Why I think Intel is in trouble.
First, I’m not an expert on CPU’s or CPU architectures, and what I write here could be completely missing the mark. But I’m basing this on logic, common sense, and how I see things.
The first reason I think Intel are in trouble, is their CPU architecture.
Intel ‘core’ architecture is now in its 7th generation, and will very soon enter the 8th generation of the Intel ‘core’ architecture. Despite the new generations, the basic ‘core’ architecture remains the same. They improve parts, and bolt bits on, but the basic architecture is still the same.
Intel has been unchallenged for about a decade, so there was no need for them to spend large amounts of money developing a new CPU architecture.
AMD on the other hand realised four years ago that the architecture that they were using was quite frankly rubbish compared to Intel’s ‘core’ architecture, lagging some 60% in terms of IPC performance when compared to Intel’s ‘core’ architecture. So AMD started with a clean sheet of paper, when they started out with their ZEN architecture. At the time, AMD announced that they would bring a 40% gain in IPC performance from their ZEN architecture. Intel made fun of them, stating that AMD didn’t have the expertise to bring about such an advancement in IPC performance. In actual fact, ZEN has a 52% gain in IPC performance over their old architecture.
This still leaves ZEN some 8% behind Intel, but Intel must have got the shock of their lives, when AMD made such high gains. You should also take into consideration that AMD RyZen is the first generation of the ZEN architecture, with ZEN 2 on target to come to market some time in 2018, built on an 7nm node size. In fact they could beat Intel to the market for a sub 14nm chip.
You would expect AMD to make further significant IPC gains, in Zen 2, just like Intel made with the 2nd and 3rd generation ‘core’ CPU’s, simply because the ZEN architecture is new, and you would expect there to be room for improvement.
In contrast, Intel’s ‘core’ architecture is now getting very long in the tooth, and they have had very little gains in IPC performance in recent generations of ‘core’. I personally think they have now hit a brick wall with ‘core’ and they need to set about designing a brand new CPU architecture. If they do take this route, that wont come to market for several years. Perhaps not as long as four years, but it wont appear tomorrow.
In the mean time, RyZen is modern, its a infant, and has loads of room to grow. But even now, it has some fantastic new technologies in the ZEN architecture, like ‘Infinity Fabric’ which is like a ultra high speed network (150GB/s) 256bit data path with ultra low latency, connecting one part of the CPU to the other. Then there is ‘SENSEMl’ which uses advanced AI, to predict what the CPU will be doing next, and loads that data into its ultra high speed cache.
The second reason I think Intel are in trouble.
Intel makes very neat CPU’s, lets take their new SkyLake X CPU’s with up to 18 cores. It’s in a very neat package, nice and small.
They are packing 18 cores into that tiny package, and then connecting it to the cooling solution via a metal heat spreader, which is relying on a TIM for thermal conductivity between the CPU die and the heat spreader.
So somehow they will have to transfer the heat away from the CPU cores, from a small surface area which doesn’t make very good thermal contact with the CPU die.
Intel made fun of AMD when they announced RyZen Thread Ripper, saying they had simply glued two Ryzen CPU dies together and put them in a single large package. That part is true, but Intel failed to mention that they would be glued together with ‘Infinity Fabric’, and with that large CPU package, AMD RyZen Thread Ripper, would have more than twice the surface area to transfer away the heat compared to Intel SkyLake X, and that those AMD RyZen cores would be soldered to the very large heat spreader, insuring the optimum heat transfer between the CPU die and the outside world…
Adding to that, RyZen is very much more energy efficient than SkyLake X. A lot of people only think of energy efficiency, as meaning the CPU will be cheaper to run in regards to the amount of fuel it consumes. But that isn’t the end of the story. The more energy a CPU consumes, the more heat it produces, heat that has to be taken away from the CPU. It’s simple physics, the larger the surface area, the more heat you can transfer to the outside world.
So which one would you think would be the easiest one to cool, AMD Thread Ripper or SkyLake X?
Then there is price, the third reason I think Intel are in trouble.
When Intel launch Coffee Lake with its 6 core variant, they will no doubt claim that they have brought 6 core CPU’s to mainstream computing. Sorry Intel, you don’t decide what is mainstream, price decides what is mainstream, and AMD have already brought very competitive high performance 6 and 8 core CPU’s to the mainstream in the shape of RyZen R5 and RyZen R7.
One thing is for sure. It’s going to be interesting to see the test results from AMD RyZen Thread Ripper when compared to Intel’s 14 to 18 core SkyLake X CPU’s. I’m looking forward to them going head to head.