...not with full graphics. The average cluster is only 8 nodes, 160 cores or so, with 6-8 GPUs...they can't be expected to run something like Crysis full out, right?
156 posts • joined 20 Nov 2009
...not with full graphics. The average cluster is only 8 nodes, 160 cores or so, with 6-8 GPUs...they can't be expected to run something like Crysis full out, right?
Yes, you are correct, this is not a UK shop. But what's the difference?
What I wanted to spotlight in the article was a company that serves the high-end geek, and provides the choices and service that you just don't see much anymore.
Snarky today, eh? That's ok, I can see how the article might read as an advertioral. But my intention was to highlight a company that is doing things right and willing to work with an increasinly small niche of the market - the power user.
Like many of you, I get requests from friends, family, and collegues to help them figure out what sort of PC or workstation they should buy. I'm geeky enough to enjoy this process, unfortunately. Over the last few years, I've been disappointed to find that fewer and fewer vendor configurator tools give prospective buyers a robust set of options. Sure, you can change the processor, amount of memory, or GPU, but fewer of them really give you the ability to configure everything - and fewer yet will take the time to either write up, or talk to you directly, about the best set of options. You used to be able to find these shops in years past, but not so much today - and I miss it.
Hey, I'm running my business on this system, so it was pretty damned important that I get a replacement. I did a bit of research into fixing the power connector however:
1) it wasn't totally clear that the problem WAS the power connector, as someone commented above, it could have been broken traces somewhere else in the board, but near the power connector
2) I'm not the most handy guy in the world. If I were to start soldering, what if I dripped solder onto other parts of the board? What if I slipped with the iron and melted something else? Scary stuff for a non-hardcore guy like me.
So tell me more about this Range Rover, my current car is dirty and the dealer told me the best course is to replace it....
Hey, no need for "Mr Olds", we're all pals here, call me Dan, or Mr. Dan if you insist on some level of formality.
Good comments above, particularly your points about solutions vs. selling parts of a solution. I've been in the business for 20 years now and one thing I know is that selling an entire solution is much more profitable than just selling the parts. The very highest deals, in fact, are those that begin with a business consulting engagement. If a company can successfully diagnose a problem or discover an opportunity on the business side, then their technical solution to that problem/opportunity, despite a high price tag, is almost unassailable by less well-informed competitors.
Bottom line: Solutions = $$
You bring up a point I should have brought out. When IBM had the chip franchise for all three game consoles, it drove a lot of volume for them. Enough volume to justify continuing and even expanding their fab operations. When the consoles went to AMD, the math changed considerably.
"Not competent journalism"? Ouch...but then again, I'm not a journalist. I'm an analyst with my own boutique IT industry market research/analysis shop, Gabriel Consulting Group.
I usually get early NDA briefings and deep dive briefings on upcoming hardware/software/initiatives from a wide variety of vendors. I have to walk a fine line when it comes to what I say in some of my public articles, trying to get out the stuff I know while not violating any NDAs.
I don't want to be one of those guys who constantly says "I can't tell you more because of my NDAs..." Maybe I'm overly sensitive, but I think that line would make me sound like a pretentious and pompous asshat. .
I would also point out that being 'in the know' doesn't necessarily mean anything. Take the Sun Rock processor example that a lot of people cite in these comments. Sun repeatedly told the analyst community, myself included, that Rock was healthy and coming along. Maybe a little late, but it's going to be awesome when it's released. We all know what happened there, right?
Wow, your scathing denunciation is actually longer than my original article, I think. I don't have time to address all of your points, but I will hit on a few here.
1. "AIX is dead" I don't see or hear this at all from IBM, anyone working at IBM, or customers - either formally or informally. There is an AIX roadmap and development on the o/s continues. Linux isn't replacing AIX, it's a complimentary o/s that allows the platform to compete more effectively with x86. The article you link in your reply is from 2003....11 years ago....and has obviously not come to pass.
2. There is also a Power processor roadmap that includes Power9 and other processors. From what I've heard from very reliable sources, there are Power9 processors running in labs today.
3. x86 vs. Power: Intel revs their processors more quickly than IBM revs Power processors. The typical pattern is that a new Power processor will offer significantly higher performance than the cream of the crop x86 chip. Then, over the next couple of years, Intel processors play catch up. IBM will do a frequency 'speed bump' sometime in that period. Towards the end of a Power generation, Intel processors are arguably at parity on most feeds/speeds, and maybe even ahead on others. But then IBM trots out a new Power generation and the game begins again. So comparing today's best Intel processors vs. older Power7 designs isn't exactly informative or fair.
4. "Why does this article exist....?" I was originally going to write about the deal between IBM and GF, but the original article I cited did a great job of summing up the deal. But I noticed in the comments that several folks seemed to believe that this transaction was a death knell for IBM's hardware business. As a guy who knows more than a little about this topic, I decided to write about it.
5. You use the words 'desperate' and 'desperately' a few times in your masterful summation. However, I have to point out that I'm not desperate to defend much of anything, other than my beloved Oregon Ducks against the hated SEC in college football.
Thank you! I knew someone would do it, thanks for being the first.
The online info on the GADGET site isn't written to be what I'd regard as understandable for the typical techie. Way too much jargon, not much clarity, not great....
In retrospect, I should probably set up an anonymous account if I'm going to shamelessly tout my own articles.
Hey John, yeah, I had some problems finding it at first too. Here's a link to the developers: http://www.mpa-garching.mpg.de/gadget/ and here's a link to the Wiki article on GADGET: http://en.wikipedia.org/wiki/GADGET
I should probably hyperlink all of the software references, that's an oversight on my part. I'll do that in the future.
And, for my money? Inspector Gadget should have been much better. Great concept, so-so execution.
Hey, this IS the most comprehensive coverage of student cluster competition I've ever seen. No need to eat a handful of spider webs, at least as far as I'm concerned. The writing literally leaps off the page, goes through my eye holes, and gets into my head - in a good way.
You are obviously very wise, and not just in the pastry and dessert arena.
Hey, I live to use those cliches...in fact, I'm looking for more of them in order to keep from using the same cliches over and over.
As you can tell, I really like these competitions. I'd like folks in technology to see these events much like the general public does college football, basketball, soccer, or whatever.
My strategy for accomplishing this overly ambitious goal is to treat these competitions as much like a major collegiate sport as I can. I want to cover the technical details (which are the "x's and o's" in other sports), and also generate some excitement and add some humor along the way.
If anyone has any good cliches for me to add into my mix, let me know by hitting my name and sending me an email. If they're good, I'll use 'em and even credit you for the contribution.
My thinking was that a brace of accelerators would be the typical maximum that you'd ever see on a typical system - ie four. Thus a double brace would be eight. But your research has opened my eyes and I'll bow to that definition in future articles.
And stop calling me Shirley
I'm not a creative professional. I run a small industry analyst firm and put up a fair amount of stuff here in El Register. You may have seen some of my videos covering the student cluster competitions and a few webcasts in my Reg HPC Blog.
When I stated doing more audio and video work in 2010, I realized that I needed some serious tools if I wanted my stuff to look and sound better. Looking at what I was doing, I figured I needed a good video editing suite, photo editing, audio processing, and even something I could use to cobble together the odd logo or two. I tried GIMP, Audacity, and some other free tools, but wasn't wild about them. I had a hard time using them, which is mainly a function of me not having enough time to devote to getting up the learning curve.
I then took a hard look at various Adobe offerings, but getting full versions of Premiere, Photoshop, Audition (which was called something else back then), plus Illustrator or maybe AfterEffects, was going to cost me anywhere from $2,000 -$3,000, plus a few hundred bucks every couple of years for updates.
That was a big enough number that I just couldn't justify it. Then CC came along and gave me all of this stuff for $50 a month. Now I have full versions that are always up to date, and I can use them on my desktop, laptop, and on big server - as long as I only run one instance at a time.
Speaking for myself, I've found CC to be one of the best purchases I've made in a long time. I haven't had any problems with the apps or with CC verification - even when I'm on extended business trips. I did have a problem once when my credit card expired, but CC gave me 30 days of use before pulling the plug. I called Adobe customer support, and it was taken care of in about 5 minutes.
Financially, I think it's about a wash in my case. I'm a guy who typically purchases updates when they're available, and I'm really glad to see that there's so much educational material out there. If I'm stuck trying to do something, I seem to always be able to find a Youtube video or tutorial that shows me the way.
I'm using these apps more than I thought I would, primarily Audition and Premiere, with a fair amount of AfterEffects and Photoshop, and a smidgen of Illustrator, Prelude, and Speedgrade.
As much as I like bitching about things in general, CC has been great for me. I think they did it right, and I think this is would be a great delivery model for other ISVs to emulate.
Good point on the price difference, but NVIDIA is a big supporter of these competitions and had been great about giving the kids whatever they need.
Thanks for the info on GeForce and CUDA. I've been working off and on with a PCIe extension box that has GeForce cards in it, trying to get it working with my laptop. So far, no success in getting the laptop to see the box or the cards, but that could be a limitation of Windows 7 or the lousy PCIMIA card I'm using to connect. I'll try connecting it to a desktop motherboard directly via PCIe.
LOL....I've been writing about this school and their cluster competitions for years and you're right!! I didn't see it before. Nice job...
Yeah, should have been 1,000x, not 1,000%....I screwed the pooch on that one....total pooch screw.
I agree, that would be pretty cool. Maybe for GTC15? Hmm...I wonder if I could get the GE guys to give me some footage of the the torture tests they put their ruggedized systems through? That might be a substitute for what you wanted to see....
That's a good point about potentially higher power requirements for general purpose gear vs. special function chips. If anyone is interested, I could talk to Dustin and see how GPUs compare to FPGAs and other custom chips in terms of performance, power, heat, etc. Might make a good webcast.
I think that the environmentals can probably be engineered around, particularly since the general purpose stuff is so much less expensive, it would free up budget for the re-engineering. But one problem I haven't considered is security. Does using general purpose gear make the device more vulnerable to hacking by an enemy? I would imagine it would, so what steps are taken to keep it secure?
The 'big deal' part of this is that we have computers that are fast enough to output the various waves, modulate them correctly, etc., then received the input, sort it, interpret it, and route it to the right place. All of this on the fly and at the same time - with a computer that's small enough to fit in a small compartment on a plane. That's pretty cool, isn't it?
Yeah, I know what you mean. I've wondered about that stuff too, whether any system could withstand the kind of punishment it would receive on a real battlefield. But, damn, they ruggedize the hell out of their gear. You can see a bit of it in this video I shot at GTC13 http://www.theregister.co.uk/2013/04/10/ge_ruggedised_systems/
They also pay a lot of attention to thermals. I don't know if you can see this in the video, but the metal casings around their system is very thick, they do this to make it stronger of course, but the enclosure also acts as a huge heat sink. It was just over room temperature when I touched it, but they assured me that it was engineered to handle much more heat than could be generated at a well attended trade show.
Ok, that cracked me up, made me spit Sprite Zero on my keyboard. Are you happy now?
Damn it! I missed out on writing about the free unicorn? They didn't have a unicorn at the show, or maybe it was in one of the customer-only whisper suites.
The way I understood this is that we're essentially talking about using the entire set of drives in the array (48? 64? or more) to rebuild the failed drive. This would be a lot faster, of course, then having five or ten drives rebuilding a single drive.
As to the internal bandwidth and the limitations there, you make a good point. However, I'm pretty sure we're talking about SATA 3 rather than SATA 2, and PCIe 3 as well. I'm not positive on these points, but it makes sense to me that they'd use the latest/greatest on this new storage box.
You guys both make some good points, but these Student Cluster Competitions aren't just about putting together computers and racing them. These kids have to run real scienfitic apps, running the gamut from weather forecasting to molecular dynamics to financial modeling. In order to do this, they have to investigate and understand the apps and the science behind them. Many of these students are majoring in scientific fields, which, as you point out, require some fairly deep knowledge of computer science.
The laptop was not connected to the same speakers as the desktop. The audio on the laptop was playing through the laptop speakers and headphone jack. Hmm....I didn't check to see if both computers were playing the SAME audio at the same time - that was an oversight on my part, probably caused by panic..lol
Yeah, you're right. I may not have a solution here at all. But so far, so good. The rest of your post scared me to the point where I almost peed my pants. But it's not anything that I haven't been wondering about.
Very interesting situation with your twisted wires picking up boradcasts. In my case, I could hear enough audio snippets to realize that they were coming from locations very far away from me - not local TV or radio stations.
You should charge admission to your demonstration of your audio problem. If it confounds and amazes co-workers, they should pay at least a little something for the experience, eh?
You make some great points above. And I was scared to death about what the rootkit might have been doing while it was distracting me with the audio. I did look at my network traffic on my switch (also 16 ports to support desktop computers and home infrastructure). I didn't see much, if any traffic coming out of the infected system, which at least sort of put my heart at ease.
I did boot the system from a recovery DVD and repair the O/S as part of my own troubleshooting. It didn't seem to work, or else the virus reinstalled itself upon boot.
I wish I had the skills to use a debugger and those kind of deep diagnostics you cite above. That would have come in very handy back then!
Damn it! Where the hell were you when I was going through all of this??!! I could have really used your help and would have paid handsomely for it.
Just to keep things straight, I wasn't going through some random security auditors or 'experts'. I contracted with a major security firm to have them fix this box. From what I could tell from watching what they were doing on my screen, you're absolutely right, they were primarily using the same tools that I was using before I contacted them.
I didn't see any evidence of them using a debugger, or checking processes, or doing any sort of deep dive into figuring out how this virus functioned. To me, I figured it would be a matter of finding out what process is compromised and then eliminating it. But, according to them, it's not that easy these days. They said the virus code could have been buried inside other processes - which would make finding it more problematic.
On the other hand, there was SOME process that was pushing audio to my sound card, would it be impossible to trace that chain back and see what was issuing those commands?
Ok, here we go, this is more like it...lol. The charge was for $100 per month and it was with one of the biggest and most reputable security firms in the industry. So i wouldn't put them on par with monkeys - unless you know some really smart and experience monkeys.
I ended up reflashing back way more than a few days and that seems to have solved the problem, while losing valuable work at the same time, of course.
You're right that my NAS wouldn't have helped me if someone had come in and encrypted all the files on the server and NAS box. That sort of 'data kidnapping' scheme is pretty tough to protect against. How would you do it?
Hold on, wait a minute. I didn't receive ANY benefit from their services. I thought I made this clear in the article, but maybe not.
The security company mainly just cost me time. They didn't tell me anything I didn't know. They weren't able to identify the virus, how it got into my system, or fix it. The did tell me that my current security status was top notch and exactly what they would recommend.
In their pitch to me, they said that they would certainly be able to fix my system without any intervention on my part. Was it an ironclad guarantee? Nope, but no one gives those types of guarantees on these types of services. The problem here is that they simply didn't have the experience or tools to handle my particular problem. Since the reason I purchased the subscription was to solve this problem, I feel that canceling it was fine ethically.
Great question. Here's what I observed:
1) I couldn't seem to find a specific process that fired up when the audio was active. The process list (hundreds of processes and services) looked identical when the virus was 'sleeping' vs. when it was actively playing sound. I also didn't see any particular process taking more memory or CPU than normal.
2) The audio played even when there wasn't any other application open and without me doing anything that would prompt a system sound. There also weren't any system issues that would cause an alert - at least nothing I could find with deep hardware and software scans.
3) The audio was unfamiliar to me. It wasn't playing anything that I had ever heard before (nothing from my media library, for example). It sounded like it was current or cached versions of web broadcasts, but the snippets weren't long enough for me to figure out exactly what they were.
NO! Not a stupid question at all! I spent hours wondering the same thing. I also closely checked network activity when the virus was active to see if it was busily sucking files out of my system or feverishly sending commands to a bot army. I didn't see much of any outbound network activity at all, and very little inbound traffic too.There also hasn't been any problems that are identity theft related either.
I think whoever penetrated me (whether I was targeted or it was random) probably used some code that was both sophisticated and amateurish at the same time. It was able to penetrate multiple layers of protection to get into my boxes and hide itself, but the payload didn't do much of anything other than annoy me.
On the back-up issue. I would heartily suggest that you keep more than three weeks of back-up images. One thing I didn't mention in the article (I didn't want to make it even longer and more tedious) was that one of the first things I tried was to use Windows System Restore to get back to an earlier version of the system. No luck on that, even with going back weeks in time.
So if I didn't have as many back-images as I did, I would have been even more screwed than I was. Storage is cheap these days, you can get a 3TB drive for less than $100. Get one and fill it up with back-ups, it's cheap insurance.
I see your point about remote vs. keyboard access. But from looking at what they were doing on my system, it looked like they did do a boot from an external source into a proprietary shell o/s.
It didn't work, of course. Which sucked. But is reformatting the only solution here? That's going to take a lot of time - longer than it took me to reimage a spare drive. I get your point that there's every possibility for reinfection, but I would hope our current set of tools can do at least a decent job of ensuring that your existing files aren't compromised.
I took a quick scan through the lastest Top500 list and, not surprisingly, didn't see any Itanium based systems on the list. It's not surprising, since these boxes were mainly sold to customers who need large SMP instances. There also aren't very many HPC apps available for HP-UX, which is a huge hurdle as well.
Yep, that's another great point and something that I've been considering. Most of the industry system components are manufacturered. Systems vendors today are much more assemblers and integrators than anything else. So how much difference is there really between a Lenovo branded system or a rebadged IBM set of boxes? Technically? Probably none. But according to procurement rules, probably it's the difference between a completed deal and a quick rejection.
You're making a great point in your comments above. Lenovo will absolutely face problems with trying to replace govt. and lab systems. How much of a problem? I'm not sure yet but am asking around. I'll post a follow up story as I find out more. Thanks for the great comment!
You think their PCs have gotten crappier over time? Hmmm...that hasn't been my experience. I had an original ThinkPad when I was with IBM for a few years in the late 90's until 2001. I was impressed with the design and durability - very efficient box. Just for fun, I priced out my business laptop as if I were buying it retail and was shocked to see how much the thing would have cost me.
After leaving Big Blue and starting my own company, I had HP, Acer, Fujitsu, and Sony laptops before giving Lenovo a try again. While the other laptops worked fine and were functional, I found that the Lenovo had that same durability and functional design that I remembered from my first ThinkPads. Moreover, when I started buying Lenovos, the prices had reached parity with the rest of the industry.
So, from my perspective and for my needs, their laptops are a cut above the rest. I haven't tried every laptop out there - or any other brand for the last three years, so maybe things have changed.
Ok, that one made me spit Sprite Zero all over my keyboard....great job....lol
Hmm...I'm not so sure it was as big a swap-out as you're saying. There are still a boatload of Lenovo folks in Raleigh, NC. How many of those are legacy IBM, I'm not sure. I also know that they've hired a number of IBM'ers to staff their earlier server efforts.
I seriously doubt that the carnage for the people moving from IBM to Lenovo will be anywhere close to what you're saying. They're picking up an ENTIRE division of completely new products. I'm assumnig that they truly do want to be a power in the x86 server market. That said, they can't afford to blow out the people they're going to need to market, sell, service, and design these boxes. It just doesn't make sense to me.
I hate the idea of an unpaid internship. If the intern is providing any sort of value, then pay them at least something to recognize that fact.
What's great about these student cluster competitions is that these kids are finding good opportunities out there - whether we're talking internships, advanced degrees, or full time jobs. I have another webcast coming up that talks with Cray's John Lee about this....
It's a short video, but longer than zero. It's working fine now - at least it did for me.
...but the name of the university is truly "National University of Defense Technology". These are the same folks who built/own Top500 chart topping boxes like Tianhe-1 and Tianhe-2
I'm a guy who needs to use a fair amount of the big Adobe tools like PS, Premiere, After Effects, Audition, and Illustrator. Before CC, I had debated buying either the package or the individual components I needed, but had always balked at the $3,000+ purchase price and then the extra money to stay current (assuming I wanted to stay current).
For my $49 per month, I get everything I need. The crossover point between me being ahead of the game financially is 61 months (3,000/49), meaning that I'm ahead of the game for five years. For me, this is a better deal than having to put out three grand to get what I need.
My analysis would be different if I had already owned any of these admittedly expensive tools. But in my case, Adobe's CC brought them a new customer that they probably wouldn't have had before.
Good idea, but that would have reduced the hyperbole in my sentence by perhaps as much as 37%. My goal is to provide all the information about the competition, but with as much hyperbole as I can possibly pack in. All of this at NO additional charge to you, the reader.
Ahhh...damn it, I should have mentioned the voltage! We're talking 26 amps at standard US power of 110 volts. They have two 13 amp circuits they can draw from. Sorry for omitting that key fact and thanks for pointing it out to me.
I see what you mean, but I remember the days when we had SMP systems using buses and crossbars with flat flat access times (ie. non-NUMA). The system I'm most familiar with, the Sun E10000, had a 16x16 crossbar with 1.3 GB/sec bandwidth at about 550 ns latency. The E10k scaled like a scorched weasel (which is pretty good) when compared with the systems of that era.
Today, you can get 56Gb FDR Infiniband that offers 6.8 GB bandwidth at around 700 ns latency. That's not too bad at all, and should be enough to ensure reasonably linear scalability on a good portion of workloads.
In any case, a SMP system should be able to scale better than a cluster running message passing. While sitting here writing out this reply, I can't think of any application that would scale better in an MPP system vs. SMP, but I could be wrong.