The Inquirer-Home

Rambus memory thrashes DDR, and it's better with hyperthreading

Granite Bay displaying doggish, piggish DDR tendencies
Sun Dec 22 2002, 13:14
AS OUR OWN Chicago savant says, the only justice you can expect in this life is to die. A life sentence means one strike and you're out, unless you believe in transmigration.

Inquirer-roadmap-symbol----which-memory-type-is-the-two-wheeler--and-which-the-one-with-better-horsepower-So just a few weeks away from when Rambus RDRAM memory looks set to die for the PC platform, it's somewhat ironical and, perhaps, unjust that reviews from three hardware web sites have concluded that the 850E performs a heck of a sight better than Intel's new DDR based Granite Bay (E7205) chipset.

Mind you, information posted here at the INQUIRER shows that Granite Bay only has a half life of around six weeks itself, before Intel swerves again with its upcoming Canterwood chipset.

A review at Ace's Hardware said that Granite Bay needs fast DDR 333 running at DDR 266 2-2-2 just to keep up with the i850e chipset, while it's more expensive too. Plus, although dual DDR 266 is faster by around 15 per cent at reading long datastreams, Rambus are up to 60% more efficient at handling writes and interspersed reads and writes. So, the Rambus chipset offers more bandwidth.

Michael, at Lost Circuits , gives Granite Bay a gentle sort of rough ride too. If HT (hyperthreading) is enabled, he reckons you need 2GB of system memory to note lose performance.

And venerable InfoWorld agrees with this assessment too. In this review, it claims systems using Rambus RDRAM perform better on hyperthreading than that old new DDR SDRAM stuff.

It says: "We also found that Hyper-Threading gains are greater when using Rambus DRAM than when using DDR (Double Data Rate) SDRAM."

Of course, Intel itself has the whole world+dog now scrabbling around after it introduced cunning plans for fast DDR 400 which, because of its market position, looks set to be the "standard" this coming year.

So where is Rambus right now? The firm is waiting for courts of appeals to release their verdicts, and a lot hangs on those decisions. Those verdicts are likely in weeks, or perhaps months.

The Federal Trade Commission is also investigating Rambus but that will take a while before it's decided, while, somewhat ironically, the Department of Justice (DoJ) is investigating Micron.

It turns out - judging from the reviews above, that Rambus and Intel were both right about the technological advantages of RMBS RDRAM. But that was never really the point, was it?

The rest of the industry was determined it shouldn't have to pay these pesky royalties and while Intel continued bolstering Rambus, inside the corporation - we know on very good authority, wiser counsels were hearing what the vendors and other partners were muttering, and were attempting to get top management to stop being so "pig headed" about persisting in support for RDRAM. Market realities appear to have prevailed inside Intel.

One source inside the corp told us that if it had listened a little bit earlier, and applied the right persuasive techniques, by now Rambus RDRAM would have been the pervasive PC standard.

As Dataquest senior semiconductor analyst Andrew Norwood never tires of saying: "Nothing moves the memory market like FUD (fear, uncertainty, doubt)"... µ

See Also
Intel-Rambus love affair reaches end of the roadmap


Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?