Speculation that Intel will announce acquisition of Nvidia today:
Intel to take over Nvidia
My own instinct is that an acquisition is probably not the best solution if all they're after is technology. A stake in Nvidia or even a JV or a MOU type arrangement to co-develop certain technologies are better paths to gaining advantage of Nvidia's technical capabilities. However, from a business standpoint having Nvidia's comprehensive product line up and brand will be a good counter to how AMD may try to leverage the huge strength of the ATI brand. Whatever it is...any kind of alliance between Intel and Nvidia is BIG!
--------------------------------------------------------------------------------
UPDATED
The weekend is here and Intel has still not bought it's bagful of (graphic) chips. I think this deal is not a deal. But something is in the works between the two. Collaboration is inevitable. The enemy of my enemy is my friend. Neither Intel nor Nvidia will want AMD/ATI to strengthen because they are now going to feed each other. Nvidia also does not want Intel to strengthen because it is also a competitor...though the lesser of the two evils. However, with Intel's intention to get into discrete graphics, they are heading for a collision. What choices does Nvidia have - become an end to end CPU+GPU company. Easier said than done. Or find a partnership with someone else who can help them get there. As I said in a post earlier, Nvidia and Intel will get closer. I'm sure something will happen by the end of the year.
Subscribe to:
Post Comments (Atom)
11 comments:
This news does not make any sense to me. why? because the regulatory authorities will not allow it! after this merger or acquisition Intel/Nvidia will have a huge segment of the GPUs market (around 70% I guess).
I guess we can wait and see what happens by the close of business on Friday. That being said it doesn't make as much sense to me for Intel as the AMD acquisition of ATI did. Apparently I am not the only one to feel that way.
"AMD's purchase of ATI allows the Intel rival to enter a new market as it currently doesn't make graphics chips.
An Intel acquisition of Nvidia wouldn't have that appeal because Intel already sells graphics chips, an area where it has been expanding by hiring engineers specialized in that field, said Pacific Growth Equities analyst Satya Chillara."
Why would Nvidia want to ally with Intel again?
"Why would Nvidia want to ally with Intel again?"
Before this Intel was touting real time 3D rendering using software and multi-core stuff. At IDF 2006, the 80 core teraflop processor must be scaring the pants off GPU companies like NVDA. Imagine a CPU capable of doing what GPUs doing now but in real time and using mere software. Plus those terabytes of bandwidth, lots more possibilities.
Now wouldn't that make NVDA nervous about the future? Already GPUs are beginning to feel the limit and bottleneck at very high resolutions. And within 5 years time, where would NVDA stand if they can't get better die shrinks and cramp more transistors into their GPUs running at low clock speeds while giants like Intel starts making multi core CPUs which clocks higher and have higher performance.
Historical lessons - remember these hardware? MIDI wavetable soundcards, DVD/MPEG decoders, hardware DSP based modems, etc? Nowadays processors became so faaast, these can be done in software. Didn't you think NVDA ever took a leaf from these historical stuff?
My 2 cents.
Mr Nvidia is too smart for Intel..
sure anyone is for sale but Nvidia will want 2-3x 38$ a share because Nvidia want to be a force in CPU and GPU they wont sell to a has-been company!.
They recently licensed from Imagination Technologies. I don't see what Nvidia has to offer.
While I don't think a merger or complete acquistion of Nvidia makes sense...if it were to be the case I wonder if Intel could make the case to the regulators that with AMD now owning ATI, they may use it to lock Intel out of certain markets and hence Intel needs to have acquire Nvidia to remain competitive.
I haven't formed this thought fully so give it some time to breathe. Any inputs welcome...
"While I don't think a merger or complete acquistion of Nvidia makes sense...if it were to be the case I wonder if Intel could make the case to the regulators that with AMD now owning ATI, they may use it to lock Intel out of certain markets and hence Intel needs to have acquire Nvidia to remain competitive.
I haven't formed this thought fully so give it some time to breathe. Any inputs welcome..."
I see what uR saying there old lad, but I dont think regulators will aprove this. It would give Intel around 75% of the grahic market. Considering they allready have 75% of the cpu market i think regulators will scream in pain..
/MOrK
ashenman said..."I think if anything Intel would be afraid of Nvidia because of their research into the general purpose GPU.
I don't see where you're getting anything about a resolution bottleneck. If anything, Nvidia is finding it hard to make their core able to process advanced shader and texture data more efficiently. If anything, an 80 core Intel processor based on older pentium technology would do an even worse job than a core 2 duo if it were given the job of graphics processing because of the lack of thread parallelism in current graphics code, and because of the specialized processing architectures currently used for vectors and textures."
OK you missed the key word - TERABYTES of bandwidth. Impossible by today's GPUs.
Thread parallelism? Didn't I mention about Intel encouraging those? Didn't IDF quad demo illustrate what Intel is trying to put acorss about multi-cores?
By comparson, present day GPUs can do only about 360 GigaFlops, about a third of a teraflop.
Even then Intel can even write themselves a DX10 driver that runs on this 80 core CPU? 1 Teraflop would mean 3X the speed of present day GPUs. And that Terbytes of bandwidth means it can do it easily at higher resolutions.
More to come...
My 2 cents.
"OK you missed the key word - TERABYTES of bandwidth. Impossible by today's GPUs.
Thread parallelism? Didn't I mention about Intel encouraging those? Didn't IDF quad demo illustrate what Intel is trying to put acorss about multi-cores?
By comparson, present day GPUs can do only about 360 GigaFlops, about a third of a teraflop.
Even then Intel can even write themselves a DX10 driver that runs on this 80 core CPU? 1 Teraflop would mean 3X the speed of present day GPUs. And that Terbytes of bandwidth means it can do it easily at higher resolutions.
More to come...
My 2 cents."
I think YOU are missing the point. The price point for a 360 Gigaflop GPU is ~$500. Can you even contemplate the cost of an 80 core CPU currently?
$10,000? $50,000? $100,000?
I can't even imagine what the yield rate would be.
GPU's are EXTREMELY inexpensive compared to a conventional CPU. For those of us that do scientific computing it's not uncommon to use two high speed GPU's for code acceleration. For a standard 2x PCI x16 system, $1000 will buy you 500+ Gigaflops for parallelizeable code.
I personally use:
http://www.acceleware.com/index.php?page=22
In terms of value, GPUs dominate on parallelizeable code.
Lastly, intel already controls ~30% of the graphics market because of their integrated graphics chip sets. Purchasing Invidia would push it past the magic 50%. Very doubtful that the FTC would allow it to happen, but I may be wrong. That is just conjecture on my part at best.
ashenman... I really like that idea of Nvidia buying VIA. That would finally bring that crucial third player into the game. I have a VIA EPIA-SP13000G system wich I run Suse Linux on. Its a good affordble way for those who want to experiment with other OS then Windows. The one thing lacking is a better graphiccard. VIAs CN400 is much to boast with.
/ze m0rk
Post a Comment