Jump to content

Nvidia GPU Technology Conference Coverage


Bosco

Recommended Posts

  • Replies 62
  • Created
  • Last Reply

Top Posters In This Topic

The term Co processing is being badied about. The GPU is not being looked at as a replacement for the CPU snce the CPU is great at serial functions while the GPU is better on the parallel functions . Meaning the two together do the work faster

:withstupid: we're not there yet fellas :lol: (and ladies)

Share this post


Link to post
Share on other sites

it does sound like a larrabee competitor , except larrabee doesnt even look too impressive anymore

 

http://www.anandtech.com/cpuchipsets/intel...?i=3367&p=7

 

Actually larrabee isnt even close to getting out and they are getting crushed. Intel had made statements about how current GPUs design would have a huge disadvantage in computing becuase they lacked cache on their cores.

 

I cant see how Nvidia can make a gpu with that much cache and it be reasonable, its usually extremely expensive and costly, defective cache is a big issue in microprocessors. I wonder how they couldve did it and remain affordable. I am sure these will be expensive. I just dont see the their average gaming gpu with all this cache. maybe their higher tesla models.

Share this post


Link to post
Share on other sites

I can't wait to see what Intel has in store for us with Larrabee. I'm somehow not expecting too impressive gaming performance due to their track record, but it should be strong on parallel computing and easier to program for since it's build with x86 cores.

 

Fermi has 3 billion transistors, there is no doubt it will be much more expensive to manufacture than RV870. It's also most likely -- hopefully -- going to be faster as well. Just like GT200 compared to RV770.

 

It's quite clear NVIDIA is confident, or rather hoping, that their Tesla division picks up with Fermi. Not only do they have CUDA and more cache, but also half-speed double precision along with a full featured C++ compiler are powerful features for general computing.

Share this post


Link to post
Share on other sites

but the nvidia guy said that he priced the gtx200 wrong and he will not make the same mistake with gtx 300 ...

 

I was talking manufacturing cost ;)

GT300 will be about the size of G80, so it'll be smaller than GT200, but it's still a very large chip. Which in turn makes it harder to scale down. Notice we still don't have low/mid-range GT200 products yet?

 

As for retail pricing, it's all speculation at this point, but it's pretty much a given the top-end Fermi core will be more expensive than a 5870. And rightly so because it should be faster as well. By Q1 2010, we should have $299-$329 5870's, so it will be interesting to see what NVIDIA comes up with.

 

The GT200 launch pricing fiasco (let's be honest, it was) won't happen again simply because NVIDIA knows what ATI has and for how much. R770 was just much faster than anyone anticipated, so they were caught off-guard.

Share this post


Link to post
Share on other sites


×
×
  • Create New...