New competitor to AMD and Intel in 2013?

"Lyric Semiconductor, a Defense Advanced Research Projects Agency (DARPA)-funded spinout from the Massachusetts Institute of Technology"

Okay, that just made this far more plausible that they have the ability to create chips to challenge Intel and AMD and have the VC backing.
 
After reading through the whole thing, we definitely arent going to see their CPUs in desktops, they will probably be in totally custom systems because of their fuzzy logic ability, but they could turn out to be very helpful for some data centers. The error correctors for flash will almost certainly show up in desktops as long as they arent going to cost an arm and a leg.
 


Why does that name seams so familiar... [:grahamlv:3]
 

jj463rd

Distinguished
Apr 9, 2008
1,510
0
19,860


You buy a LCD screen recently, guess what, LCD tech wasn't made by private business. Neither was the internet, transistors, or even computers in general.

About 90% of the underlying technology used today can be traced back to government or University origin. Hence, your argument fails. Nevermind this chip will liekly compete in integrated systems, and not the desktop market.
 

dont_look@me

Distinguished
Aug 6, 2010
81
0
18,630


I'm not saying that wiki is always right but this looks credible.
 
Well it can't compete in the desktop or notebook market unless it's X86, duh. So good luck competing. LOL

no, it would have a very hard time competing in the consumer market if it isn't x86, personally if it is that much faster i would buy another desktop/laptop to use for software development, but it would have to be a lot faster
 


Unless its 1000 times better is accurate, then it wont compete in the x86 market at all. All the software is coded on and for x86 currently. Until something comes out that can emulate x86 at 100% or better, they wont get very far.
 

werxen

Distinguished
Sep 26, 2008
1,331
0
19,310


Rofl +1.

My school has an old supercomputer with 256 cores that needed the power from a nuclear reactor to run and also claimed figures such as these. I doubt these new chips will be relavant considering the serverside market for Intel and and is so strong.
 

sonoran

Distinguished
Jun 21, 2002
315
0
18,790


Probabilities? Hmm, so you walk into the bank that's using these CPU's, and the teller says "Sorry sir, the computer says you probably don't have any money."

I'm sure it has its uses, but I don't think banks or stock exchanges will be run on these.
 
Thats not the part of the bank it would be used for, fuzzy logic probability systems could be invaluable for a bank deciding which stocks it should invest in and whether its a good time to sell them off of buy more. It would be used more in the back end of the bank with their investments and risks rather than with regular accounts.
 


i think they meant 16KiB of L1 D-Cache per core in the module (so basically per thread), but it concerns me as well, but maybe the L2 is faster than K10 so it could make up for it
 
I'm already suspecting that memory access speeds are going to be VERY important for Bulldozer, especially with the size of that L1 cache...I'd be VERY interested to see the first memory access benchmarks with various RAM kits, to say the least.

I also want to see the effects of not replicating the Floating Point Units. I suspect significantly decreased performance (versus Intel) in floating point tests. For some coorporations, this could be a deal-breaker.

We'll see; I'm not real impressed so far.
 
idk, it would have the same 16 KiB L1 Data per thread that Intel currently has for the i7

as for the floating point capability, it should be a lot better (as compared to AMD right now) as the 128 bit FPU's are what is currently used and each module has two of them, they work together on certain instructions though
 
overall I think that all cpu's will drop in price because of competition which will be good for the consumer.
the question is, who will prevail as superior.?
reports do say that the AMD Bulldozer and Bobcat cpu's are going to be good.

Reports said that K10 was going to be amazing. Look how that turned out. K10.5 was better but still not amazing.

Reports also say that SB will be good.

But reports have been known to be very wrong.