[HN Gopher] How AlphaChip transformed computer chip design
       ___________________________________________________________________
        
       How AlphaChip transformed computer chip design
        
       Author : isof4ult
       Score  : 175 points
       Date   : 2024-09-27 15:56 UTC (7 hours ago)
        
 (HTM) web link (deepmind.google)
 (TXT) w3m dump (deepmind.google)
        
       | yeahwhatever10 wrote:
       | Why do they keep saying "superhuman"? Algorithms are used for
       | these tasks, humans aren't laying out trillions of transistors by
       | hand.
        
         | epistasis wrote:
         | Google is good at many things, but perhaps their strongest
         | skill is media positioning.
        
           | jonas21 wrote:
           | I feel like they're particularly bad at this, especially
           | compared to other large companies.
        
             | pinewurst wrote:
             | Familiarity breeds contempt. They've been pushing the
             | Google==Superhuman thing since the Internet Boom with
             | declining efficacy.
        
           | lordswork wrote:
           | The media hates Google.
        
             | epistasis wrote:
             | It a love/hate relationship. Which benefits Google and the
             | media greatly.
        
         | jeffbee wrote:
         | This is floorplanning the blocks, not every feature. We are
         | talking dozens to hundreds of blocks, not billions-trillions of
         | gates and wires.
         | 
         | I assume that the human benchmark is a human using existing EDA
         | tools, not a guy with a pocket protector and a roll of tape.
        
           | yeahwhatever10 wrote:
           | Floorplanning algorithms and solvers already exist https://li
           | msk.ece.gatech.edu/course/ece6133/slides/floorplan...
        
             | jeffbee wrote:
             | The original paper from DeepMind evaluates what they are
             | now calling AlphaChip versus existing optimizers, including
             | simulated annealing. They conclude that AlphaChip
             | outperforms them with much less compute and real time.
             | 
             | https://www.cl.cam.ac.uk/~ey204/teaching/ACS/R244_2021_2022
             | /...
        
               | foobarian wrote:
               | Randomized algorithms strike again!
        
               | sudosysgen wrote:
               | This is moreso amortized optimization/reinforcement
               | learning, not randomized algorithms.
        
               | hulitu wrote:
               | > They conclude that AlphaChip outperforms them with much
               | less compute and real time.
               | 
               | Of course they do. I'm waiting for their products.
        
         | fph wrote:
         | My state-of-art bubblesort implementation is also superhuman at
         | sorting numbers.
        
           | xanderlewis wrote:
           | Nice. Do you offer API access for a monthly fee?
        
             | int0x29 wrote:
             | I'll need 7 5 gigawatt datacenters in the middle of major
             | urban areas or we might lose the Bubble Sort race with the
             | Chinese.
        
               | gattr wrote:
               | Surely a 1.21-GW datacenter would suffice!
        
               | therein wrote:
               | Have we decided when are we deprecating it? I'm already
               | cultivating another team in a remote location to work on
               | a competing product that we will include into Google
               | Cloud a month before deprecating this one.
        
               | dgacmu wrote:
               | Surely you'll be able to reduce this by getting TSMC to
               | build new fabs to construct your new Bubble Sort
               | Processors (BSPs).
        
           | HPsquared wrote:
           | Nice. Still true though! We are in the bubble sort era of AI.
        
         | jayd16 wrote:
         | "superhuman or comparable"
         | 
         | What nonsense! XD
        
         | thomasahle wrote:
         | Believe it or not, but there was a time where algorithms were
         | worse than humans at layout out transistors. In particular at
         | the higher level design decisions.
        
         | negativeonehalf wrote:
         | Prior to AlphaChip, macro placement was done manually by human
         | engineers in any production setting. Prior algorithmic methods
         | especially struggled to manage congestion, resulting in chips
         | that weren't manufacturable.
        
       | mirchiseth wrote:
       | I must be old because first thing I thought reading AlphaChip was
       | why is deepmind talking about chips in DEC Alpha :-)
       | https://en.wikipedia.org/wiki/DEC_Alpha.
        
         | mdtancsa wrote:
         | haha, same!
        
         | sedatk wrote:
         | I first used Windows NT on a PC with a DEC Alpha AXP CPU.
        
       | dreamcompiler wrote:
       | Looks like this is only about placement. I wonder if it can be
       | applied to routing?
        
         | amelius wrote:
         | Exactly what I was thinking.
         | 
         | Also: when is this coming to KiCad? :)
         | 
         | PS: It would also be nice to apply a similar algorithm to graph
         | drawing (e.g. trying to optimize for human readability instead
         | of electrical performance).
        
           | tdullien wrote:
           | The issue is that in order to optimize for human readability
           | you'll need a huge number of human evaluations of graphs?
        
             | amelius wrote:
             | Maybe start with minimization of some metric based on
             | number of edge crossings, edge lengths and edge bends?
        
       | hinkley wrote:
       | TSMC made a point of calling out that their latest generation of
       | software for automating chip design has features that allow you
       | to select logic designs for TDP over raw speed. I think that's
       | our answer to keep Dennard scaling alive in spirit if not in
       | body. Speed of light is still going to matter, so physical
       | proximity of communicating components will always matter, but I
       | wonder how many wins this will represent versus avoiding thermal
       | throttling.
        
       | mikewarot wrote:
       | I understand the achievement, but can't square it with my belief
       | that uniform systolic arrays will prove to be the best general
       | purpose compute engine for neural networks. Those are almost
       | trivial to route, by nature.
        
         | ilaksh wrote:
         | Isn't this already the case for large portions of GPUs? Like,
         | many of the blocks would be systolic arrays?
         | 
         | I think the next step is arrays of memory-based compute.
        
           | mikewarot wrote:
           | Imagine a bit level systolic array. Just a sea of LUTs, with
           | latches to allow the magic of graph coloring to remove all
           | timing concerns by clocking everything in 2 phases.
           | 
           | GPUs still treat memory as separate from compute, they just
           | have wider bottlenecks than CPUs.
        
       | pfisherman wrote:
       | Questions for those in the know about chip design. How are they
       | measuring the quality of a chip design? Does the metric that
       | Google is reporting make sense? Or is it just something to make
       | themselves look good?
       | 
       | Without knowing much, my guess is that "quality" of a chip design
       | is multifaceted and heavily dependent on the use case. That is
       | the ideal chip for a data center would look very different from
       | those for a mobile phone camera or automobile.
       | 
       | So again what does "better" mean in the context of this
       | particular problem / task.
        
         | q3k wrote:
         | This is just floorplanning, which is a problem with fairly well
         | defined quality metrics (max speed and chip area used).
        
           | Drunk_Engineer wrote:
           | Oh man, if only it were that simple. A floorplanner has to
           | guestimate what the P&R tools are going to do with the
           | initial layout. That can be very hard to predict -- even if
           | the floorplanner and P&R tool are from the same vendor.
        
         | Drunk_Engineer wrote:
         | I have not read the latest paper, but their previous work was
         | really unclear about metrics being used. Researchers trying to
         | replicate results had a hard time getting reliable
         | details/benchmarks out of Google. Also, my recollection is that
         | Google did not even compute timing, just wirelength and
         | congestion; i.e. extremely primitive metrics.
         | 
         | Floorplanning/placement/synthesis is a billion dollar industry,
         | so if their approach were really revolutionary they would be
         | selling the technology, not wasting their time writing blog
         | posts about it.
        
           | IshKebab wrote:
           | > Floorplanning/placement/synthesis is a billion dollar
           | industry
           | 
           | Maybe all together, but I don't think automatic placement
           | algorithms are a billion dollar industry. There's so much
           | more to it than that.
        
             | Drunk_Engineer wrote:
             | Yes in combination. Customers generally buy these tools as
             | a package deal. If the placer/floorplanner blows everything
             | else out of the water, then a CAD vendor can upsell a lot
             | of related tools.
        
           | negativeonehalf wrote:
           | The original paper reports P&R metrics (WNS, TNS, area,
           | power, wirelength, horizontal congestion, vertical
           | congestion) -
           | https://www.nature.com/articles/s41586-021-03544-w
           | 
           | (no paywall): https://www.cl.cam.ac.uk/~ey204/teaching/ACS/R2
           | 44_2021_2022/...
        
             | Drunk_Engineer wrote:
             | From what I saw in the rebuttal papers, the Google cost-
             | function is wirelength based. You can still get good TNS
             | from that if your timing is very simplistic -- or if you
             | choose your benchmark carefully.
        
               | negativeonehalf wrote:
               | They optimize using a fast heuristic based on wirelength,
               | congestion, and density, but they evaluate with full P&R.
               | It is definitely interesting that they get good timing
               | without explicitly including it in their reward function!
        
       | ilaksh wrote:
       | How far are we from memory-based computing going from research
       | into competitive products? I get the impression that we are
       | already well passed the point where it makes sense to invest very
       | aggressively to scale up experiments with things like memristors.
       | Because they are talking about how many new nuclear reactors they
       | are going to need just for the AI datacenters.
        
         | HPsquared wrote:
         | And think of the embedded applications.
        
         | sroussey wrote:
         | The problem is that the competition (our current von neumann
         | architecture) has billions of dollars of R&D per year invested.
         | 
         | Better architectures without the yearly investment train will
         | no longer be better quite quickly.
         | 
         | You would need to be 100x to 1000x better in order to pull the
         | investment train onto your tracks.
         | 
         | Don't has been impossible for decades.
         | 
         | Even so, I think we will see such a change in my lifetime.
         | 
         | AI could be that use case that has a strong enough demand pull
         | to make it happen.
         | 
         | We will see.
        
           | ilaksh wrote:
           | I think it's just ignorance and timidity on the part of
           | investors. Memristor or memory-computing startups are surely
           | the next trend in investing within a few years.
           | 
           | I don't think it's necessarily demand or any particular
           | calculation that makes things happen. I think people
           | including investors are just herd animals. They aren't
           | enthusiastic until they see the herd moving and then they
           | want in.
        
             | foota wrote:
             | I don't think it's ignorant to not invest in something that
             | has a decade long path towards even having a market, much
             | less a large market.
        
       | colesantiago wrote:
       | A marvellous achievement from DeepMind as usual, I am quite
       | surprised that Google acquired them for a significant discount of
       | $400M, when I would have expected it to be in the range of $20BN,
       | but then again Deepmind wasn't making any money back then.
        
         | dharma1 wrote:
         | it was very early. probably one of their all time best
         | acquisitions in addition to YouTube.
         | 
         | Re:using RL and other types AI assistance for chip design,
         | Nvidia and others are doing this too
        
           | sroussey wrote:
           | Applied Semantics for $100m which gave them their advertising
           | business seems like their best deal.
        
       | amelius wrote:
       | Can this be abstracted and generalized into a more generally
       | applicable optimization method?
        
       | loandbehold wrote:
       | Every generation of chips is used to design next generation. That
       | seems to be the root of exponential growth in Moore's law.
        
         | bgnn wrote:
         | That's wrong. Chip design and Moore's law have nothing to do
         | with each other.
        
           | smaddox wrote:
           | To clarify what the parent is getting at: Moore's law is an
           | observation about the density (and, really about the cost) of
           | transistors. So it's about the fabrication process, not about
           | the logic design.
           | 
           | Practically speaking, though, maintaining Moore's law would
           | have been economically prohibitive if circuit design and
           | layout had not been automated.
        
         | negativeonehalf wrote:
         | Definitely a big part of it. Chips enable better EDA tools,
         | which enable better chips. First it was analytic solvers and
         | simulated annealing, now ML. Exciting times!
        
       | red75prime wrote:
       | I hope I'll still be alive when they'll announce AlephZero.
        
       | idunnoman1222 wrote:
       | So one other designer plus Google is using alpha chip for their
       | layouts? - not sure on that title, call me when amd and nvidia
       | are using it
        
       | kayson wrote:
       | I'm pretty sure Cadence and Synopsys have both released
       | reinforcement-learning-based placing and floor planning tools.
       | How do they compare...?
        
         | hulitu wrote:
         | They don't. You cannot compare reality (Cadence, Synopsys) with
         | hype (Google).
        
           | pelorat wrote:
           | So you're basically saying that Google should have used
           | existing tools to layout their chip designs, instead of their
           | ML solution, and that these existing tools would have
           | produced even better chips than the ones they are actually
           | manufacturing?
        
         | RicoElectrico wrote:
         | Synopsys tools can use ML, but not for the layout itself,
         | rather tuning variables that go into the physical design flow.
         | 
         | > Synopsys DSO.ai autonomously explores multiple design spaces
         | to optimize PPA metrics while minimizing tradeoffs for the
         | target application. It uses AI to navigate the design-
         | technology solution space by automatically adjusting or fine-
         | tuning the inputs to the design (e.g., settings, constraints,
         | process, flow, hierarchy, and library) to find the best PPA
         | targets.
        
       | cobrabyte wrote:
       | I'd love a tool like this for PCB design/layout
        
         | onjectic wrote:
         | First thing my mind went to as well, I'm sure this is already
         | being worked on, I think it would be more impactful than even
         | this.
        
           | bgnn wrote:
           | why do you think that?
        
             | bittwiddle wrote:
             | Far more people / companies are designing PCBs than there
             | are designing custom chips.
        
               | foota wrote:
               | I think the real value would be in ease of use. I imagine
               | the top N chip creators represent a fair bit of the
               | marginal value in pushing the state of the art forward.
               | E.g., for hobbyists or small shops, there's likely not
               | much value in tiny marginal improvements, but for the big
               | ones it's worth the investment.
        
       | abc-1 wrote:
       | Why aren't they using this technique to design better transformer
       | architectures or completely novel machine learning architectures
       | in general? Are plain or mostly plain transformers really peak? I
       | find that hard to believe.
        
         | jebarker wrote:
         | Because chip placement and the design of neural network
         | architectures are entirely different problems, so this solution
         | won't magically transfer from one to the other.
        
           | abc-1 wrote:
           | And AlphaGo is trained to play Go? The point is training a
           | model through self play to build neural network
           | architectures. If it can play Go and architect chip
           | placements, I don't see why it couldn't be trained to build
           | novel ML architectures.
        
       | vighneshiyer wrote:
       | This work from Google (original Nature paper:
       | https://www.nature.com/articles/s41586-021-03544-w) has been
       | credibly criticized by several researchers in the EDA CAD
       | discipline. These papers are of interest:
       | 
       | - A rebuttal by a researcher within Google who wrote this at the
       | same time as the "AlphaChip" work was going on ("Stronger
       | Baselines for Evaluating Deep Reinforcement Learning in Chip
       | Placement"): http://47.190.89.225/pub/education/MLcontra.pdf
       | 
       | - The 2023 ISPD paper from a group at UCSD ("Assessment of
       | Reinforcement Learning for Macro Placement"):
       | https://vlsicad.ucsd.edu/Publications/Conferences/396/c396.p...
       | 
       | - A paper from Igor Markov which critically evaluates the
       | "AlphaChip" algorithm ("The False Dawn: Reevaluating Google's
       | Reinforcement Learning for Chip Macro Placement"):
       | https://arxiv.org/pdf/2306.09633
       | 
       | In short, the Google authors did not fairly evaluate their RL
       | macro placement algorithm against other SOTA algorithms: rather
       | they claim to perform better than a human at macro placement,
       | which is far short of what mixed-placement algorithms are capable
       | of today. The RL technique also requires significantly more
       | compute than other algorithms and ultimately is learning a
       | surrogate function for placement iteration rather than learning
       | any novel representation of the placement problem itself.
       | 
       | In full disclosure, I am quite skeptical of their work and wrote
       | a detailed post on my website: https://vighneshiyer.com/misc/ml-
       | for-placement/
        
         | s-macke wrote:
         | When I first read about AlphaChip yesterday, my first question
         | was how it compares to other optimization algorithms such as
         | genetic algorithms or simulated annealing. Thank you for
         | confirming that my questions are valid.
        
         | gdiamos wrote:
         | Criticism is an important part of the scientific process.
         | 
         | Whichever approach ends up winning is improved by careful
         | evaluation and replication of results
        
         | jeffbee wrote:
         | It seems like this is multiple parties pursuing distinct
         | arguments. Is Google saying that this technique is applicable
         | in the way that the rebuttals are saying it is not? When I read
         | the paper and the update I did not feel as though Google
         | claimed that it is general, that you can just rip it off and
         | run it and get a win. They trained it to make TPUs, then they
         | used it to make TPUs. The fact that it doesn't optimize
         | whatever "ibm14" is seems beside the point.
        
         | Workaccount2 wrote:
         | To be fair, some of these criticisms are a few years old. Which
         | normally would be fair game, but the progress in AI has been
         | breakneck. Criticism of other AI tech from 2021 or 2022 are
         | pretty dated today.
        
           | jeffbee wrote:
           | It certainly looks like the criticism at the end of the
           | rebuttal that DeepMind has abandoned their EDA efforts is a
           | bit stale in this context.
        
         | porphyra wrote:
         | The Deepmind chess paper was also criticized for unfair
         | evaluation, as they were using an older version of Stockfish
         | for comparison. Apparently, the gap between AlphaZero and that
         | old version of Stockfish (about 50 elo iirc) was about the same
         | as the gap between consecutive versions of Stockfish.
        
         | nemonemo wrote:
         | What is your opinion of the addendum? I think the addendum and
         | the pre-trained checkpoint are the substance of the
         | announcement, and it is surprising to see little mention of
         | those here.
        
         | negativeonehalf wrote:
         | FD: I have been following this whole thing for a while, and
         | know personally a number of the people involved.
         | 
         | The AlphaChip authors address criticism in their addendum, and
         | in a prior statement from the co-lead authors:
         | https://www.nature.com/articles/s41586-024-08032-5 ,
         | https://www.annagoldie.com/home/statement
         | 
         | - The 2023 ISPD paper didn't pre-train at all. This means no
         | learning from experience, for a learning-based algorithm. I
         | feel like you can stop reading there.
         | 
         | - The ISPD paper and the MLcontra paper both used much larger
         | older technology node sizes, which have pretty different
         | physical properties. TPU has a sub 10nm technology node size,
         | whereas ISPD uses 45nm and 12nm. These are really different
         | from a physical design perspective. Even worse, MLcontra uses a
         | truly ancient benchmark with >100nm technology node size.
         | 
         | Markov's paper just summarizes the other two.
         | 
         | (Incidentally, none of ISPD / MLcontra / Markov were peer
         | reviewed - ISPD 2023 was an invited paper.)
         | 
         | There's a lot of other stuff wrong with the ISPD paper and the
         | MLcontra paper - happy to go into it - and a ton of weird
         | financial incentives lurking in the background. Commercial EDA
         | companies do NOT want a free open-source tool like AlphaChip to
         | take over.
         | 
         | Reading your post, I appreciate the thoroughness, but it seems
         | like you are too quick to let ISPD 2023 off the hook for
         | failing to pre-train and using less compute. The code for pre-
         | training is just the code for training --- you train on some
         | chips, and you save and reuse the weights between runs. There's
         | really no excuse for failing to do this, and the original
         | Nature paper described at length how valuable pre-training was.
         | Given how different TPU is from the chips they were evaluating
         | on, they should have done their own pre-training, regardless of
         | whether the AlphaChip team released a pre-trained checkpoint on
         | TPU.
         | 
         | (Using less compute isn't just about making it take longer -
         | ISPD 2023 used half as many GPUs and 1/20th as many RL
         | experience collectors, which may screw with the dynamics of the
         | RL job. And... why not just match the original authors'
         | compute, anyway? Isn't this supposed to be a reproduction
         | attempt? I really do not understand their decisions here.)
        
           | clickwiseorange wrote:
           | Oh, man... this is the same old stuff from the 2023 Anna
           | Goldie statement (is this Anna Goldie's comment?). This was
           | all addressed by Kahng in 2023 - no valid criticisms. Where
           | do I start?
           | 
           | Kahng's ISPD 2023 paper is not in dispute - no established
           | experts objected to it. The Nature paper is in dispute.
           | Dozens of experts objected to it: Kahng, Cheng, Markov,
           | Madden, Lienig, Swartz objected publically.
           | 
           | The fact that Kahng's paper was invited doesn't mean it
           | wasn't peer reviewed. I checked with ISPD chairs in 2023 -
           | Kahng's paper was thoroughly reviewed and went through
           | multiple rounds of comments. Do you accept it now? Would you
           | accept peer-reviewed versions of other papers?
           | 
           | Kahng is the most prominent active researcher in this field.
           | If anyone knows this stuff, it's Kahng. There were also five
           | other authors in that paper, including another celebrated
           | professor, Cheng.
           | 
           | The pre-training thing was disclaimed in the Google release.
           | No code, data or instructions for pretraining were given by
           | Google for years. The instructions said clearly: you can get
           | results comparable to Nature without pre-training.
           | 
           | The "much older technology" is also a bogus issue because the
           | HPWL scales linearly and is reported by all commercial tools.
           | Rectangles are rectangles. This is textbook material. But
           | Kahng etc al prepared some very fresh examples, including
           | NVDLA, with two recent technologies. Guess what, RL did
           | poorly on those. Are you accepting this result?
           | 
           | The bit about financial incentives and open-source is
           | blatantly bogus, as Kahng leads OpenROAD - the main open-
           | source EDA framework. He is not employed by any EDA
           | companies. It is Google who has huge incentives here, see
           | Demis Hassabis tweet "our chips are so good...".
           | 
           | The "Stronger Baselines" matched compute resources exactly.
           | Kahng and his coauthors performed _fair comparisons_ between
           | annealing and RL, giving the same resources to each. Giving
           | greater resources is unlikely to change results. This was
           | thoroughly addressed in Kahng 's FAQ - if you only could read
           | that.
           | 
           | The resources used by Google were huge. Cadence tools in
           | Kahng's paper ran hundreds times faster and produced better
           | results. That is as conclusive as it gets.
           | 
           | It doesn't take a Ph.D. to understand fair comparisons.
        
             | negativeonehalf wrote:
             | For AlphaChip, pre-training is just training. You train,
             | and save the weights in between. This has always been
             | supported by the Google's open-source repository. I've read
             | Kahng's FAQ, and he fails to address this, which is
             | unsurprising, because there's simply no excuse for cutting
             | out pre-training for a learning-based method. In his setup,
             | every time AlphaChip sees a new chip, he re-randomizes the
             | weights and makes it learn from scratch. This is obviously
             | a terrible move.
             | 
             | HPWL (half-perimeter wirelength) is an approximation of
             | wirelength, which is only one component of the chip
             | floorplanning objective function. It is relatively easy to
             | crunch all the components together and optimize HPWL ---
             | minimizing actual wirelength while avoiding congestion
             | issues is much harder.
             | 
             | Simulated annealing is good at quickly converging on a bad
             | solution to the problem, with relatively little compute. So
             | what? We aren't compute-limited here. Chip design is a
             | lengthy, expensive process where even a few-percent
             | wirelength reduction can be worth millions of dollars. What
             | matters is the end result, and ML has SA beat.
             | 
             | (As for conflict of interest, my understanding is Cadence
             | has been funding Kahng's lab for years, and Markov's
             | LinkedIn says he works for Synopsis. Meanwhile, Google has
             | released a free, open-source tool.)
        
               | clickwiseorange wrote:
               | It's not that one needs an excuse. The Google CT repo
               | said clearly you don't need to pretrain. "supported"
               | usually includes at least an illustration - no such thing
               | there before Kahng's paper.
               | 
               | Everything optimized in Nature RL is an approximation.
               | HPWL is where you start, and RL uses it in the objective
               | function too. As shown in "Stronger Baselines", RL loses
               | a lot by HPWL - so much that nothing else can save it. If
               | your wires are very long, you need routing tracks to
               | route them, and you end up with congestion too.
               | 
               | SA consistently produces better solutions than RL for
               | various time budgets. That's what matters. Both papers
               | have shown that SA produces competent solutions. You give
               | SA more time, you get better solutions. In a fair
               | comparison, you give equal budgets to SA and RL. RL
               | loses. This was confirmed using Google's RL code and two
               | independent SA implementations, on many circuits. Very
               | definitively. No, ML did not have SA beat - please read
               | the papers.
               | 
               | Cadence hasn't funded Kahng for a long time. In fact,
               | Google funded Kahng more recently, so he has all the
               | incentives to support Google. Markov's LinkedIn page says
               | he worked at Google before. Even Chatterjee, of all
               | people, worked at Google.
               | 
               | Google's open-source tool is a head fake, it's
               | practically unusable.
               | 
               | Update: I'll respond to the next comment here since
               | there's no Reply button.
               | 
               | 1. The Nature paper said one thing, the code did
               | something else, as we've discovered. The RL method does
               | some training as it goes. So, pre-training is not the
               | same as training. Hence "pre". Another problem with
               | pretraining in Google work is data contamination - we
               | can't compare test and training data. The Google folks
               | admitted to training and testing on different versions of
               | the same design. That's bad. Rejection-level bad.
               | 
               | 2. HPWL is indeed a nice simple objective. So nice that
               | Jeff Dean's recent talks use it. And all commercial
               | placers without exception optimize it and report it. All
               | EDA publications report it. Google's RL optimized HPWL +
               | density + congestion
               | 
               | 3. This shows you aren't familiar with EDA. Simulated
               | Annealing was the king of placement from mid 1980s to mid
               | 1990s. Most chips were placed by SA. But you don't have
               | to go far - as I recall, the Nature paper says they used
               | SA to postprocess macro placements.
        
               | negativeonehalf wrote:
               | The Nature paper describes the importance of pre-training
               | repeatedly. The ability to learn from experience is the
               | whole point of the method. Pre-training is just training
               | and saving the weights -- this is ML 101.
               | 
               | I'm glad you agree that HPWL is a proxy metric.
               | Optimizing HPWL is a fun applied math puzzle, but it's
               | not chip design.
               | 
               | I am unaware of a single instance of someone using SA to
               | generate real-world, usable macro layouts that were
               | actually taped out, much less for modern chip design, in
               | part due to SA's struggles to manage congestion,
               | resulting in unusable layouts. SA converges quickly to a
               | bad solution, but this is of little practical value.
        
               | sijnapq wrote:
               | Google can shove open-source in its a$$! Evidence will
               | appear gradually. Tell your wife to be ready.
        
           | sijnapq wrote:
           | Gabe. Shut up! Your field is not even hardware.
        
         | smokel wrote:
         | I don't really understand all the fuss about this particular
         | paper. Nearly _all_ papers on AI techniques are pretty much
         | impossible to reproduce, due to details that the authors don 't
         | understand or are trying to cover up.
         | 
         | This is what you get if you make academic researchers compete
         | for citation counts.
         | 
         | Pretraining seems to be an important aspect here, and it makes
         | sense that such pretraining requires good examples, which
         | unfortunately for the free lunch people, is not available to
         | the public.
         | 
         | That's what you get when you let big companies do fundamental
         | research. Would it be better if the companies did not publish
         | anything about their research at all?
         | 
         | It all feels a bit unproductive to attack one another.
        
         | bsder wrote:
         | EDA claims in the digital domain are fairly easy to evaluate.
         | Look at the picture of the layout.
         | 
         | When you see a chip that has the datapath _identified_ and laid
         | out properly by a computer algorithm, you 've got something. If
         | not, it's vapor.
         | 
         | So, if your layout still looks like a random rat's nest? Nope.
         | 
         | If even a random person can see that your layout actually
         | follows the obvious symmetric patterns from bit 0 to bit 63,
         | maybe you've got something worth looking at.
         | 
         | Analog/RF is a little tougher to evaluate because the smaller
         | number of building blocks means you can use Moore's Law to
         | brute force things much more exhaustively, but if things "looks
         | pretty" then you've got something. If it looks weird, you
         | don't.
        
           | glitchc wrote:
           | That doesn't mean the fabricated netlist doesn't work. I'm
           | not supporting Google by any means, but the test should be:
           | Does it fabricate and function as intended? If not, clearly
           | gibberish. If so, we now have computers building computers,
           | which is one step closer to SkyNet. The truth is probably
           | somewhere in between. But even if some of the samples, with
           | the terrible layouts, are actually functional, then we might
           | learn something new. Maybe the gibberish design has reduced
           | crosstalk, which would be fascinating.
        
       | lordswork wrote:
       | Some interesting context on this work: 2 researchers were bullied
       | to the point of leaving Google for Anthropic by a senior
       | researcher (who has now been terminated himself):
       | https://www.wired.com/story/google-brain-ai-researcher-fired...
       | 
       | They must feel vindicated by their work turning out to be so
       | fruitful now.
        
       | FrustratedMonky wrote:
       | So AI designing it's own chips. Now that is moving towards
       | exponential growth. Like at the end of "Colossus" the movie.
       | 
       | Forget LLM's. What DeepMind is doing seems more like how an AI
       | will rule, in the world. Building real world models, and applying
       | game logic like winning.
       | 
       | LLM's will just be the text/voice interface to what DeepMind is
       | building.
        
       | DrNosferatu wrote:
       | Yet, their "frontier" LLM lags all the others...
        
       | 7e wrote:
       | Did it, though? Google's chips still aren't very good compared
       | with competitors.
        
       | ninetyninenine wrote:
       | What occupation is there that is purely intellectual that has no
       | chance of an AI ever progressing to a point where it can take it
       | over?
        
         | alexyz12 wrote:
         | anything that needs very real-time info. AI's will always be
         | limited by us feeding them info, or them collecting it
         | themselves. But humans can travel to more places than an AI
         | can, until robots are everywhere too I suppose
        
       | QuadrupleA wrote:
       | How good are TPUs in comparison with state of the art Nvidia
       | datacenter GPUs, or Groq's ASICs? Per watt, per chip, total cost,
       | etc.? Is there any published data?
        
         | jeffbee wrote:
         | MLPerf is a good place to start. The only problem is you don't
         | have any verifiable information about TPU energy consumption.
         | https://mlcommons.org/benchmarks/inference-datacenter/
        
         | wslh wrote:
         | I have some company notes from early 2024 which cannot be
         | accurate but could help,
         | 
         | TPU v5e [1]: not available for purchase, only through GCP,
         | storage=5B, LLM-Model=7B, efficiency=393TFLOP.
         | 
         | [1] https://cloud.google.com/tpu/docs/v5e
        
       | ur-whale wrote:
       | Seems to me the article is claiming a lot of things, but is very
       | light on actual comparisons that matter to you and me, namely:
       | how does one of those fabled AI-designed chop compare to their
       | competition ?
       | 
       | For example, how much better are these latest gen TPU's when
       | compared to NVidia's equivalent offering ?
        
       | thesz wrote:
       | Eurisco [1], if I remember correctly, was once used to perform
       | placement-and-route task and was pretty good at it.
       | 
       | [1] https://en.wikipedia.org/wiki/Eurisko
       | 
       | What's more, Eurisco was then used in designing Traveler TCS'
       | game fleet of battle spaceships. And Eurisco used symmetry-based
       | placement learned from VLSI design in the design of the
       | spaceships' fleet.
       | 
       | Can AlphaChip's heuistics be used anywhere else?
        
       ___________________________________________________________________
       (page generated 2024-09-27 23:00 UTC)