rev. Bill Gates

Bront has it right: for the time being, RIMM4200 modules are essentially 2 single-sided RIMMs back to back. They are not "single-channel" 32bit they are dual-channel 16bit packaged onto a single 32bit RIMM. Yes, we could argue the sematics of dual or single channel, but the chips or "devices" are 16bit and are identical to the devices on the 16bit modules. The latency of one 32bit module, therefore, will be identical to installing 2 16bit RIMMs with the same chips.

If the thought I thought I thought had been the thought I thought, I wouldn't have thought so much.

Would be interesting if it was possible to Dual Channel those RIMMS with another stick, 8.4GB/sec of bandwidth, nice.

--
Thunderbirds in wintertime, Northwoods in summertime!  

Half of which would be unused, but it would be nice to say I had 8.4GB/s of bandwidth :cool:

<font color=blue>If you don't buy Windows, then the terrorists have already won!</font color=blue> - Microsoft

Hmm... since latency increases with the addition of more RAM in a RDRAM system, wouldn't having only half the RIMMs reduce latency by about the same about that a dual-channel would have? Just conjecture on my part, really.

-SammyBoy