Profitable innovation doesn’t occur in a vacuum. Somebody should make a substantial acquire for it to achieve traction. In any other case, it was simply an fascinating thought.
This week’s Design Automation Convention is all in regards to the new issues which can be happening within the business, each challenges and alternatives.
By this time this weblog goes stay, I’ll have moderated a panel about why EDA has not been open to disruption. Whereas making ready for that, plenty of ideas emerged in my thoughts. First, we have now to keep in mind that EDA is a enterprise whose function is to help the semiconductor business. They must be worthwhile, and they should present what their clients demand. In fact, that is the proper setup to be disrupted in accordance with Clayton Christenson, writer of the favored Innovator’s Dilemma.
What the semiconductor business has demanded is the flexibility to fill a chip with performance at a considerably fastened value by way of manpower. For that they have been keen to pay a sure degree of ‘tax’. That tax price stayed pretty fixed for a number of a long time, and so EDA income grew at roughly the identical price as the event value aspect of the semiconductor business. Other than a number of durations the place they fell behind, they’ve typically succeeded in offering the mandatory productiveness features.
Disruption requires a major change and that may come from both the semiconductor business or by new EDA applied sciences being developed.
Disruptions to EDA
Alongside the way in which, varied design methods failed, such because the efficiency curve of processors, or reminiscence entry charges. These created unintended disruptions for designers. The entrenched business would have been a lot happier seeing processor efficiency improve with each node. As a substitute, current processor distributors have been pressured to spend money on homogenous multi-processor chips, whereas the remainder of the business used it as a possibility to spend money on heterogenous multi-processor techniques built-in into an SoC. However this and different such modifications didn’t disrupt EDA as a result of they solely referred to as for incremental shifts.
I can’t consider any change within the design neighborhood that disrupted EDA. There have been a number of makes an attempt, such because the introduction of asynchronous design, and to some minds this failed as a result of EDA didn’t correctly help it and no startup rose to fill the void. We have now seen a steady strain for analog design to be made extra predictable, or easier, however no one has been capable of crack that downside. Manufacturing has provided a protracted record of latest points and considerations with each node, and EDA has provide you with the mandatory incremental options.
Maybe one of the best instance of a disruption is the rise in recognition of the RISC-V processor. EDA has responded and they’re creating instruments to assist with the design and verification of these processors. Firms like Synopsys have began to help this although they beforehand had their very own proprietary ISA. Synopsys has additionally bought one of many few startups, Imperas, that was creating instruments for these cores.
Disruption inside EDA
So, what about disruptive EDA improvements? There haven’t been lots of these over the previous couple of a long time, and there may be appreciable argument about who’s in charge. Nevertheless, it appears to me that disruption was not essential and probably not wished.
The EDA business spent lots of time and assets on the event of digital system degree (ESL) instruments and flows. This was primarily based on analysis that had began fairly a very long time earlier than and promised improved productiveness, and the flexibility to deal with extra advanced techniques. It got here at about the identical time because the IP business was taking off. This was a battle between top-down and bottom-up flows, and everyone knows which one succeeded. Nevertheless, it could have solely delayed the inevitable, as a result of an growing variety of corporations at the moment are wanting extra top-down focus.
Twenty years in the past, verification modified from human-in-the-loop to model-in-the-loop. This was precipitated by constrained random check sample technology. It definitely disrupted the way in which during which verification occurred within the business, and it virtually disrupted the EDA corporations. A brand new entrant rapidly took the lead and left huge EDA fairly a manner behind. However the incumbents acknowledged what was taking place and mainly thwarted the upstart by creating an ordinary that marginalized the benefit the startup had. That was the creation of SystemVerilog. Precisely the identical factor occurred a few years in the past with the creation of PSS. They have been technical disruptions however didn’t trigger a enterprise disruption just like the change to RTL did.
At the moment, there may be lots of buzz round 2.5D and 3D design, however it’s not clear if that may trigger disruptions inside EDA. It definitely will disrupt the packaging business, and we see foundries and OSATs preventing over the newly created territory. It could additionally disrupt the IP business, making silicon IP extra enticing than mushy IP. With that may come a major enterprise change. However it’s unlikely to disrupt EDA, which cautiously is creating the mandatory items of know-how that could be required. I say cautiously as a result of it’s not clear but if the semiconductor business will demand the whole lot that’s being thought-about. For instance, is place-and-route throughout stacked vertical dies wanted within the close to future?
Lots of people are speaking about AI, and it’s unclear for the time being if AI is best than the normal algorithms which have been optimized over the previous 30 years. Actually, if AI might be skilled on all the info that was used to make these optimizations, then it could in all probability be higher, however that knowledge is not obtainable. Solely a really small fraction of it’s, and to displace these algorithms it must be so much higher. When these instruments are re-factored, which appears to occur each 10 or 15 years, I’m positive they are going to take a look at including it. We additionally see lots of AI getting used to switch folks within the loop for optimization flows. The large query right here is, what’s the finish worth they supply? In case you want extra licenses, extra compute energy, and doubtless extra time, how a lot enchancment do you get for that? It is a switch of prices from manufacturing to growth, and whereas this will likely work for high-volume gadgets, it’s much less enticing to smaller manufacturing runs.
Disruption occurs for a motive — somebody will get a substantial acquire. This can be from a brand new market or new product alternative. May a change in EDA create a brand new market? Probably, if they may cut back growth prices by 10X or extra, however that’s solely a tiny fraction of the entire growth value. Wouldn’t it create a market 10X the dimensions it’s at present? I doubt it. Low-cost EDA has been tried earlier than and failed. 10X higher EDA could stand extra probability.
