Realistically now, with the AMD/ATI Deal. Nvidia to do CPUs?

demonic

Regular
There was a rumor awhile ago. That Nvidia would get into the CPU arena.

In alot of sites I have read, AMD have bought ATI to survive.

Anyway..

In terms of CPU's, Chipsets and GPU's. I would say AMD have a leg up against Intel. Although I am not discounting Intel at the moment.

But what about Nvidia. They do great GPU's and Chipsets. Will they also go the route of GPU, CPU, Chipset convergence like AMD is proposing?

Or are they just going to be carrying on as normal and compete/compliment intels and amds lineup as a 2nd choice offering?

Nvidias CEO did say, they wanted to be as big as Intel. Perphaps they are just waiting for things to slot into place, before they step up to the plate?

Thoughts?
 
Considering Nvidia has no fabs, I'd say it's rather unlikely they'll be touching the CPU market any time soon -- there is no way they could feasibly compete. I'm not quite sure people want a third CPU choice anyways.

It seems to me they'll just continue to do the chipset thing for Intel and AMD and produce GPUs like always. They've been very successful with their margins and I don't really see the ATI/AMD deal affecting that any time soon (at least not drastically). Should be interesting to see what kind of changes do happen in the next 5 years or so.
 
CPU market is a rough one at the moment and I cant imagine Nvidia trying to go into it.

They will continue to focus on their core market and probably be quite successful at it.
 
Nvidia dont just lack fabrication space to manufacture cpu's, they also lacks experienced personel to make cpu's, and own nothing of existing PC cpu pattents, or technology.

Who CAN come into this space, is IBM, if they want. Nvidia could presumably buy a small processor firm ,and found devlopment of a new type of cpu, but, whats the point?
 
Gosh, I hope not, it would be a big distraction to their core business. I hate this AMD deal, I feel like it is dangerous and damaging to competition in the GPU market and I fear it will hurt ATI's competitiveness and AMDs.

During the P4 debacle, I loved AMD. I though they were owning Intel on all fronts, but Intel's new chips are sweet and now AMD looks like they have been resting on their laurels. I'm worried that ATI's revenues might help remove some of the urgency of AMD to retake the CPU lead it had, or vice-versa, that one misstep by the ATI business unit might cause AMD to hack off the high-end and plunge resources into low-end or integrated solutions (including the oft-talked about CPU/GPU merger, which I do not view as a high end part)

I would rather have two companies laser focused on CPUS fighting it out, and two companies laser focused on GPUs fighting it out, and I feel the buyout could potentially be harmfull to competiveness in these markets.
 
DemoCoder said:
I would rather have two companies laser focused on CPUS fighting it out, and two companies laser focused on GPUs fighting it out, and I feel the buyout could potentially be harmfull to competiveness in these markets.

Agreed, that pretty much sums it up.
 
kimg said:
Nvidia dont just lack fabrication space to manufacture cpu's, they also lacks experienced personel to make cpu's, and own nothing of existing PC cpu pattents, or technology.

Now this can change drastically. I just did a little research on Sun's Niagara and guess what. It's wide, has high TLP, and clocked very conservatively - remind you of anything?

The more I read up on the future of CPU's it looks like they will resemble today's GPU's a lot more than they resemble today's CPU's. Maybe a fabless IHV might have a fighting chance in the CPU wars of the next decade. However, that might be a near impossiblity as long as x86 remains the de facto instruction set. If this should change it will open up the market to more players.
 
DemoCoder said:
I would rather have two companies laser focused on CPUS fighting it out, and two companies laser focused on GPUs fighting it out, and I feel the buyout could potentially be harmfull to competiveness in these markets.

I really don't understand your point of view. Intel was never "laser focused on CPUs". It says right in their company motto that they want to dominate all applications of integrated circuits. Intel is a CPU, GPU, chipset and more company, a platform company. They are not and never were focused on CPUs 100%. Likewise, AMD until only recently was one of the biggest players in the NOR flash space. Also, ATI and Nvidia make chipsets, TV chips, mobile chips and more. Even they aren't "laser focused on GPUs".

The computation war is about just that, computation. It's not about GPUs, CPUs or DSPs. It's about the overall computational package you can supply, your platform. After all, wheather or not you multiply two floats on a GPU or CPU, it's still the same operation. The foundation is the same, computation. If anything, applying your philosophy to any of these companies is only a sure way to garuntee their bankruptcy. Reality doesn't analouge romantic fantasy.

In this more rational regard, I think there's a decent chance Nvidia will do CPUs. It's not garunteed, but it will fit their platform nicely. I'm pretty sure they have enough savings to build a fab or two. Even if it takes a few years, I don't see what's stopping them.
 
Last edited by a moderator:
DudeMiester said:
In this more rational regard, I think there's a decent chance Nvidia will do CPUs. It's not garunteed, but it will fit their platform nicely. I'm pretty sure they have enough savings to build a fab or two. Even if it takes a few years, I don't see what's stopping them.

Now I have no real knowledge or experience on the subject, but it would seem to me that Nvidia's experience with GPUs would be more suited to CPUs than their experience designing for other fabs would be suited to running their own fab.


What would seem like a likely/ideal path for Nvidia to follow might be a CPU very much like Sun's Niagara II, and then bill it as a multimedia processor. I wouldn't expect Nvidia to necesarily excel at office type work loads but if they leveraged their GPU experience and built a very parallel processor they'd most certainly have the fasted video encoding/decoding and physics processor around. The only problem I can see is that the (likely) poor branching performance would hurt them in games, and would work to tarnish the "multimedia" aspect of it all.
 
chavvdarrr said:
if they buy VIA, they can try... :D

Why VIA? 'cause they'll need x86 license.


Except that VIA is already owned by someone, and is probably bigger than nVidia...

But VIA builds CPU's, and is fabless, Transmeta builds CPU's, and is fabless, granted, neither competes in the consumer desktop space, they compete in the blade server space, thin'n'light space, network applicance, etc

If nVidia wanted to buy a company to build CPU's, Transmeta would be a much more likely canidate at least imo (it's a lot smaller and it isn't doing super duper well, it fits the profile of a "small CPU company")

I could see nVidia going into CPU's, they have ULi + their own chipset design, which has to give them something, I mean, they can already design excellent memory controllers, I/O controllers, graphics subsystems, audio subsystems, all their missing is the actual ALU/FPU processor (I mean this in a very rough and very very broad sense)
And as a final kicker, they've probably gotta have at least some tidbit of experience with CPU on GPU (which I've heard isn't that amazing currently)
 
Nvidia could easily buy VIA for cpus. That way nvidia could sell a complete system, focused on the graphics chip with a decent enough cpu.

Though more likely, I just see them pushing to get more workload shifted onto gpus, thus lessening the importance of cpus.

BTW, I think the ATi buy was good for AMD, since ATI has a lot more room for growth than AMD. AMD really only has one market, ATI is in everything.

so does that mean that VIA is worth 500 mln US $ !?

Last I checked, I believe VIA was worth close to $2 billion (but maybe I was off by a factor of 10?), with the conglomerate that owns them being much much larger than that.
 
obobski said:
If nVidia wanted to buy a company to build CPU's, Transmeta would be a much more likely canidate at least imo (it's a lot smaller and it isn't doing super duper well, it fits the profile of a "small CPU company")

And I imagine that if Transmeta's translation software were modified to treat a GPU as a massively parallel x86 CPU you could have a pretty interesting setup.
 
Read somewhere also, that Nvidia is working with Sony on Cell. That Cell is already working on Linux? I guess, if they made Cell to work on X86. That could be a possibility too?
 
demonic said:
Read somewhere also, that Nvidia is working with Sony on Cell. That Cell is already working on Linux? I guess, if they made Cell to work on X86. That could be a possibility too?

x86 is a totally different architecture then Cell, the best they could do is some sort of emulator and that wouldn't fly.

Crazy theory: maybe their entry into "computers" is the PS3? Imagine the faces at MS if PS3 should have a full-blown OS which could actually somewhat compete with windows :p
 
Back
Top