News Chip design legend Jim Keller aims for Tenstorrent wins in markets 'not well served by Nvidia

[Admin]

CEO of Tenstorrent believes that although Nvidia dominates the market of AI processors, there is still a place for Tenstorrent.

Chip design legend Jim Keller aims for Tenstorrent wins in markets 'not well served by Nvidia : Read more

 

[-Fran-]

Are they going to produce cheap consumer GPUs?

Man... One can definitely dream, no?

Regards

 

[Taslios]

I've read a bunch of speculation that Sifive has a lot of connections with Intel and many expect it to eventually be acquired by them.

in interviews Jim speaks much more highly of his two stints at AMD than he does of his time at Intel. Now this may be because of the recency of his time at Intel and various NDA's he may still be under, but I have to wonder if he as the CEO might not be angling for the M&A route. AMD is buying up AI startups....

 

[dalauder]

It's hard to imagine that Jim Keller doesn't know what he's talking about--so much so that we've been talking about Tenstorrent for years with no real info or products. My guess is that he just talks big and sells this off--finally cash in his name. He's probably more important to computing than guys like Samueli, maybe more important than Jensen Huang. And those guys are all big billionaires. He doesn't even have enough to retire.

 

[bit_user]

I've read a bunch of speculation that Sifive has a lot of connections with Intel and many expect it to eventually be acquired by them.

Intel seems all-in on Habana/Gaudi. I don't see why they'd buy Tenstorrent, except to snuff out a competitor and keep it out of anyone else's hands. Right now, I think Intel's investors probably wouldn't look too kindly on that use of the company's funds. Their dividend still isn't back to where it used to be, and the company doesn't look free of headwinds.

in interviews Jim speaks much more highly of his two stints at AMD than he does of his time at Intel. Now this may be because of the recency of his time at Intel and various NDA's he may still be under,

He complained that Intel put him in a sort of CTO position, from what I recall. He was too far removed from where the actual work happened. So, he basically played evangelist, for a couple years, to try and spur innovation and technology development at Intel, and then I think he jumped when he decided that he did everything in that role that he could.

I have to wonder if he as the CEO might not be angling for the M&A route. AMD is buying up AI startups....

In the startup game, the the top prize is to IPO and then get acquired. If they'd be acquired now, it would come at too high a price for their current revenues. That's probably why neither they nor Cerebras have been snapped up.

As for AMD acquiring them, maybe it makes more sense. The Xilinx IP comprising XDNA is really inference-oriented. AMD is relying on GPUs to do training, but that could put them at a disadvantage vs. a purpose-built AI accelerator. If Jim is right about their cost, performance, and efficiency advantages, then it's quite plausible AMD would buy them.

 

[bit_user]

I wish TH would stop exaggerating and calling this boastful narcissist a "chip design legend". Yeah, he was a lead architect for AMD K8 and he co-authored Hypertransport but that's old stuff,

You forgot a few things:

• He co-architected the DEC Alpha 21264 - a 64-bit 4-way superscalar OoO CPU. It was the fastest & most complex architecture in the world, back in 1998.
• He spearheaded Apple's internal CPU core design effort.
• He basically headed up the effort to build Zen.
• He then went to Tesla to design their self-driving processor.

After that, he spent the time at Intel and then moved on to Tenstorrent.

Clearly someone at TH is a fangirl given how much space and praise he is getting.

Name someone more accomplished.

 

[bit_user]

This is not an honest rebuttal. Instead of countering with facts, you're countering with doubts. I will not debate you on this, because no matter what facts I present, it's clear that you've already made up your mind and changing it would be impossible.

Honestly, I'm okay with you believing whatever you want. All I'd say is that others should just beware not to take your comments at face value.

Not sure how he did that, because it was Daniel Dobberpuhl who founded PA Semi and received awards for low power designs which led to them being bought by Apple.

I do just have to clarify one thing: Keller went to Apple before the P.A. Semi acquisition! For all we know, it Keller actually had a hand in it.

Keller described the process of establishing the architecture of both Apple's and AMD's respective cores like this:

"So when you build a new computer, and Zen was a new computer, there was already work underway. You build in basically a roadmap, so I was thinking about what we were going to do for five years, chip after chip. We did this at Apple too when we built the first big core at Apple - we built big bones [into the design]. When you make a computer faster, there's two ways to do it - you make the fundamental structure bigger, or you tweak features, and Zen had a big structure. Then there were obvious things to do for several generations to follow. They've been following through on that."

Source: https://www.anandtech.com/show/16762/an-anandtech-interview-with-jim-keller-laziest-person-at-tesla
​

What he did? Cheerleading stunts?

If you honestly cared, you'd do pretty well to read the above interview. Not just skim it, but actually read it. I was skeptical of his rock star status, until I read that and a couple other of his interviews. He sounds pretty down to earth and really doesn't strike me as pretentious.

Given his resume, I don't think he's that over-hyped, but I'm sure you have your reasons for thinking otherwise. If you want to pick on an engineer for being over-hyped, I think Raja Koduri should be on the list. I can sure think of some non-technical people in business, media, sports, etc. who are way over-hyped. There are also certain youtubers, but let's not get off-track.

 

[thestryker]

He complained that Intel put him in a sort of CTO position, from what I recall. He was too far removed from where the actual work happened. So, he basically played evangelist, for a couple years, to try and spur innovation and technology development at Intel, and then I think he jumped when he decided that he did everything in that role that he could.

There were a lot of rumors to that effect, but from what I gather that absolutely wasn't the case. His job at Intel was extremely time intensive so it's quite believable that a personal reason could have him step away. I wouldn't be surprised if he was a driving factor for Intel keeping fabs instead of spinning or selling off. A lot of wall street investors have been screaming for it since around Ice Lake launch. I'm not convinced we've even seen his Intel impact in retail parts yet.

I think the only reason he's mostly just known in tech circles is that he always seems to want to be designing something new. It's really hard to do that at any publicly traded company so it seems like he makes his impact and moves along.

 

[bit_user]

Based on a cursory patent search (and assuming that's the same Jim Keller), he did some work on storage device to host communication with other people.

LOL - you literally searched for "Jim Keller"?? Try James B. Keller.

• Combined transparent/non-transparent cache - US US10241705B2 James B. Keller Apple Inc.
• Cache implementing multiple replacement policies - US US8719509B2 James B. Keller Apple Inc.
• Trust zone support in system on a chip having security enclave processor - US US8775757B2 James B. Keller Apple Inc.
• L1 flush mechanism to flush cache for power down and handle coherence during … - US US8171326B2 James B. Keller Apple Inc.
• Non-blocking address switch with shallow per agent queues - WO EP US CN JP AT DE ES TW US7970970B2 James B. Keller Apple Inc.
• Block-based non-transparent cache - WO EP US CN JP KR EP2452265B1 James B. Keller Apple Inc.
• Digital phase relationship lock loop - US US8078772B2 James B. Keller Apple Inc.
• Combined buffer for snoop, store merging, load miss, and writeback operations - US US7398361B2 James B. Keller P.A. Semi, Inc.
• Replay reduction for power saving - US US8255670B2 James B. Keller Apple Inc.
• Memory controller with loopback test interface - US US8301941B2 James B. Keller Apple Inc.
• High speed bus system that incorporates uni-directional point-to-point buses - US US7668997B2 James B. Keller Hewlett-Packard Development Company, L.P.
• Zero cycle move - US US9575754B2 James B. Keller Apple Inc.
• Electronic system and method of bandwidth management - WO US TW TWI512474B James B Keller Apple Inc
• Branch Predictor for Wide Issue, Arbitrarily Aligned Fetch - US US20160048395A1 James B. Keller Apple Inc.
• Uncacheable load merging - US US20080086594A1 James B. Keller P.A. Semi, Inc.
• System cache with data pending state and method for optimizing system cache - WO US TW TWI492055B James B Keller Apple Inc
• Instruction set architecture mode dependent sub-size access of register with … - US US9317285B2 James B. Keller Apple Inc.
• Receiving data from virtual channels - US US20070189299A1 James Keller Broadcom Corporation, A California Corporation
• Hierarchical Fabric Control Circuits - US US20120182902A1 James B. Keller Saund Gurjeet S
• Fabric limiter circuits - US US8744602B2 James B. Keller Apple Inc.
• Multi-Destination Instruction Handling - US US20140089638A1 James B. Keller Apple Inc.
• Segmented interconnect for connecting multiple agents in a system - WO EP US CN JP AT DE TW US7426601B2 James B. Keller P.A. Semi, Inc.
• Stacked memory device having a scalable bandwidth interface - US US8611127B1 James B. Keller Apple Inc.
• Memory controller with programmable configuration - EP US US6625685B1 James B. Keller Broadcom Corporation
• Implementing locks in a distributed processing system - US US6473849B1 James B. Keller Advanced Micro Devices, Inc.
• Circuit and method for maintaining order of memory access requests initiated by … - US US6167492A James B. Keller Advanced Micro Devices, Inc.
• Maintaining cache coherency during a memory read operation in a multiprocessing … - US US6490661B1 James B. Keller Advanced Micro Devices, Inc.
• Flexible probe/probe response routing for maintaining coherency - WO EP US JP KR BR DE EP1141839B1 James B. Keller Advanced Micro Devices, Inc.
• System and method for implementing a separate virtual channel for posted … - US US6950438B1 James B. Keller Advanced Micro Devices, Inc.
• System and method of maintaining coherency in a distributed communication … - US US7069361B2 James B. Keller Advanced Micro Devices, Inc.
• Scheduler capable of issuing and reissuing dependency chains - US US6542984B1 James B. Keller Advanced Micro Devices, Inc.
• Computer system implementing a system and method for tracking the progress of … - US US6721813B2 James B. Keller Advanced Micro Devices, Inc.
• Establishing an operating mode in a processor - US US7124286B2 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for optimizing the performance of LDxL and STxC interlock … - US US6141734A James Keller Compaq Computer Corporation
• Pipeline elements which verify predecode information - US US6502185B1 James B. Keller Advanced Micro Devices, Inc.
• Memory cancel response optionally cancelling memory controller's providing of … - US US6370621B1 James B. Keller Advanced Micro Devices, Inc.
• Messaging scheme to maintain cache coherency and conserve system memory … - US US6275905B1 James B. Keller Advanced Micro Devices, Inc.
• System having interfaces and switch that separates coherent and packet traffic - EP US AT DE US6941406B2 James B. Keller Broadcom Corporation
• Cache which provides partial tags from non-predicted ways to direct search if … - WO EP US CN JP KR DE EP1244970B1 James B. Keller Advanced Micro Devices, Inc.
• Double prefix overrides to provide 16-bit operand size in a 32/64 operating … - US US6560694B1 James B. Keller Advanced Micro Devices, Inc.
• Adaptive retry mechanism - EP US DE US6633936B1 James B. Keller Broadcom Corporation
• Programmably disabling one or more cache entries - EP US US6848024B1 James B. Keller Broadcom Corporation
• Store to load forward predictor training using delta tag - US US6622237B1 James B. Keller Advanced Micro Devices, Inc.
• Computer system implementing a system and method for ordering input/output (IO) … - US US6557048B1 James B. Keller Advanced Micro Devices, Inc.
• Circuit and method for selectively stalling interrupt requests initiated by … - US US6389526B1 James B. Keller Advanced Micro Devices, Inc.
• Store load forward predictor training - US US6694424B1 James B. Keller Advanced Micro Devices, Inc.
• Scheduler capable of issuing and reissuing dependency chains - WO EP CN JP KR DE EP1244962B1 James B. Keller Advanced Micro Devices, Inc.
• Response virtual channel for handling all responses - US US6888843B2 James B. Keller Advanced Micro Devices, Inc.
• Virtual channels and corresponding buffer allocations for deadlock-free … - WO EP US JP KR DE US6938094B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for developing multiprocessor cache control protocols by … - US US6397302B1 James B. Keller Compaq Information Technologies Group, L.P.
• Methods and apparatus for processing load instructions in the presence of RAM … - US US6374344B1 James B. Keller Compaq Information Technologies Group L.P. (Citg)
• Data cache having store queue bypass for out-of-order instruction execution and … - US US6360314B1 James B. Keller Compaq Information Technologies Group, L.P.
• Method and apparatus for determining availability of a queue to which a program … - US US7093105B2 James Keller Hewlett-Packard Development Company, L.P.
• Apparatus and method for replaying decoded instructions - US US5012403A James B. Keller Digital Equipment Corporation
• Method and apparatus for optimizing bcache tag performance by inferring bcache … - US US6401173B1 James B. Keller Compaq Information Technologies Group, L.P.
• Profile directed simulation used to target time-critical crossproducts during … - US US6212493B1 James B. Keller Compaq Computer Corporation
• Method and apparatus for maximizing utilization of an internal processor bus in … - US US5924120A James Keller Digital Equipment Corporation
• Method and apparatus for developing multiprocessore cache control protocols … - US US20010029574A1 James B. Keller Rahul Razdan
• Method and apparatus for resolving probes in multi-processor systems which do … - US US6295583B1 James B. Keller Compaq Information Technologies Group, L.P.
• Host bridge translating non-coherent packets from non-coherent link to coherent … - US US6714994B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for using a control signal on a packet based communication … - US US6748442B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for developing multiprocessor cache control protocols … - US US6349366B1 James B. Keller Compaq Information Technologies Group, L.P.
• Method and apparatus for minimizing pincount needed by external memory control … - US US6199153B1 James B. Keller Digital Equipment Corporation
• System and method for separate virtual channels for posted requests in a … - WO EP JP AU DE TW EP1314094B1 James B. Keller Advanced Micro Devices, Inc.
• Unload counter adjust logic for a receiver buffer - US US6434640B1 James B. Keller Advanced Micro Devices, Inc.
• Determination of execution resource allocation based on concurrently executable … - US US6704854B1 James B. Keller Advanced Micro Devices, Inc.
• Line predictor which caches alignment information - WO EP US JP KR US20040168043A1 James Keller Advanced Micro Devices, Inc.
• Method and apparatus for minimizing dcache index match aliasing using hashing … - US US6253285B1 James B. Keller Compaq Computer Corporation
• Method and apparatus for performing speculative memory fills into a … - US US6493802B1 James B. Keller Compaq Information Technologies Group, L.P.
• Conversation of distributed memory bandwidth in multiprocessor system with … - US US6393529B1 James B. Keller Advanced Micro Devices, Inc.
• Physical rename register for efficiently storing floating point, integer, … - US US6393546B1 James B. Keller Advanced Micro Devices, Inc.
• Computer system implementing flush operation - US US6553430B1 James B. Keller Advanced Micro Devices, Inc.
• Training line predictor for branch targets - US US6647490B2 James B. Keller Advanced Micro Devices, Inc.
• Method of controlling a shared memory bus in a multiprocessor system for … - US US5202973A James B. Keller Digital Equipment Corporation
• Computer system implementing system and method for ordering write operations … - US US6529999B1 James B. Keller Advanced Micro Devices, Inc.
• Apparatus and method for responding to an aborted signal exchange between … - WO EP US CN JP KR AU CA ES IL IN US4858173A James B. Keller Digital Equipment Corporation
• Line predictor entry with location pointers and control information for … - US US6546478B1 James B. Keller Advanced Micro Devices, Inc.
• Apparatus and method for providing a cache memory unit with a write operation … - WO EP US JP AU CA DE EP0292501B1 James B. Keller Digital Equipment Corporation
• Method and circuit for initializing a de-skewing buffer in a clock forwarded … - EP US AT DE US6952791B2 James B. Keller Broadcom Corporation
• Multiple processor integrated circuit having configurable packet-based … - US US20040019704A1 James Keller Barton Sano
• Conservation of system memory bandwidth and cache coherency maintenance using … - WO EP JP KR EP2320322A2 James Keller Advanced Micro Devices, Inc.
• Scheduler which discovers non-speculative nature of an instruction after … - US US6564315B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for balancing load vs. store access to a primary data … - US US6163821A James B. Keller Compaq Computer Corporation
• Predictor miss decoder updating line predictor storing instruction fetch … - US US6636959B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for developing multiprocessor cache control protocols … - US US6314496B1 James B. Keller Compaq Computer Corporation
• Method and apparatus for a dedicated physically indexed copy of the data cache … - US US6253301B1 James B. Keller Compaq Computer Corporation
• Method and apparatus for accelerated addition of sliced addends - WO US JP US4878193A James B. Keller Digital Equipment Corporation
• Implementing locks in a distributed processing system - WO EP US JP AU DE US7640315B1 James B. Keller Advanced Micro Devices, Inc.
• Apparatus and method for providing a settling time cycle for a system bus in a … - US US5029076A James B. Keller Digital Equipment Corporation
• Method and apparatus for developing multiprocessor cache control protocols … - US US6651144B1 James B. Keller Hewlett-Packard Development Company, L.P.
• Transmitting data from a plurality of virtual channels via a multiple processor … - US US20040017813A1 James Keller Manu Gulati
• Scheduler which retries load/store hit situations - US US6622235B1 James B. Keller Advanced Micro Devices, Inc.
• High speed bus system - EP JP AU CA DE EP0464708B1 James B. Keller Digital Equipment Corporation

That's just what I got from filtering through the first 250 hits. There are about 4 or 5 of those I listed which are just "James Keller", but most correspond to places he was known to work and the dates matched. Note that Compaq bought DEC, which explains why some of his Alpha patents are filed under that company name.

If your chosen quote is representative of that interview then rest assured I have no desire to read the rest of it -- he said a bunch of words without explaining anything, including his role in any of it. It simply can't get any more vague than this.

It was for a mass audience. He's not lecturing to graduate students or mentoring interns. Plus, the interview was already pretty long, as is. The deeper he got into the technical details would've expanded the scope exponentially.

Also, to fully appreciate what he said, you'd want to look at the surrounding context. I didn't included it in my quote, because it wasn't material to the point I was trying to make. However, if you're going to pick on the informational content of the quote, then you'll want to include the context.

Yeah, let's not -- you asked me to provide others who are more accomplished and I did, funny how you skipped responding to that.

I told you I didn't consider your reply to be serious and fact-based.

 

[bit_user]

Nothing on that list is revolutionary and worthy of the hype he's getting.

See? This is why it's pointless to debate this, with you. You've already made up your mind and will find some way to dismiss any evidence you're shown.

For whatever reason, maybe out of jealousy, I think you hate to see someone else in tech elevated like this. No matter that all other sectors of society do it all the time, and sports or media stars get like a billion followers on social media, while being about as smart as a bag of hammers and not really doing anything that people before them didn't already do. We're not even talking about mass media here, but I guess even being given a little notoriety, in the tech press, is too much.

I just have this question: do you hold other tech CEOs to the same standard? When someone does an interview with the Qualcomm CEO, and they talk about the technical & competitive landscape, do you bristle at their lack of technical accomplishments? Or do they get a "pass", having come from the business side and not rising through the ranks as an engineer?

I didn't ask for a deep technical dive -- I would have appreciated knowing on which subsystems he worked and contributed to.

It was a wide-ranging interview. It wasn't meant as a biography, but rather to get some tidbits about various projects he worked on and hear his take on numerous other topics. I think sitting down for a biographical interview would've been a far more vain exercise, wouldn't it?

So the people I named as examples with more accomplishments are imaginary?

This discussion lost its integrity when you decided to raise doubts to protect your beliefs, instead of seeking facts that would help us arrive at a greater understanding of the matter. You've predetermined the outcome, so why should I waste my time? I just picked a few points I wanted to address, but I really don't see any point in continuing this discussion.

 

[JTR17]

LOL - you literally searched for "Jim Keller"?? Try James B. Keller.

• Combined transparent/non-transparent cache - US US10241705B2 James B. Keller Apple Inc.
• Cache implementing multiple replacement policies - US US8719509B2 James B. Keller Apple Inc.
• Trust zone support in system on a chip having security enclave processor - US US8775757B2 James B. Keller Apple Inc.
• L1 flush mechanism to flush cache for power down and handle coherence during … - US US8171326B2 James B. Keller Apple Inc.
• Non-blocking address switch with shallow per agent queues - WO EP US CN JP AT DE ES TW US7970970B2 James B. Keller Apple Inc.
• Block-based non-transparent cache - WO EP US CN JP KR EP2452265B1 James B. Keller Apple Inc.
• Digital phase relationship lock loop - US US8078772B2 James B. Keller Apple Inc.
• Combined buffer for snoop, store merging, load miss, and writeback operations - US US7398361B2 James B. Keller P.A. Semi, Inc.
• Replay reduction for power saving - US US8255670B2 James B. Keller Apple Inc.
• Memory controller with loopback test interface - US US8301941B2 James B. Keller Apple Inc.
• High speed bus system that incorporates uni-directional point-to-point buses - US US7668997B2 James B. Keller Hewlett-Packard Development Company, L.P.
• Zero cycle move - US US9575754B2 James B. Keller Apple Inc.
• Electronic system and method of bandwidth management - WO US TW TWI512474B James B Keller Apple Inc
• Branch Predictor for Wide Issue, Arbitrarily Aligned Fetch - US US20160048395A1 James B. Keller Apple Inc.
• Uncacheable load merging - US US20080086594A1 James B. Keller P.A. Semi, Inc.
• System cache with data pending state and method for optimizing system cache - WO US TW TWI492055B James B Keller Apple Inc
• Instruction set architecture mode dependent sub-size access of register with … - US US9317285B2 James B. Keller Apple Inc.
• Receiving data from virtual channels - US US20070189299A1 James Keller Broadcom Corporation, A California Corporation
• Hierarchical Fabric Control Circuits - US US20120182902A1 James B. Keller Saund Gurjeet S
• Fabric limiter circuits - US US8744602B2 James B. Keller Apple Inc.
• Multi-Destination Instruction Handling - US US20140089638A1 James B. Keller Apple Inc.
• Segmented interconnect for connecting multiple agents in a system - WO EP US CN JP AT DE TW US7426601B2 James B. Keller P.A. Semi, Inc.
• Stacked memory device having a scalable bandwidth interface - US US8611127B1 James B. Keller Apple Inc.
• Memory controller with programmable configuration - EP US US6625685B1 James B. Keller Broadcom Corporation
• Implementing locks in a distributed processing system - US US6473849B1 James B. Keller Advanced Micro Devices, Inc.
• Circuit and method for maintaining order of memory access requests initiated by … - US US6167492A James B. Keller Advanced Micro Devices, Inc.
• Maintaining cache coherency during a memory read operation in a multiprocessing … - US US6490661B1 James B. Keller Advanced Micro Devices, Inc.
• Flexible probe/probe response routing for maintaining coherency - WO EP US JP KR BR DE EP1141839B1 James B. Keller Advanced Micro Devices, Inc.
• System and method for implementing a separate virtual channel for posted … - US US6950438B1 James B. Keller Advanced Micro Devices, Inc.
• System and method of maintaining coherency in a distributed communication … - US US7069361B2 James B. Keller Advanced Micro Devices, Inc.
• Scheduler capable of issuing and reissuing dependency chains - US US6542984B1 James B. Keller Advanced Micro Devices, Inc.
• Computer system implementing a system and method for tracking the progress of … - US US6721813B2 James B. Keller Advanced Micro Devices, Inc.
• Establishing an operating mode in a processor - US US7124286B2 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for optimizing the performance of LDxL and STxC interlock … - US US6141734A James Keller Compaq Computer Corporation
• Pipeline elements which verify predecode information - US US6502185B1 James B. Keller Advanced Micro Devices, Inc.
• Memory cancel response optionally cancelling memory controller's providing of … - US US6370621B1 James B. Keller Advanced Micro Devices, Inc.
• Messaging scheme to maintain cache coherency and conserve system memory … - US US6275905B1 James B. Keller Advanced Micro Devices, Inc.
• System having interfaces and switch that separates coherent and packet traffic - EP US AT DE US6941406B2 James B. Keller Broadcom Corporation
• Cache which provides partial tags from non-predicted ways to direct search if … - WO EP US CN JP KR DE EP1244970B1 James B. Keller Advanced Micro Devices, Inc.
• Double prefix overrides to provide 16-bit operand size in a 32/64 operating … - US US6560694B1 James B. Keller Advanced Micro Devices, Inc.
• Adaptive retry mechanism - EP US DE US6633936B1 James B. Keller Broadcom Corporation
• Programmably disabling one or more cache entries - EP US US6848024B1 James B. Keller Broadcom Corporation
• Store to load forward predictor training using delta tag - US US6622237B1 James B. Keller Advanced Micro Devices, Inc.
• Computer system implementing a system and method for ordering input/output (IO) … - US US6557048B1 James B. Keller Advanced Micro Devices, Inc.
• Circuit and method for selectively stalling interrupt requests initiated by … - US US6389526B1 James B. Keller Advanced Micro Devices, Inc.
• Store load forward predictor training - US US6694424B1 James B. Keller Advanced Micro Devices, Inc.
• Scheduler capable of issuing and reissuing dependency chains - WO EP CN JP KR DE EP1244962B1 James B. Keller Advanced Micro Devices, Inc.
• Response virtual channel for handling all responses - US US6888843B2 James B. Keller Advanced Micro Devices, Inc.
• Virtual channels and corresponding buffer allocations for deadlock-free … - WO EP US JP KR DE US6938094B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for developing multiprocessor cache control protocols by … - US US6397302B1 James B. Keller Compaq Information Technologies Group, L.P.
• Methods and apparatus for processing load instructions in the presence of RAM … - US US6374344B1 James B. Keller Compaq Information Technologies Group L.P. (Citg)
• Data cache having store queue bypass for out-of-order instruction execution and … - US US6360314B1 James B. Keller Compaq Information Technologies Group, L.P.
• Method and apparatus for determining availability of a queue to which a program … - US US7093105B2 James Keller Hewlett-Packard Development Company, L.P.
• Apparatus and method for replaying decoded instructions - US US5012403A James B. Keller Digital Equipment Corporation
• Method and apparatus for optimizing bcache tag performance by inferring bcache … - US US6401173B1 James B. Keller Compaq Information Technologies Group, L.P.
• Profile directed simulation used to target time-critical crossproducts during … - US US6212493B1 James B. Keller Compaq Computer Corporation
• Method and apparatus for maximizing utilization of an internal processor bus in … - US US5924120A James Keller Digital Equipment Corporation
• Method and apparatus for developing multiprocessore cache control protocols … - US US20010029574A1 James B. Keller Rahul Razdan
• Method and apparatus for resolving probes in multi-processor systems which do … - US US6295583B1 James B. Keller Compaq Information Technologies Group, L.P.
• Host bridge translating non-coherent packets from non-coherent link to coherent … - US US6714994B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for using a control signal on a packet based communication … - US US6748442B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for developing multiprocessor cache control protocols … - US US6349366B1 James B. Keller Compaq Information Technologies Group, L.P.
• Method and apparatus for minimizing pincount needed by external memory control … - US US6199153B1 James B. Keller Digital Equipment Corporation
• System and method for separate virtual channels for posted requests in a … - WO EP JP AU DE TW EP1314094B1 James B. Keller Advanced Micro Devices, Inc.
• Unload counter adjust logic for a receiver buffer - US US6434640B1 James B. Keller Advanced Micro Devices, Inc.
• Determination of execution resource allocation based on concurrently executable … - US US6704854B1 James B. Keller Advanced Micro Devices, Inc.
• Line predictor which caches alignment information - WO EP US JP KR US20040168043A1 James Keller Advanced Micro Devices, Inc.
• Method and apparatus for minimizing dcache index match aliasing using hashing … - US US6253285B1 James B. Keller Compaq Computer Corporation
• Method and apparatus for performing speculative memory fills into a … - US US6493802B1 James B. Keller Compaq Information Technologies Group, L.P.
• Conversation of distributed memory bandwidth in multiprocessor system with … - US US6393529B1 James B. Keller Advanced Micro Devices, Inc.
• Physical rename register for efficiently storing floating point, integer, … - US US6393546B1 James B. Keller Advanced Micro Devices, Inc.
• Computer system implementing flush operation - US US6553430B1 James B. Keller Advanced Micro Devices, Inc.
• Training line predictor for branch targets - US US6647490B2 James B. Keller Advanced Micro Devices, Inc.
• Method of controlling a shared memory bus in a multiprocessor system for … - US US5202973A James B. Keller Digital Equipment Corporation
• Computer system implementing system and method for ordering write operations … - US US6529999B1 James B. Keller Advanced Micro Devices, Inc.
• Apparatus and method for responding to an aborted signal exchange between … - WO EP US CN JP KR AU CA ES IL IN US4858173A James B. Keller Digital Equipment Corporation
• Line predictor entry with location pointers and control information for … - US US6546478B1 James B. Keller Advanced Micro Devices, Inc.
• Apparatus and method for providing a cache memory unit with a write operation … - WO EP US JP AU CA DE EP0292501B1 James B. Keller Digital Equipment Corporation
• Method and circuit for initializing a de-skewing buffer in a clock forwarded … - EP US AT DE US6952791B2 James B. Keller Broadcom Corporation
• Multiple processor integrated circuit having configurable packet-based … - US US20040019704A1 James Keller Barton Sano
• Conservation of system memory bandwidth and cache coherency maintenance using … - WO EP JP KR EP2320322A2 James Keller Advanced Micro Devices, Inc.
• Scheduler which discovers non-speculative nature of an instruction after … - US US6564315B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for balancing load vs. store access to a primary data … - US US6163821A James B. Keller Compaq Computer Corporation
• Predictor miss decoder updating line predictor storing instruction fetch … - US US6636959B1 James B. Keller Advanced Micro Devices, Inc.
• Method and apparatus for developing multiprocessor cache control protocols … - US US6314496B1 James B. Keller Compaq Computer Corporation
• Method and apparatus for a dedicated physically indexed copy of the data cache … - US US6253301B1 James B. Keller Compaq Computer Corporation
• Method and apparatus for accelerated addition of sliced addends - WO US JP US4878193A James B. Keller Digital Equipment Corporation
• Implementing locks in a distributed processing system - WO EP US JP AU DE US7640315B1 James B. Keller Advanced Micro Devices, Inc.
• Apparatus and method for providing a settling time cycle for a system bus in a … - US US5029076A James B. Keller Digital Equipment Corporation
• Method and apparatus for developing multiprocessor cache control protocols … - US US6651144B1 James B. Keller Hewlett-Packard Development Company, L.P.
• Transmitting data from a plurality of virtual channels via a multiple processor … - US US20040017813A1 James Keller Manu Gulati
• Scheduler which retries load/store hit situations - US US6622235B1 James B. Keller Advanced Micro Devices, Inc.
• High speed bus system - EP JP AU CA DE EP0464708B1 James B. Keller Digital Equipment Corporation

That's just what I got from filtering through the first 250 hits. There are about 4 or 5 of those I listed which are just "James Keller", but most correspond to places he was known to work and the dates matched. Note that Compaq bought DEC, which explains why some of his Alpha patents are filed under that company name.


It was for a mass audience. He's not lecturing to graduate students or mentoring interns. Plus, the interview was already pretty long, as is. The deeper he got into the technical details would've expanded the scope exponentially.

Also, to fully appreciate what he said, you'd want to look at the surrounding context. I didn't included it in my quote, because it wasn't material to the point I was trying to make. However, if you're going to pick on the informational content of the quote, then you'll want to include the context.


I told you I didn't consider your reply to be serious and fact-based.

I agree with your assessment of Jim Keller, that in addition to his impressive patent history, he has a reputation for innovation (assumably bringing a talented team of engineers) at semiconductor firms. Is it just a coincidence that Keller led engineering R&D at DEC, SiByte (acquired by Broadcom), AMD (Athlon), PA Semi (acquired by Apple--A4+A5), AMD (again--Zen), Tesla (Autopilot HW), Intel, and now Tenstorrent, when they were pivoting with major architectural changes?

I saw the earlier comment that SiFive might get acquired. They've been struggling for a while--RISC-V customization is a tough market to compete in--and maybe SiFive's acquisition price is more reasonable for an acquisition by the likes of AMD. However, AMD's recent acquisitions have mostly been smaller AI HW+SW startups (less expensive to acquire), after their hugely expensive acquisition of Xilinx ($49 Billion), which reminded everyone that ROI is something to weigh more heavily, echoing AMD's backbreaking acquisition of ATI. I wonder if Microchip might acquire SiFive--I'd have to look into whether SiFive's business model/clients would fit into Microchip's strategic planning.

 

[bit_user]

Is it just a coincidence that Keller led engineering R&D at DEC, SiByte (acquired by Broadcom), AMD (Athlon), PA Semi (acquired by Apple--A4+A5), AMD (again--Zen), Tesla (Autopilot HW), Intel, and now Tenstorrent, when they were pivoting with major architectural changes?

Absolutely. This is a result-based industry. If you're a clown, you will be quickly found out and run out on a rail. I think the main reasons for Keller's notoriety are that he earns the respect of the people who work with him and he delivers results.

And let me add: based on what little I know about them, I'd sure as heck rather work for Jim Keller than Steve Jobs. Keller doesn't seem like a tyrant, the way that many anecdotes have painted Jobs. I think/hope it shows that you can have a high-performing team without destroying the lives of its members, but I'll confess that I'm well into the realm of speculation, at this point.

 

[bit_user]

And besides, I don't hate that he is given notoriety -- I am protesting the way it is done.

He is basically being mystified as some great tech guru, articles published don't go in any depth about his work, they are shallow and vague, full of bombastic statements wiithout substance -- they don't critically analyze any of his claims, just parrot them and expect us to take them for granted.

Sites like Toms are aimed at the largest audience possible, within their niche. I don't see them go into that kind of detail with anyone. If you want to see more substantive interviews, then perhaps you're looking in the wrong places.

To me those hype articles with their low effort way in which they are presenting him and Jensen and the others are actually putting an equal sign between sport / media celebrities and people who do actual science and engineering. Frankly, I am surprised that such equalization doesn't bother you.

First, I have also frequently taken issue with Jensen being deified and his words treated as particularly insightful.

Second, this article isn't about Jim Keller. It's about a startup offering a new product - namely, their prebuilt AI PCs - which was announced by their CEO. That CEO happens to be Jim Keller, and he's definitely trading some of his name cred to help his company. It happens all the time. Still, the article isn't about him. It's about their products. So, your remarks about him are somewhat missing that point.

It wouldn't. It need not be a full biography, he could use ChatGPT to summarize important bits of his work.

That Anandtech article I linked was written 3 years ago. ChatGPT wasn't a thing, back then. LLMs existed, but not of a caliber that could be used to generate such content.

Now, I would love it, if there were a LLM that could read all of his patents, other publications, and every other tidbit of detail published about the products he worked on, and then be prompted to talk about salient aspects of them, highlighting his contributions. Or, someone could just ask him and get it from the horse's mouth. He doesn't seem like a terribly self-aggrandizing person, but maybe you can actually read (or watch) some of the interviews with him, and decide for yourself.

That's what the proper interviews and real journalists are for -- to tell us why he is important by mentioning most important accomplishments, not by saying "he is important because he is a CPU god".

Except, isn't it kind of rude to do an investigative piece - especially one involving interviews - on whether person XYZ is really such a big deal? Who would even sit down, for such an interview?

What you might find is people doing retrospectives on some of the projects he happened to be involved in, like DEC Alpha or AMD K8. You might learn some things about him, from that.

Because you asked me to give those names which I did and it's a basic curtesy to respond if I already bothered to fulfill your request.

At the same time as you did that, you also posted FUD that caused me to lose faith that we were engaging in an intellectually honest discussion. If I can't trust what you're saying then I'm not going to invest time in it. It's really that simple.

You are avoiding to respond to that because it won't help your argument that he deserves to be hyped

Nope. I don't care what you think about him. Believe whatever you like.

 

[CmdrShepard]

Now, I would love it, if there were a LLM that could read all of his patents, other publications, and every other tidbit of detail published about the products he worked on, and then be prompted to talk about salient aspects of them, highlighting his contributions.

You can do that with llama3 and RAG model such as nomic-embed-text (extract and index embeddings from patent documents so they are searchable by LLM and can be referenced).

Or, someone could just ask him and get it from the horse's mouth.

The trouble is, given the audience, nobody is asking the right questions.

Except, isn't it kind of rude to do an investigative piece - especially one involving interviews - on whether person XYZ is really such a big deal? Who would even sit down, for such an interview?

That's not what I meant. I meant that he should be interviewed by someone with engineering background who can understand what he was doing and ask the right questions to get the most important bits of what he did out.

At the same time as you did that, you also posted FUD that caused me to lose faith that we were engaging in an intellectually honest discussion.

I don't post FUD, and accusing me of doing so isn't exactly honest either.

If I can't trust what you're saying then I'm not going to invest time in it. It's really that simple.

I never asked you to trust me, I only wanted TH to verify stuff they publish and give at least some details to bolster credibility of those claims instead of us having to blindly trust their X regurgitations of tech celebrities. Too bad you can't see that.

 

[bit_user]

I only wanted TH to verify stuff they publish and give at least some details to bolster credibility of those claims instead of us having to blindly trust their X regurgitations of tech celebrities. Too bad you can't see that.

I do understand that you have extremely high standards (unrealistic, IMO), for any sort of praise or notoriety.

I happen to think his resume, alone, clears this bar. As far as that goes, here's a full list of the positions he's held, over the course of his career. No, it doesn't give any details about what he did there, but at least it's more authoritative than a Wikipedia page.

https://www.linkedin.com/in/jimbkeller/details/experience/

That's not to mention the 95 patents I cited (and no, don't try to tell me you actually read even the abstracts of more than a handful). You can't possibly know that none represented real breakthroughs.

We're at an impasse, so we'll have to agree to disagree.