A Glimpse Under the Hood of Larrabee

rgMekanic

[H]ard|News
Joined
May 13, 2013
Messages
6,943
Back in the early 2000's, Intel was working on a revolutionary CPU and GPU hybrid called Larrabee. Think of a GPU that can use the x86 instruction set. Unfortunately in 2010, it was announced that Larrabee would never make it to the public. Fast forward to November of last year, when [H]ardForum member, and GPU collector erek spotted a Larrabee prototype on eBay. Unfortunately erek was unable to win the auction (read the thread for reasons), but somehow the card from the auction ended up in the hands of Linus Tech Tips who did a somewhat underwhelming video on the card, and tried to get it running unsuccessfully.

Fortunately for us, another [H]ardForum user, Slaventus86, also has a Larrabee card, and he isn't scared to show under it's skirt.

Check out the video.

Awesome stuff, I really hope he manages to get it working so we can see a glimpse of what it can do.

My try to launch Intel's NEVER RELEASED prototype - Intel Larrabee card. This card is also known as Knights Ferry or Aubrey Isle. It has 32 cores, 2 Gigs of RAM, DVI, HDMI, DP outputs
 
lrb_img166_small.jpg
 
Why can't people in other countries pronounce the letter R. I cannot understand them when they speak! People from the middle east have the "rolling R" syndrome really bad. China replaces the R with L. I know English is a confusing and difficult language, but what makes the R so difficult?

On another note, this is freaking cool. I thank the guy for showing us a piece of hidden technology history.
 
  • Like
Reactions: erek
like this
I'd hit up Tom Forsyth

He was the first guy I contacted as soon as I got the card. He helped me a lot but unfortunatelly he doesn't have any drivers/software.

Why can't people in other countries pronounce the letter R. I cannot understand them when they speak! People from the middle east have the "rolling R" syndrome really bad. China replaces the R with L. I know English is a confusing and difficult language, but what makes the R so difficult?

This is how teachers teach us at school :) Sorry, my bad, need to listen more to natives I think

UPD: Just found out that Tom Forsyth has returned to Intel. Ok now, I'll make one more try to contact him
 
Last edited:
Did intel ever give a reason for its cancellation or did it just get dropped and not mentioned again? I'd have to guess performance wasn't upto snuff for them to shitcan it.
 
  • Like
Reactions: erek
like this
Did intel ever give a reason for its cancellation or did it just get dropped and not mentioned again? I'd have to guess performance wasn't upto snuff for them to shitcan it.

Story on the street is it became their Phi cards. Which apparently sell really well, so while it didn't enter the discrete graphics market, it was still profitable.
 
  • Like
Reactions: erek
like this
Slaventus86

I'd like to see the information from the "event log" that it mentioned in your video ... can you locate it?

upload_2018-7-4_16-4-27.png



also i'd be interested in messing around with the Linux image it's trying to boot from Knights Corner: bzImage-knightscorner.bin

upload_2018-7-4_16-3-32.png
 
Let me look at this one: D:\Program Files\Intel\MPSS\k1om-mpss-linux\boot\vmlinux-2.6.38.8+mpss3.3.4 ...


vmlinux-2.6.38.8+mpss3.3.4: ELF 64-bit LSB executable, Intel K1OM, version 1 (SYSV), statically linked, BuildID[sha1]=57d7d491dcf0c61141904a1cf1dfa858218cd401, with debug_info, not stripped, too many notes (256)


vmlinux-2.6.38.8+mpss3.3.4: file format elf64-k1om
architecture: k1om

versus:

/boot/vmlinuz-4.15.0-23-generic: Linux kernel x86 boot executable bzImage, version 4.15.0-23-generic (buildd@lgw01-amd64-055) #25-Ubuntu SMP Wed May 23 18:02:16 UTC 2018, RO-rootFS, swap_dev 0x7, Normal VGA


Obviously wont work:

upload_2018-7-4_17-2-0.png
 
Slaventus86

X:\Program Files\Intel\MPSS\global.xml

"
<!--Remove the quiet element for more verbose boot logging-->


<quiet/>"


-- Can you remove that and see?
 
Why can't people in other countries pronounce the letter R. I cannot understand them when they speak! People from the middle east have the "rolling R" syndrome really bad. China replaces the R with L. I know English is a confusing and difficult language, but what makes the R so difficult?

On another note, this is freaking cool. I thank the guy for showing us a piece of hidden technology history.

I was introduced to the English language about 25 years ago and I still can't wrap my head around not having an actual nice and clean R like they used to long ago. I actively avoid pronouncing it pLopeLly - that's how you talk
 
I think Slaventus will regret his willingness.

Nukester: "Syndrome"? Please. How many languages do you speak? And how many of those without an accent? Yes, it's part of an accent: "R" is not formed in the same way in all languages (mostly it appears as a rolling R: English is actually the exception). Call it a transfer error or just an accent, so long as it doesn't interfere with intelligibility, there's no problem. That you can't understand might reflect less a problem on the speaker's end and more your lack of exposure to people from other countries. Travel a bit. Or watch some YT vids at least.
 
I think Slaventus will regret his willingness.

I'm just hoping to figure out why it won't even boot the Linux kernel image for the Xeon Phi coprocessors. I already took a look at the init ramfs image, there's no sign of the "DirectXGfx":


erek@ubuntu:~/Desktop/1$ ls
bin boot dev etc home init initramfs-knightscorner.cpio.gz lib lib64 media mnt proc sbin sys tmp usr var
erek@ubuntu:~/Desktop/1$ grep -ri directx *
erek@ubuntu:~/Desktop/1$ find . -type f -name "*irect*"
erek@ubuntu:~/Desktop/1$
 
essentially micctrl.exe is intel's replacement for grub and can be configured via that global.xml or per mic#.xml (each iteration of a xeon phi coprocessor you have installed)... would like to see the boot up with verbose logging enabled ... everything is compiled to intel's specific hardware architecture: k1om
 
Did intel ever give a reason for its cancellation or did it just get dropped and not mentioned again? I'd have to guess performance wasn't upto snuff for them to shitcan it.
They changed direction with the product, markets were opening up for highly threaded, very task specific processors. The cores on these allowed them to pack a ton on a simple add-in card and sell.

From my understanding it did well for them. I don’t understand why they still didn’t try a GPU release though.

Also to add they needed to keep NV from dominating this area (NV ended up dominating) to ensure they maintained certain contracts.
 
  • Like
Reactions: erek
like this
Might be a stupid question... why do you expect a KNC kernel to boot on KNF?
Tom said they're all the same chip:

Tom Forsyth said:
When I say "Larrabee" I mean all of Knights, all of MIC, all of Xeon Phi, all of the "Isle" cards - they're all exactly the same chip and the same people and the same software effort

https://tomforsyth1000.github.io/blog.wiki.html

if there's something like what Slav was doing in terms of forcing the drivers to install by editing the Device ID related to how this public software stack works then maybe we can get the Linux environment at least initialized on it
 
Why can't people in other countries pronounce the letter R. I cannot understand them when they speak! People from the middle east have the "rolling R" syndrome really bad. China replaces the R with L. I know English is a confusing and difficult language, but what makes the R so difficult?

On another note, this is freaking cool. I thank the guy for showing us a piece of hidden technology history.

They are called phonemes look up about them they are interesting one of the most interesting ones is the TH in the, actually that is really rare from what I have experienced in most other languages. But just the same there is a ton you can't pronounce. Most people in America don't get it cause you know Americans can only speak one language. The English R is often different in other languages. But just the same when you try to pronounce the closest phoneme to an R such as the ror rua in thai you will botch it up and they will laugh at you. Right now people generally feel that once you pass the age of 12 you lose the ability to learn new phonemes this is why people who speak a language at a young age speak with no accent but no matter how many years you work on a language you can never shed an accent if you did not learn it before 12.

What would be really cool is if some government, (hint the USA government) would make an initiative to move all or almost all the phonemes into the English language and teach them in primary school if they did this you could actually create a society which was capable of learning almost any language with no or very little accent without needing them to learn that language early.
 
Last edited:
Here is boot log:

Event[25]:
Log Name: MPSSLog
Source: micctrl
Date: 2018-07-05T11:31:08.302
Event ID: 20
Task: N/A
Level: Error
Opcode: Info
Keyword: Classic
User: N/A
User Name: N/A
Computer: DESKTOP-B4ET9RR
Description:
Timeout node 0 From state MIC_BOOT to state MIC_ONLINE. Current state: MIC_BOOT

Event[26]:
Log Name: MPSSLog
Source: micctrl
Date: 2018-07-05T11:31:08.302
Event ID: 50
Task: N/A
Level: Error
Opcode: Info
Keyword: Classic
User: N/A
User Name: N/A
Computer: DESKTOP-B4ET9RR
Description:
Exception during: System.Exception: Timeout node 0 From state MIC_BOOT to state MIC_ONLINE. Current state: MIC_BOOT
at MpssLib.Host.WaitForTransition(UInt32 node, mic_status FromState, mic_status ToState)
at MpssLib.Host.WaitForAllTransitions(List`1 list, mic_status FromState, mic_status ToState)
at MpssLib.Host.Start(List`1 BootList)

I don't see any difference with verbose logging enabling. I also tried to send "micctrl -l mic0" during boot, but just got full screen of symbols without sense.

UPD: difference is that log output goes directly to console. So above is the full log.
 
Last edited:
Even if the card can never fully be initialized, this is such a happier ending to this than I could have thought possible.
Keep up the great work Slaventus86, and thanks for erek for sharing this for us all from the very beginning!
 
Cool stuff and awesome efforts by all. Can only hope one day a binder or disk surfaces somewhere with the needed info for this. Fascinating read here in the thread about this unique critter.
 
even without drivers for windows, it should still boot up and work in VGA mode during BIOS POST. that was odd
 
Why can't people in other countries pronounce the letter R. I cannot understand them when they speak! People from the middle east have the "rolling R" syndrome really bad. China replaces the R with L. I know English is a confusing and difficult language, but what makes the R so difficult?

It's probably not a priority to them and they didn't realize it would trigger you so much. English is more of a tool for communication to them than something they're happy to master, etc.
 
They are called phonemes look up about them they are interesting one of the most interesting ones is the TH in the, actually that is really rare from what I have experienced in most other languages. But just the same there is a ton you can't pronounce. Most people in America don't get it cause you know Americans can only speak one language. The English R is often different in other languages. But just the same when you try to pronounce the closest phoneme to an R such as the ror rua in thai you will botch it up and they will laugh at you. Right now people generally feel that once you pass the age of 12 you lose the ability to learn new phonemes this is why people who speak a language at a young age speak with no accent but no matter how many years you work on a language you can never shed an accent if you did not learn it before 12.

What would be really cool is if some government, (hint the USA government) would make an initiative to move all or almost all the phonemes into the English language and teach them in primary school if they did this you could actually create a society which was capable of learning almost any language with no or very little accent without needing them to learn that language early.

Thanks for that post. Interesting about how difficult it is to learn phonemes after a certain age.
 
  • Like
Reactions: erek
like this
They are called phonemes look up about them they are interesting one of the most interesting ones is the TH in the, actually that is really rare from what I have experienced in most other languages. But just the same there is a ton you can't pronounce. Most people in America don't get it cause you know Americans can only speak one language. The English R is often different in other languages. But just the same when you try to pronounce the closest phoneme to an R such as the ror rua in thai you will botch it up and they will laugh at you. Right now people generally feel that once you pass the age of 12 you lose the ability to learn new phonemes this is why people who speak a language at a young age speak with no accent but no matter how many years you work on a language you can never shed an accent if you did not learn it before 12.

What would be really cool is if some government, (hint the USA government) would make an initiative to move all or almost all the phonemes into the English language and teach them in primary school if they did this you could actually create a society which was capable of learning almost any language with no or very little accent without needing them to learn that language early.

With all due respect, and knowing full well that an internet forum doesn't necessarily constitute any kind of research-level writing, most of your post is nonsense (and ironically, very poorly written for a person espousing linguistic knowledge). No need to get into why in great detail here, but suffice it to say that a 'phoneme' simply refers to the individual sounds produced by letters in a given alphabet. Every language that is spoken thus contains phonemes and to say that after 12 people lose the ability to learn new phonemes is untrue. The 'critical period' to achieve native-like fluency in a given language is passed well before 12, yet people can acquire languages to fluent levels, albeit via different cognitive means than how it is achieved within the critical period. This ties in directly to accent; learning a language after the aforementioned critical period, we mimic the sounds instead of naturally producing (not the best way to characterize this process, but I'll go with it here) them which is why again, learning a language later in life might lead to some phonetic inaccuracy (lack of sufficient exposure to received or 'standard' pronunciation and correction). This in turn leads to fossilization, or a kind of permanence in pronunciation (think of the adage "practice makes perfect" and correct it to "practice makes permanent"): when a mature speaker reaches the point of intelligibility, they tend to stop trying to improve, and quite rightly so, as this is the only standard of world English pronunciation. Thus, the rolling R is not incorrect at all, nor is it any kind of "syndrome": it's just the English pronunciation of many Slavic non-native speakers of English and quite acceptable. English is taught in many, many countries at the elementary school level, but without more immersion, exposure and correction around the clock, it is tremendously difficult to overcome the differences which you see as problematic and I see simply as accents. Nobody owns a language and it is all of our jobs to become accustomed to the variations. Just as an American, an Australian and a speaker from the UK all have different English pronunciations, so too do the Japanese, Russians and ze Germans. :)

Sorry to go off-topic, but a little information is certainly a dangerous thing.
 
With all due respect, and knowing full well that an internet forum doesn't necessarily constitute any kind of research-level writing, most of your post is nonsense (and ironically, very poorly written for a person espousing linguistic knowledge). No need to get into why in great detail here, but suffice it to say that a 'phoneme' simply refers to the individual sounds produced by letters in a given alphabet. Every language that is spoken thus contains phonemes and to say that after 12 people lose the ability to learn new phonemes is untrue. The 'critical period' to achieve native-like fluency in a given language is passed well before 12, yet people can acquire languages to fluent levels, albeit via different cognitive means than how it is achieved within the critical period. This ties in directly to accent; learning a language after the aforementioned critical period, we mimic the sounds instead of naturally producing (not the best way to characterize this process, but I'll go with it here) them which is why again, learning a language later in life might lead to some phonetic inaccuracy (lack of sufficient exposure to received or 'standard' pronunciation and correction). This in turn leads to fossilization, or a kind of permanence in pronunciation (think of the adage "practice makes perfect" and correct it to "practice makes permanent"): when a mature speaker reaches the point of intelligibility, they tend to stop trying to improve, and quite rightly so, as this is the only standard of world English pronunciation. Thus, the rolling R is not incorrect at all, nor is it any kind of "syndrome": it's just the English pronunciation of many Slavic non-native speakers of English and quite acceptable. English is taught in many, many countries at the elementary school level, but without more immersion, exposure and correction around the clock, it is tremendously difficult to overcome the differences which you see as problematic and I see simply as accents. Nobody owns a language and it is all of our jobs to become accustomed to the variations. Just as an American, an Australian and a speaker from the UK all have different English pronunciations, so too do the Japanese, Russians and ze Germans. :)

Sorry to go off-topic, but a little information is certainly a dangerous thing.

I don't think I every espoused linguistic knowledge I don't even have any training I just looked up some stuff while studying another language and wondering what the answer was to a basic question why is it so impossible for any older person to speak a new language correctly. However my basic knowledge was not improved in your paragraph. If the age is not by 12 for which someone must learn to speak as I say without an accent but you say native like, then why don't you just tell us what that age is? If it is not phonemes that they are unable to replicate then also what is it?

I don't know what people owning or not owning a particular language has to do with it. Does that dismiss the point that there are sounds that you cannot replicate after a certain age?
 
I don't think I every espoused linguistic knowledge I don't even have any training I just looked up some stuff while studying another language and wondering what the answer was to a basic question why is it so impossible for any older person to speak a new language correctly. However my basic knowledge was not improved in your paragraph. If the age is not by 12 for which someone must learn to speak as I say without an accent but you say native like, then why don't you just tell us what that age is? If it is not phonemes that they are unable to replicate then also what is it?

I don't know what people owning or not owning a particular language has to do with it. Does that dismiss the point that there are sounds that you cannot replicate after a certain age?

I will stick to your questions as they appear.

The critical period is between 0 and 7, but some say it ends at 5, or when a child begins to use particular linguistic structures (as an indicator). If I were to write out for you all of the factors that might account for all of the difficulties adult learners have, it would take entire volumes. I didn't set out to do this and I think it isn't a reasonable request. If you're interested, there is a huge mass of data on the internet or in your local library. Here is a light introduction to the topic. You can do further searches for the names mentioned in the article: it's kind of a top-ten of the important authors and researchers in the fields mentioned. Bibliographies are your friend.

The fundamental reason older people have trouble learning new languages is the same reason why they have trouble with almost anything that is entirely new and again, I haven't the time (or inclination) to detail this further. It is searchable and there are reams of information about learning - not just language - as an adult. Finally, age in and of itself has nothing to do with making sounds or the ability to do so and pronunciation is only one aspect of learning a language. I would claim that generally, people fail to learn languages simply because they aren't devoted to the task. It takes daily work - I mean literally a couple of hours a day minimum - to see significant growth and even then over months. So, motivation is key and then of course is quality instruction and opportunity to use the target language in a meaningful way.

I don't and didn't mean to come off sounding like a prick, but your post needed correction, plain and simple. I also have learned languages (non-Romantic) as an adult and am well aware of how hard it is. But it's tough for everyone who missed the critical period. If you really want to do it, you can, but honestly, most people don't set realistic, achievable goals, don't want it bad enough, lack quality instruction, lack intrinsic motivation and have nowhere near enough opportunity to approach even moderate fluency let alone bilingualism. I only mentioned "ownership" of language because it is a frame: most people believe that only Germans speak German, for example, when the truth is that anyone who learns German speaks German. I know that sounds something like a tautology, but again, as a healthy, optimistic frame of mind, it makes sense to think that way.

You're starting point is self-defeating: by "wondering what the answer was to a basic question why is it so impossible for any older person to speak a new language correctly", you have set yourself up to fail. I have outlined the major hurdles people face, but all of those can be overcome.
 
Here is boot log:

Event[25]:
Log Name: MPSSLog
Source: micctrl
Date: 2018-07-05T11:31:08.302
Event ID: 20
Task: N/A
Level: Error
Opcode: Info
Keyword: Classic
User: N/A
User Name: N/A
Computer: DESKTOP-B4ET9RR
Description:
Timeout node 0 From state MIC_BOOT to state MIC_ONLINE. Current state: MIC_BOOT

Event[26]:
Log Name: MPSSLog
Source: micctrl
Date: 2018-07-05T11:31:08.302
Event ID: 50
Task: N/A
Level: Error
Opcode: Info
Keyword: Classic
User: N/A
User Name: N/A
Computer: DESKTOP-B4ET9RR
Description:
Exception during: System.Exception: Timeout node 0 From state MIC_BOOT to state MIC_ONLINE. Current state: MIC_BOOT
at MpssLib.Host.WaitForTransition(UInt32 node, mic_status FromState, mic_status ToState)
at MpssLib.Host.WaitForAllTransitions(List`1 list, mic_status FromState, mic_status ToState)
at MpssLib.Host.Start(List`1 BootList)

I don't see any difference with verbose logging enabling. I also tried to send "micctrl -l mic0" during boot, but just got full screen of symbols without sense.

UPD: difference is that log output goes directly to console. So above is the full log.

Slaventus86 would you be interested in selling your Larrabee for $700-800 USD?
 
Back
Top