[H]ardocp (Kyle Bennett) thoughts on ATI/AMD, Intel, and Nvidia.

But how can they get into that sector if their chipsets are "buggy", I thought the server market stuff needed to be rock stable.
 
But how can they get into that sector if their chipsets are "buggy", I thought the server market stuff needed to be rock stable.

That is, of course, the contradiction i was talking about.
I haven't seen much complaining regarding "severe hardware bugs" in the majority of AMD Opteron's server chipsets (made by Nvidia).
 
But how can they get into that sector if their chipsets are "buggy", I thought the server market stuff needed to be rock stable.


I think the bugs are a combination of different driver settings and OS's, I noticed from going from windows xp to vista, the bugs I had (would cause missing hard drives) on nf4 disappeared. Also the missing hard drives only happened with in XP with certian graphics drivers.

As Inkster said though, highly profitable area, and their server platforms are noted to be very solid for Opterons.
 
Chipset stuff, not unsurprising. Keeps them in the game and not requiring their own x86 chip, for now.

The Xeon thing is pretty big, hell huge if you think about it. Of course depending on how much penetration they can get with the OEMs.
 
Yeah, that's all great.. except in his musing Charlie seems to have left out an important part: What would Intel get out of this? They don't go giving away access to their market out of goodness of their heart and they don't NEED Nvidia to fill those roles.
 
Yeah, that's all great.. except in his musing Charlie seems to have left out an important part: What would Intel get out of this? They don't go giving away access to their market out of goodness of their heart and they don't NEED Nvidia to fill those roles.
Well, the first thing that comes to mind is "Making sure NVIDIA is NOT doing CPU R&D", but that doesn't strike me as a terribly likely scenario. The Inq's article is clearly named "What Nvidia got from Intel", not the other way around though. So perhaps Charlie is really hinting at a Part Deux...
 
Yeah, that's all great.. except in his musing Charlie seems to have left out an important part: What would Intel get out of this? They don't go giving away access to their market out of goodness of their heart and they don't NEED Nvidia to fill those roles.

SLI.
How can Intel sell motherboards and high-end CPU's that are competitive against AMD, if it requires giving a helping hand precisely to that same competitor ?
The halo effect would be lost, and X38 would've had an awkward time supporting AMD's Crossfire, indirectly funding the CPU competition.
Besides, SLI had its debut in an Intel chipset (Xeon's E7525).

The "Conroe" family didn't get this far by being cheap since launch, they got here because it trounced the Athlon 64 X2 especially in the... gaming department.
 
They would need to get their SLI license for free, because even at a reduced fee they would still pay nVidia and gain nothing from this deal...
 
Which ones? Not to put too fine a point on it, but Nvidia IS GPU patents. If they cross-license too much away, they might as well get naked on their high horse and go home*.







*not sure what that metaphor means yet, but Copyright 2007 Geefrocer just in case
 
Which ones? Not to put too fine a point on it, but Nvidia IS GPU patents. If they cross-license too much away, they might as well get naked on their high horse and go home*.

On the other hand, if they signed a royalty-type deal (like the one with Sony for the production of RSX), then Intel could sell many more GPU's than they could ever have imagined, and NV collects a few bucks in the process, together with their own existing products of that time.

The "lingering poison pill" strategy.
"If you want to compete with me later, you must agree to give me a share of all your profits, because they'll use my technology at its core."
 
Which ones? Not to put too fine a point on it, but Nvidia IS GPU patents. If they cross-license too much away, they might as well get naked on their high horse and go home.


Well, as Arun pointed out to me when we discussed this, there is a difference between "cross-licensing" and "technology transfer". The first provides you legal cover only. You still have to "reinvent the wheel" yourself, and most patent filings are carefully vague and general.
 
Possibly related to my page 4 post...

I didn't know that Barcelona is intended to have split power planes and thus split clocks, with the Northbridge (read: memory?) controller volted and so clocked higher than the rest of the CPU. Could Inq be confusing what Anandtech says is the MC's "200-400MHz increase" over the CPU core clock with a higher general CPU clock? I'm not sure Charlie or his sources would make a mistake this basic, but I'm equally skeptical that AMD can pull 20% higher clocks out of its hat at the last minute.

I barely knew about Barcelona when it was speculated that R600 would be delayed to coincide with a platform launch, though, so forgive me if none of this is new.
 
Back
Top