How come Trinity got a much higher density than Vishera? Remember Vishera got a whole 16MB cache with 4x2MB L2 and 1x8MB L3.
And again, I doubt the transistor count for Kaveri is even remotely correct.
We've already said this multiple times. The reported transistor count on Kaveri is off by a considerable amount. The GPU, which accounts for nearly half the die and consists of very dense logic, cannot take up more than 1 billion transistors. The rest of the die on average cannot be as dense as the GPU, so therefore the rest of the die can't be more than 1 billion either.So how come A10-7850K has higher transistor density than 4770K, when also taking into account that they both have similar percentage of die area allocated to high density blocks (GPU and cache)?
Are you saying AMD's cache, CPU cores, and/or GPU cores are denser by design than Intel's? Or what other explanation are you suggesting?
Please keep your irrelevant quips out of the conversation. He's claiming the process is denser. It is factually not denser.He could as well have used a single other parameter as straw,
why not connexions thickness or gate legnth or whatever
other part that seems falsely favourable.
As you point it what matters is the sum of all thoses metrics
down to the bottom line wich is the transistors/mm2, all the
rest is useless debate about details.
You should know better that Vishera doesnt have an iGPU and the topography is optimized for lower temps sacrificing die size. You cannot compare Vishera to Trinity.
We've already said this multiple times. The reported transistor count on Kaveri is off by a considerable amount. The GPU, which accounts for nearly half the die and consists of very dense logic, cannot take up more than 1 billion transistors. The rest of the die on average cannot be as dense as the GPU, so therefore the rest of the die can't be more than 1 billion either.
Kaveri may be a denser chip than Haswell, but GloFo's 28nm process is not denser than Intel's 22nm process. Period. Even Intel's high performance process, which is what Haswell would use, has an SRAM density of .108um2.
Well, if it's true, then I suppose that we can assume that their dGPU transistor counts are fabrications.Until AMD change it, that number still holds true. Not you nor anyone in this forum has the knowledge to refute it.
It wasnt targetted at you, and it was a leading question. Same way as you cant compare any design with another design and make a statement about the process node. Despite what someone is looping around and around to falsely prove.
Well, if it's true, then I suppose that we can assume that their dGPU transistor counts are fabrications.
Our reasoning is sound. I suggest you actually go and read what we've said.
We've already said this multiple times. The reported transistor count on Kaveri is off by a considerable amount. The GPU, which accounts for nearly half the die and consists of very dense logic, cannot take up more than 1 billion transistors. The rest of the die on average cannot be as dense as the GPU, so therefore the rest of the die can't be more than 1 billion either.
Please keep your irrelevant quips out of the conversation. He's claiming the process is denser. It is factually not denser.
Kaveri may be a denser chip than Haswell, but GloFo's 28nm process is not denser than Intel's 22nm process. Period. Even Intel's high performance process, which is what Haswell would use, has an SRAM density of .108um2.
Cape Verde has 1.5 billion transistors with two 64 bit memory controllers, VCE, PCI-E PHYs, etc.I suggest to have a look at this first, then rethink about the higher transistor count in the iGPU.
Keep in mind that GloFo's 0.120um2 density is the highest density they offer, which is a number that should be compared only to Intel's 0.092um2. Obviously high performance transistors would less dense.0.108 to 0.120 is very close but yes the Intel 22nm has higher density as a process.
I'm actually the one with the least-biased track record in this conversation, except for maybe Fjodor and mikk.If your biaised opinion
Except Intel's not fudging their numbers. Intel claims a 0.108um2 density for high performance SRAM. And guess what Chipworks found? 0.108um2.If your biaised opinion must be taken at face value despite
AMD claiming 2.41bn then i can equaly say that intel s
datas are flawed and that the sizes they claim are wrong,
hence their lower density as proved by the transistors
count per unity of area , the very definition of density
that you keep denying for trolling purposes...
Cape Verde has 1.5 billion transistors with two 64 bit memory controllers, VCE, PCI-E PHYs, etc.
And larger.You totally forget that Kaveri's gpu ain't same as discrete gpu, it's hsa and totally cache coherent with cpu cores. It's a big deal and make gpu much more complex to implement.
Cape Verde has 1.5 billion transistors with two 64 bit memory controllers, VCE, PCI-E PHYs, etc.
Seriously, the comparison being made here is absolutely ridiculous.
Keep in mind that GloFo's 0.120um2 density is the highest density they offer, which is a number that should be compared only to Intel's 0.092um2. Obviously high performance transistors would less dense.
He is comparing an APU to an APU, you compared a 300mm2 8 core CPU to 246mm2 APU. It is like comparing 8 core Ivybridge-E to 4core + iGPU Ivybridge. Not comparable.
Great, glad you agree. Looks like I'm right. So it sounds like we can put this subject to rest.I really dont have any more data so i can only guess at this time but as i have said so many times before, Intels 22nm has higher density.
You totally forget that Kaveri's gpu ain't same as discrete gpu, it's hsa and totally cache coherent with cpu cores. It's a big deal and make gpu much more complex to implement.
Try compare iGPU IBs with IB-E, EP and EX. Might expand your horizont
Our reasoning is sound. I suggest you actually go and read what we've said.
You clearly haven't read the thread.I'm just curious: what sort of club you're representing and could you name members of it?
Lets do that,
4C Ivybridge (Core i7 3770K) = 1.2Bil transistors with 160mm2 die size = 7.5Mil transistors per mm2
6C IV-E (Core i7 4960X) = 1.86Bil transistors with 257mm2 die size = 7.2Mil transistors per mm2.
You were saying ???
And dont forget that 4C IvyBridge iGPU occupied only 27% of the die size. If it was at 47% the transistor density difference vs the 6C IV-E would be even higher.
You clearly haven't read the thread.
So you just proved you can compare and iGPU+CPU with a CPU. Yet you dont want to with Vishera
Let me help you further. I am sure you already did this math, but forgot to post it.
EN/EP-10:
2.86B/346.5mm2=8.25M/mm2.
EX-15:
4.3B/541mm2=7.95M/mm2.
Both higher densities than the iGPU part. And before you say cache, remember Vishera got 16MB of L2+L3.
Again, utterly rubbish to compare anything but the SRAM only densities.
There could be as well 128SP more than official max for redundancy purposes on an early and perhaps averagely efficient process yields wise, at this point we dont know , we ll get a better picture in a week or so.