TR Forums

When I hear the name "Gelsinger" I think of words like Itanium, Netburst, and Rambus.

Is that unfair? Or is Intel now more doomed than ever?

Rambus was excellent tech...just too early for its time and far too expensive for the gains to be had.

Rambus the product was just overpriced.
Rambus the company were some underhanded scum-sucking patent trolls.

Agreed 100%.

JustAnEngineer wrote:

Rambus the product was just overpriced.
Rambus the company were some underhanded scum-sucking patent trolls.

Like Netburst (turn bandwidth up to 11; completely ignore IPC), it was the wrong solution to the market problem trying to be solved.

It was a marketing solution to a technical problem. It worked well in certain niche scenarios where bandwidth was more important than latency, but for very general-purpose solutions, it was a solution in search of a problem.

Waco wrote:

Rambus was excellent tech...just too early for its time and far too expensive for the gains to be had.

I'd like to hear more about this perspective...

My recollection is that DDR was the excellent part, and obviously that's been widely adopted. But I seem to recall that RDRAM had a narrow bus (16 bits or 32 bits) that ran at a high clock speed, resulting in terrible performance/watt. Basically the opposite of HBM.

But high clock speeds, but with RDRAM and Netburst, fit Intel's (and really, the whole industry's) late 90s marketing narrative. Gelsinger brought us the 486, which was a fine product but also accelerated the marketing of clock rates.

So I'm just wondering... what was Gelsinger's role in pushing Intel down the RDRAM/Netburst path?

I'm trying to figure out if his return is good news for Intel or not...

IIRC, the real issue with RAMBUS was they submitted a pile of technology and got them adopted as standards on the Intel-based motherboards of the time, then just after those technologies were accepted they announced that they had patents on all of them and started charging an arm and a leg to use them. I don't recall any technical problems with their tech, just that they used some underhanded methods to try to turn it into a cash cow by creating a captive market.

I think I'm mixing up RDRAM and XDR RAM in my head. The latter is fast as hell but mired in licensing ****.

The narrative, largely rehashed in the thread above, was that Rambus conspired with Intel to take over the memory industry via Intel's Netburst architecture, and submarined patents behind JEDEC's back to subordinate the traditional DRAM makers. But there were some articles about Rambus and the JEDEC fracas, largely written after the fact, indicating that it wasn't necessarily that Rambus was a bunch of patent trolls, but rather that the rest of the industry had colluded against them to try and prevent Intel from gaining control over the dominant RAM standards, and actually did appropriate legitimate Rambus IP in the DDR standard. Some of Rambus' supposed patent trolling was arguably defensive and retaliatory. All things considered there were no good guys in the memory industry at that time.

The irony is that a couple generations later, Intel was using DDR2 FB-DIMMs on its Xeon workstations for a couple years -- which were, for all practical purposes, Rambus RIMMs parked on a DDR2 interface.

This sounds positive:

https://www.anandtech.com/show/16438/ne ... architects