Intel's Mooly Eden just disclosed Ivy Bridge's transistor count to a captive audience at IDF this morning: 1.4 billion transistors. That's presumably for the quad-core version compared to 1.16B in Sandy Bridge. Don't pay attention to the die shot above, that's not an accurate representation of Ivy Bridge. Intel is holding off on showing the die until closer to launch. Why? Likely to avoid early disclosure of how much of the die is dedicated to the GPU. As you'll see in our Ivy Bridge architecture piece later today, the lion's share of those transistors are improvements in the GPU.

Update: Intel has provided us with updated numbers for both Ivy Bridge and Sandy Bridge. Transistor counts are only comparable when they're done the same way. Intel is saying that SNB is 1.16B and Ivy is 1.4B. The gap between these numbers is only 20%, which is much more in line with what we'd expect from a tick. I'm waiting on a response from Intel as to why the SNB number went up since launch.

Update 2: This is why 1.16B and 995M are both correct for Sandy Bridge.

POST A COMMENT

16 Comments

View All Comments

  • imaheadcase - Wednesday, September 14, 2011 - link

    Also does sandy bridge new one coming out in a couple months..that have a gpu or no? Reply
  • DanNeely - Wednesday, September 14, 2011 - link

    Sandybridge-E (LGA2011) doesn't have an IGP; this probably means that IvyBridge-E won't either when it launches next year. Reply
  • phatboye - Wednesday, September 14, 2011 - link

    What sucks the most is that even though Ivy bridge will have 50% more transistors which as Anand points out will more then likely be dedicated toward the GPU, people like me won't likely notice a difference since I use a dedicated graphics card connected via PCI-e. I although I doubt it would ever happen for the mainstream consumer line of CPUs I really wish Intel would go back to making CPUs without the on chip GPU. I don't want to spend the money to buy an expensive LGA 2011 platform yet the LGA 1155 platform waste die space on a GPU that I will probably never utilize. That extra die space could be used for more CPU cores, more cache or even just removed entirely so as to cut cost and the other benefits of having a smaller die. Reply
  • ImSpartacus - Wednesday, September 14, 2011 - link

    Yeah, it's kinda silly that the high end CPUs get the best graphics, yet they rarely utilize them.

    While the low end processors get the neutered GPUs, but those are the systems that will probably go without dedicated graphics.
    Reply
  • silverblue - Wednesday, September 14, 2011 - link

    The only benefit of this sort of GPU to such a powerful CPU is QuickSync. Why couldn't they just throw in a tiny section to the die to handle that?

    AMD are doing it a different way. BD won't even have a GPU, and Trinity is only a dual-module design, leaving the FX models to be paired with a discrete card. One thing I found of interest is that AMD believes Trinity to be 50% faster at floating point calculations than Llano, which points squarely at the GPU and not the CPU, though you'd expect Enhanced Bulldozer to be equal to the CPU part of Llano at the very least, with the added bonus of turbo to lift up lightly threaded workloads. Then again, saying all that, AMD has been known for building up its products far too much in the past - Barcelona 40% faster than Clovertown? Hell no. Bulldozer up to 50% faster than Thuban and i7 950? Doubtful - benchmarks or GTFO.

    Back to the original point, I'd really like to see what all this extra real estate on die is going to do for graphics, even if it's a strange idea to put what could end up being a 6600-class GPU with a 4C/8T monster.
    Reply
  • Slaimus - Wednesday, September 14, 2011 - link

    I've always though the main point of AVX is to have twice the encoding speed of SSE4. Why is AVX not being used for encoding, but even more additional logic? Reply
  • gamerk2 - Wednesday, September 14, 2011 - link

    Time to develop software, lack of automatic compiler optimizations, etc.

    Really, how many programmers do you think actually use SSE on a day to day basis? 90% of the time, its the code the compiler spits out that contains SSE, not because the developer put it in.
    Reply
  • DanNeely - Wednesday, September 14, 2011 - link

    In general the more specialized a piece of hardware is the better it will perform at a given task. Quicksync is a dedicated decoder for several common formats and is blazingly fast for them; but won't do anything for more obscure formats or those developed in the future. AVX are general purpose vector instructions; they can be used for anything that needs vector processing not just arbitrary video en/decoding. Reply
  • fic2 - Wednesday, September 14, 2011 - link

    "One thing I found of interest is that AMD believes Trinity to be 50% faster at floating point calculations than Llano, which points squarely at the GPU and not the CPU, though you'd expect Enhanced Bulldozer to be equal to the CPU part of Llano at the very least"

    I don't think that necessarily points at the GPU. I would hope that BD is at least close to 50% in fp than Llano since the cpu is really just a Phenom II. Maybe it does but I didn't think the truly merged cpu would come out for a couple of years.
    Reply
  • Taft12 - Wednesday, September 14, 2011 - link

    Looks like they're doing it the right way to me - incentivizing you to buy a CPU more powerful (and expensive) than what you need, or conversely, making you pay for the GPU upgrade that you won't even use.

    Cha-ching!!!
    Reply

Log in

Don't have an account? Sign up now