Slashdot Mirror
Intel to Take Online Suggestions for New Chips
hhavensteincw writes "Intel has quietly launched a new online community that it plans to use to take feedback and suggestions from OEMs and end users for new features in its vPro chips and management software. Intel envisions that the community will grow to allow users to get answers from other community members faster than Intel's support group can answer questions."
i want 2,000 cores!!!!!
Perhaps rather than hoping the community can outpace their support division, Intel should strive to improve their support division so they can always provide timely assistance to their customers?
I am a geek attorney, but not your geek attorney unless you've already retained me. This is not legal advice.
"Robert Duffy, Intel's online communities strategist, added that some of the impetus behind creating the community was to boost online traffic to Intel."
How about a 1Thz CPU with on board 1TB cache that only needs 1mw of power
Has anyone else noticed how great the AMD-Intel marketshare battle has been for consumers? Intel, in particular, seems to have woken up and begun providing really good CPU's, as well as trying to reach out to the community through things like this.
AMD/Intel should stand as a primary example of why honest competition is great for a market.
- Scott
...and full of floating-point errors. Lets hope C2D is just a blip and Intel gets back to doing what it does best.
And bring back slot 1. That really future-proofed a ton of motherboards.
""Intel has quietly launched a new online community that it plans to use to take feedback and suggestions from OEMs and end users for new features in its vPro chips and management software. "
Will I get a cut of the profits from my ideas?
The chip would have it's own personality.
Then, when I boot up Chippy, I'd hear "How may I serve you master?" I'd then boot Windows, open Word and begin typing. I suppose Chippy may interrupt and say "Do you really need me to handle this? It's rather simple." I'd then open seventy five applications and begin decoding the genome.
Chippy would interject "This is a lot for me to handle master. Can you not have me work so hard? It's getting hot in here!"
I'd then open up the interface and change it's name to "Pinky". Sure, Pinky may protest, but unless he kept quiet, I'd open 30 pages of Flash.
Does this mean I can say pretty please and intel will put altivec into their chips so h.264 encoding isn't such a dog?
Get a web developer
Femmebot parts.
And be quick about it!
There exists no way of exchanging information without making judgments. --Bene Gesserit Axiom
*(Cores are process-shrinked versions of the Intel 8088) I'd like to see Intel try making some massively multicore CPU, even if it's just a 64XScale. A joint venture with a company whose name sounds like it comes out of superhero comics would have to be called Super-Duper-Threading.
..........would you like to take survey?
and Ranch, please.
... that's really crispy and comes in a can. Oh yeah, nacho cheese flavor wouldn't be bad either.
That is all.
Drop the Treacherous Computing chip?
Even though Intel is not going to do this in the foreseable future, at least not in a non-EU release (there's a chance our legislators may wisen up... oh well, whom am I kidding?), yelling loud enough and often enough may at least give Intel a hint that they're doing something wrong.
The creatures outside looked from Alt-Right to Antifa; but already it was impossible to say which was which.
"Drop the Treacherous Computing chip?"
Repeat after me. TPM isn't DRM! TPM isn't DRM! Got it? Good!
If Intel wants to serve the community, I vote for an on-die women interpreter.
If you mod this up, your slashdot background will turn into a beautiful sunset!
Intel has nothing to lose by documenting all the instruction sets, architecture designs etc. They have such a big brand name - it doesn't really matter if their designs became public.
It is quite sad that despite their chips being 100s of times faster than a few years ago, so-called 'partners' and OEMs like Microsoft have given the x86 series a bad name - resulting in little or no incremental performance gains for the user community.
Like HP made winprinters and some vendors made winmodems to the customer's ire... and the perennial problems faced by video and audio device mfrs. including big names like Creative... it is clear that the biggest OEM, namely Microsoft determines what customers get to see of "Intel Inside".
The recent thrust towards Open Source drivers for wireless cards from Intel is a very small and incomplete step. Recently at my firm, we talked to Intel for sourcing a 1000 laptops for students joining our colleges. Intel said they would share roadmaps and plans under NDA!!
This is a far cry from 20 years ago when Intel gave out the complete instruction sets and architecture layouts for their 8080; I sort-of remember the Zilog Z-80 did a better job of implementing them. Unless Intel come clean in favour of the truly Open source model, they risk small time players making it big in niche segments - including the biggest niche of them all - the PC market. If not Negroponte, someone else will come out with a non-Intel platform for under $100 and Intel will go down pulling others like Microsoft behind them.
If you keep throwing chairs, one day you'll break windows....
They should make a chip out of a potato. A potato chip, if you will.
Or how about a chip out of paint?
Perhaps a chip from someone's shoulder...
I'd like a chip with a higher clock speed. I'd like a chip that doesn't cause the lights to dim around the house when I power it up. I'd like a chip that doesn't require a heatsink the size of Guatemala and a fan with the power of a small tornado. I'd like a chip that doesn't glow like the surface of the sun if you remove the heatsink.
I've read that the reason Intel / AMD are going parallel rather than increasing clock rate is due to the problem of heat dissipation. Multi-core is great for some apps (web-server farms, simulation), but is not going to speed up most (single-threaded) apps. Dual core is nice. About the time the industry is going from 16 to 32 cores, I doubt most users will care - or bother to upgrade. And if the heat problem is not solvable - that may be a serious marketing problem for chip makers and computer manufacturers.
[Insert pithy quote here]
You made it that way... you deal with it!
ROFL!!!
They're so naive
All able to be stack pointers or be the program counter, as well as containing arithmetic and logical operands.
Make them out of potatoes... and etch them with salt and vinegar.
Sounds like they need new material and need to down more craft.
I certainly hope Intel is willing to financially compensate the people whose suggestions they end up using! "Thanks for the idea! Now please step back so we can reap the billions!"
I'd like to see something like an FPGA onboard with a compiler (or device driver model) that can allow us to take some time consuming things such as CODECs and push them off into hardware.
The last 4 front page stories are from IDG shills.
Faster, cheaper, less power!!!! What else is there to ask for?
There's the rants from the green party I suppose - and the "stop acting like a monopolist" crowd.
No sig today...
I hope for real innovation, like in the cell-phone market. I want a CPU in blue and yellow with a camera and another in pink with sparkles. OMG could they make it in the shape of a pony!
It must have been something you assimilated. . . .
Give the on board video chips some of there OWN RAM you can use a system like ati hypermemory and nvidia turbocache.
Open up the xeon cpu to chipset links so you have more choice in chipsets like AMD systems do.
Dump FB-DIMMS from xeon systems or make the same chipset with FB-DIMMS or DDR 2/3 ECC. The new xeno chipset with 2 pci-e 2.0 x16 slots should be FB-DIMM or DDR ECC.
Make the new chipsets with all pci-e 2.0 slots not some 2.0 and the rest 1.1 yes the new xeon chip with pci-e 2.0 will only have 2 slots with pci-e 2.0.
Go to true quad-core not 2 dual's linked by FSB.
Dump the FSB and go to the HT bus.
Intel, if you could make the computer boot straight into AOL on startup, that'd be great. And do something about all those viruses and credit card fraud, and make it so the games on Pogo.com win more often.
We at slashdot are scientists, specialists and kernel hackers. Your FUD will be found out.
...sitting on the EV9 Alpha, finish it and release it?
They have the technology. They can rebuild it. They'll be better, faster, stronger than they were before.
Oh, and use less power too.
The higher the technology, the sharper that two-edged sword.
Many people use slow dynamic languages, like Python, today, because they are so much nicer and easier than static languages.
Like LISP machines Intel could throw in some dynamic type instructions, reference counting/garbage collection, or even hash/dictionary mechanisms. Then our dynamic languages could fly, assuming someone wrote a compiler to support all that...
Intel doesn't listen to its own engineers (mm I am posting Anonymous, wonder why?)
What makes anyone think that they would listen to the general public?
At Intel you can fuck up major projects costing millions of dollars and still get to keep your job. However, if you are not on your group's staff and you offer up painful criticism your career is over. And remember you can't transfer out of a group at Intel unless your leaving manager OKs it.
I guess what I'd like to see is a Tri core CPU setup with two proccessors and something akin to a graphics enhancement chip that could be combined with a GPU from either nvidia or ATI to increase graphics speeds by sharing specific parts of the graphics load and/or keep things in the laptop cool so us even casual gamers aren't burning vital organs when we leave it on our laps.
1. Ranch
2. BBQ
3. Salt & Vinegar
4. Nacho cheese
5. Ridges!
Slashdot should never say something was done "quietly". Once it's been on Slashdot, the quietness ends.
Like Sun with the Niagara Processor. That would be cool.
I'd be quiet about this too if I were Intel. This is a stupid idea. Half your end users (including me) couldn't care less about what chip they have in their computer as long at works. The other half of your end users want the chips in pink or with an integrated LED. Either way a forum like this will just piss people off, because even the good suggestions aren't going to mesh with their five-year development schedule.
You mean die laughing at the Apple systems that refuse to boot on the new Intel chips?
And the Apple group will then say the same about Windows, who, in turn, will say it about Linux users, who, realizing that they're heading into an infinite loop, change it up a bit and pass the ball to the BSD camp.
Of course, they decide the ball should be useable by everyone and that changes to the ball need not be distributed back to the community.
Oh my God. Fellow Linux users, I implore you, when the ball comes to you the 2nd time, enter that infinite loop. If you pass it to the BSD camp, random namecalling and flaming will insue and we'll have nobody to blame but ourselves.
</post>
APK quotes people (including myself) without context and should not be trusted. Just thought you should know.
if they would just give me DWIM functionality in hardware. It's just so slow running that in software, you know.
I want an Donald Knuth signature MMIX chip.m l
http://www-cs-faculty.stanford.edu/~knuth/mmix.ht
that gives worms to ex-girlfriends.
Spork.
P.S. Spork.
I would be happy if they released a motherboard with a user programmable TPM chip. In particular, I am looking for a chip that can be used for general purpose cryptographic functions, that can be reprogrammed with a different (user known) endorsement keys, and that can permanently disable remote attestation and other chip dependent remote and/or configuration based DRM functions.
-Valen
I've read chatter that SSE5 is suppose to help in this area, as it is altivec-like?
No, I didn't think so.
How about DS-UWB?
No, I'm not surprised about that, either.
Don't forget: Cow chips!
*fart*
Then some corporate drones looked at what was happening and though "how can we take advantage?" So they got the "each contribute a small amount" part but overlooked the "everyone takes advantage" part. The corporate version is more like "everyone contributes a small amount and the corporation takes advantage". Many corporations have tried this plan and they've been left wondering "what went wrong?"
So here comes Intel - they're asking the people to contribute ideas and then they'll take advantage of them. We've seen this play out before and the result is always the same. Hey, Intel - if you really want people to do your work for you, you need to include some way to compensate them in your plan. You didn't really expect them to do this for you for free, did you?
I suspect they did - and when this plan fails miserably they'll pick some unfortunate person in their corporation to take the blame for the failure. They'll never for a moment think that their plan was flawed and doomed to failure from the start...
Why not actually enter the GPU market?
I don't mean the current minor onboard garbage they're putting out now. I mean real, high end chips to combat the GeForce 8800 series or the Radeon x2900 series. With their own GPU development department, and their open drivers, they could really blow open the market.
Why not?
when was the Slashdot effect so nerfed that it's now considered "quiet"?
one suggestion I would make is bring down the cost of mainstream CPUs to a more affordable price, like $10 or so. That would be nice. Thanks Intel.
With all that DEC intellectual property they managed to grab, bring back the Alpha and make a 72-bit PDP-10.
"How come there are 35 requests for implementing multimedia extensions to accelerate the translation of Esperanto to Klingon?"
"Because they all came from AMD's ip space?"
They needs to add some mad V-TACH yo! All the useless carbon fiber they can fit around it! 20's with spinners and a momo wheel!
Balderdash!
I doubt anything interesting would come out of it. But ... let them try.
My personal wish-list was always made of improved string operations:
- support more operations (or rather allow any operation - add, sub, mul to stringified),
- support source and destination increments to allow string operations to work on structures,
- handle all the alignment idiotism internally: when possible source and destination pointers should be aligned internally to get most out of string operations on plain arrays,
- remove implicit registers, allow registers to be supplied by programmer (though i'm not sure how that valid on IA-64 - on i386 it was always pain to restructure whole program to make special registers available in some place).
Also, I would appreciate a string instruction to perform quick search in array. For example such operation could perform a single step of dichotomy search on array pointed by EDI of ECX size. The op should update EDI and ECX to be usable with REP prefix. If the op would also support increment - to allow to search in array of structures sorted by field, there would be no end to my happiness.
String operations became kind of bastards now, since they are so stupid that nobody uses them. What's more, 200 op implementation of memcpy() is faster compared to plain REP MOVSB/W/D, what is really really stupid. Why we need the string operation altogether then???
All hope abandon ye who enter here.
Make something with the equivalent power usage of Via's Eden 15000, but faster. Surely Intel has the research budget to accomplish it too.
I want a small, fanless computing appliance that is going to last 20 years or more with zero maintenance other than software. No dust, no noise, no ticking time bomb spinning parts and electrolytic capacitors. Something that will not require me buying a huge solar panel if I want to go that route. If I have data storage needs, USB, firewire or eSATA external hard drive enclosures will suffice.
If I have seen further it is by stealing the Intellectual Property of giants.
Some on board reed-solomon decoding, and AES and PGP decryption/encryption.
How about a version without the TCPA / Trusted computing crap? The vPro platform was the first to fully implent this technology.
Great job on the new interface taco... I accidentally clicked "redundant" instead of "insightful" so now I have to post to undo my moderation.
Nothing to see here. Move along.
Here's to the crazy ones
... of closed proprietary profits and IP.
Does anyone keep a prior art log of the suggestions?
Why do you want to remove implicit registers? The nice thing about implicit registers is that it allows version n+1 of a chip to have more registers than version n, and allow code compiled for version n of the chip to take advantage of them. I'd like to go in the other direction, and make all instructions vector instructions with an arbitrary (power of two) length on both operands, and have ones with too long vectors decomposed into short vector / scalar micro-ops, so you can add 256, or even 512-bit vector units to a later chip and automatically use them.
I am TheRaven on Soylent News
Since Sun Microsystems is GPLing their latest processor designs (T1 and T2 at opensparc.net) they reaised the bar for a 100% OpenSource systems (OpenSolaris/OpenSparc). Something that Linux/x86 can`t achieve becouse of x86 closed ISA and closed implementations.
We should all request a GPL implementation of their latest processors..
I would like to see a good system of identification for hardware (either PCI, AGP or USB). I often have to install older equipment from which the installation disks are lost. It is often difficult to find the drivers. Some network cards don't even carry decent names on them.
Each piece of hardware should carry:
1) a link to a website where drivers can be found
2) a unique ID so that if the website if offline (company broke or domain hijacked) you can still search in an easy way on driver sites like drivershq.
3) if there is place (and with the falling memory prices that should be increasingly the case: a mini driver for the most common operating systems.
I'm not saying to remove default.
Think about it. Now you can use only one register set as input/output to string operations. Let's call it default. But why to limit ops only to one set of registers?
Main problem of IA-32/64 optimization was always lack of registers. And this is also what contributes to sparse use of advanced CPU commands - that they often require special set of registers. And at the point where particular op might be useful, other optimization could have been already made and required registers are already used up. Or worse: only one of the required registers is already used. The whole optimization is wasted due to single register overlap.
I like string ops on PowerPC. They are not that cool as on Intel, they are fewer, but they are more useful. Let say, in 4 years I have programmed assembler on Intel, I used string ops on fewer occasions that I used string ops while I coded stuff on PowerPC in half of year. On PowerPC, string ops are limited to loading/storing content of up to 8 registers (== 32 byte read/write). And that's very convenient.
All hope abandon ye who enter here.
An anonymous source from AMD had this to say; "Yeah! Well, we're gonna go build our own vPro chips. With blackjack and hookers. In fact, forget the vPro chips."
Carbon based humanoid in training.
make them faster? add more cores? optimize cache hit/miss/latency. don't add any more instructions (for god's sake is a damn risc processor underneath all of that x86 goo). provide for virtual machine monitor special cases... faster--
Has anyone else noticed how great the AMD-Intel marketshare battle has been for consumers? ... AMD/Intel should stand as a primary example of why honest competition is great for a market.
... Intel, in particular, seems to have woken up and begun providing really good CPU's, as well as trying to reach out to the community through things like this ...
AMD is no friend in the sense that they relegated us to the x86 architecture, hampering the periodic move from one CPU architecture to another. Intel tried to drop x86 and move on to something new. Under Intel's "plan" if you wanted 64-bit you were supposed to go to Itanium. It was AMD that relegated us to x86 by introducing 64-bit x86 and continuing to innovate that line, forcing Intel to do likewise. Whether or not Itanium was a good design is irrelevant. Intel would have put pressure on developers and businesses to migrate off of x86, and some may have moved to Alpha, PowerPC, etc rather than Itanium. We would have had a more diverse and healthier CPU architecture ecosystem.
I don't think so. While AMD caused Intel to continue x86 development, I don't think AMD had that great of an effect on the progress of this development once it was undertaken. I think Intel stumbled with the Pentium 4, it turned out to be a limited design. I believe they would have backed up and continued down the Pentium 3 path, Core Duo originates from the P3 line not the P4, regardless of what AMD was doing in their CPUs. What AMD did was make things go a little faster.
The problem with the instruction set is not due to the chipmakers but because there is an awful lot of proprietary software ( in particular windows ) which relies on it. Just have a look at Linux, the BSDs and Solaris. They have all been ported to numerous architectures, but this just isn't possible with a closed source application unless the vendor decides to do it. ...
... As a consequence Intel and AMD has no choice but to continue using x86 because so much software depends on it, and it would be suicidal for them to stop supporting it.
7 1&cid=20522223, perhaps it would be best to discuss this subtopic there rather than have a redundant thread.
Proprietary is largely a non-issue. The Windows NT 4 retail CD contained x86, MIPS, Alpha, and PowerPC binaries. Customers who wanted performance, very few, went Alpha. Customers who wanted price, nearly all, went ix86. Vendors of proprietary software that catered to these two markets developed accordingly. Also, the vast majority of proprietary software that locks businesses into Windows and hampers migration to Linux is internally developed. Businesses could have ported their software to whatever CPU they preferred. Hardly anyone was locked into x86, they chose x86. Much as those who chose Linux and BSD overwhelmingly chose x86.
Not true. Intel tried to drop x86 and move on to something new. Under Intel's "plan" if you wanted 64-bit you were supposed to go to Itanium. It was AMD that relegated us to x86 by introducing 64-bit x86 and continuing to innovate that line, forcing Intel to do likewise. I've commented on this in a different thread, http://hardware.slashdot.org/comments.pl?sid=2914
All that SSE-bazillion stuff is nice but how about:
A better timekeeping feature than TSC and HPET. TSC isn't necessarily synced between cores, and HPET isn't fast enough or ubiquitous enough (it needs to be on a mandatory chip).
And also stuff that'll help make-
synchronization easier (and across cluster nodes too)- mutex, locks, semaphores etc
doing things atomically easier.
Things like epoll/select more efficient (or allow the creation of something even better?).
"Wait for Event" easier/more efficient - maybe something like "When memory location A changes to hold the value V, raise interrupt and set a register X to value F". Doesn't look scalable though so hopefully someone smarter can solve that (make it hierachical?). Maybe it won't help much?
It would also be good to have a separate stack for data/parameters from that used for return addresses, currently it seems common practice to mix return addresses and parameters in the same stack which is "bad hygiene" (poorer security). I'm also guessing that not mixing the stacks could make it easier to do branch prediction.
Or just ignore me and talk to people who'd know what hardware features would make O/Ss, clustering, databases more efficient (and safer).