Go Back   AnandTech Forums > Hardware and Technology > Motherboards

Forums
· Hardware and Technology
· CPUs and Overclocking
· Motherboards
· Video Cards and Graphics
· Memory and Storage
· Power Supplies
· Cases & Cooling
· SFF, Notebooks, Pre-Built/Barebones PCs
· Networking
· Peripherals
· General Hardware
· Highly Technical
· Computer Help
· Home Theater PCs
· Consumer Electronics
· Digital and Video Cameras
· Mobile Devices & Gadgets
· Audio/Video & Home Theater
· Software
· Software for Windows
· All Things Apple
· *nix Software
· Operating Systems
· Programming
· PC Gaming
· Console Gaming
· Distributed Computing
· Security
· Social
· Off Topic
· Politics and News
· Discussion Club
· Love and Relationships
· The Garage
· Health and Fitness
· Merchandise and Shopping
· For Sale/Trade
· Hot Deals
· Free Stuff
· Contests and Sweepstakes
· Black Friday 2013
· Forum Issues
· Technical Forum Issues
· Personal Forum Issues
· Suggestion Box
· Moderator Resources
· Moderator Discussions
   

Reply
 
Thread Tools
Old 11-13-2007, 03:20 AM   #1
MVR
Junior Member
 
Join Date: Sep 2007
Posts: 17
Default Advanced Disk caching technology built into MB BIOS

I've had an idea running through my head for a couple years now. I've approached various SATA raid card manufacturers with it and they've all replied that they just don't see a market for it. That or they've told me that only the enterprise market (EMC, IBM Storage, etc) will ever feature such technology.

Basically, I'd like a raid controller with, say 8+ DIMM slots (compatible with 1GB-8GB dimms), that runs some advanced caching software for functionality such a pre-caching (most used, associated with x, x data dependant on y data, data types, etc) files, directories, applications, etc. I won't go too much into caching technology since this isn't the complex piece. There is plenty of disk cache software and technology out there.

Since none of the raid controller card makers seem to want to take the leap, perhaps next generation motherboards could implement this. Since memory is getting so cheap and motherboards now seem to support 2GB natively with 4GB and 8GB dimm support coming quickly. Why not have an enthusiast class motherboard that allows you to set aside x GB of memory to feed a BIOS level caching program?

This article gives me hope. I like the idea of vitualization becoming a commodity. IMO, there is no reason that a motherboard shouldn't feature a usefull level of pre-OS boot technology. Especially when 90% of the technology is already open source or licensable.

Now for those of you about to reply "That is what these awesome SSD's are for man!", don't bother. SSD performance is still, IMO, rediculously poor. Also, I am aware there are some ram disk products such as the HyperDrive4, but I don't see these being practical unless they reach commodity pricing. Not only that, I think I'd rather the data not go through several transport layers to achieve what a bit of hardware on a PCIe card or motherboard could pull off.

Also, I realize that 64bit Vista will support all the ram on the motherboard. But I still feel BIOS level virtualization and caching technology would be useful as most people will be stuck on 32bit OS for quite some time.

Since motherboards will always have a limited number of DIMM slots, and there is a need for a secondary memory resource pool, I also have an idea for a multipurpose memory expansion card I've overviewed here.

Thoughts?
MVR is offline   Reply With Quote
Old 11-13-2007, 07:58 AM   #2
QuixoticOne
Golden Member
 
Join Date: Nov 2005
Posts: 1,855
Default Advanced Disk caching technology built into MB BIOS

Hmm

In general I agree with you.

In specific --

* I don't think most people will be "stuck" with a 32 bit OS that long.
Micro$haft will try hard to RAM Vista and its successor down peoples' throats
whether they like it or not (e.g. ensuring MOST PCs will be bundled with
Vista or their OS lavor of the day without any real consumer choice).

LINUX is already trivial to install on just about any 64 bit capable PC
(i.e. most anything made in the last 3 years). I've run LINUX 64 for years.

I'm sure when the motherboards are SO powerful and the RAM is SO cheap
as it is today, and this powerful technology percolates out into the consumer
desktop area for another year, Microsoft et. al. will have no choice BUT to
provide cheap (relative to MS's usual) upgrades to a next generation
OS since it will be painfully clear that 32 bit is a pathetic joke, incapable of
using more than a trivial amount of RAM, dealing with the
$200 TERABYTE drives people will have, dealing with the 3rd
generation DX10+ video cards, etc.

MS is already RUSHING "MS HOME SERVER" OS to market because they
realize they're pathetically unequipped to control / satisfy the home
file-server, media-center / media-server
(digital photos, documents, movies, TV, games), et. al. markets for personal use.
Admittedly it's an abortion of a product, but even so you see media center types
enthusiastic about it because .... well ... it's the only game in town if you want to
use your Microsoft DRM infested media and XBOX media center stuff etc. etc.
Clearly they'll soon realize it sucks and Microsoft will have a more polished
generation 2 product in a year or two.


* Memory is dirt cheap now, true, and it's quite trivial to get 8-32 GB of
fast RAM if you want it; I'm already at the 32GB mark and climbing for just
some of the reasons you mentioned wrt. performance and memory / data hungry
computing. Clearly digital video, DVR uses, digital photography, and
"digital lifestyle" permanent archival desires for documents / photos etc. will
continue to drive the exploding growth of storage capacity RAM / disk, but also
the sophistication of applications / devices to USE that capacity.
Intel and Microsoft et. al. are big on the concept, and recently there's
been news from some Intel guy about terabyte data stores being needed and
"coming soon" for the average home to satisfy these consumer media needs.

* Problem is even though RAM is DIRT cheap, Hard disk space is like
*EMPTY SPACE* cheap -- RAM: 12 Gigabytes = same cost as DISK: 1 TERABYTE,
so that's like 83x cheaper for disk vs. ram and it'll continue to spread out in
disk's favor since disk makers are selling drives for more or less the money
they WANT to sell them for, whereas memory makers are lately selling on
uncomfortably thin margins due to a glut in the market and they WANT RAM
prices to climb, WANT DDR2 to be obsolete, and WANT to start selling more
expensive RAM like DDR3 to consumer / OEM markets at much higher revenue / GB.
So given the 83:1 cost difference between disk and RAM, RAM "cacheing" of the
disk or replacement for the disk becomes RELATIVELY unaffordable and useless.
What good is a "cache" that's like 1% of the size that you would use in less than
2 hours -- chances are high the data you WANT won't be in RAM!
I.e. 1 DVD = 4.7GB = $100 of cheap RAM, or $7 worth of DISK.
Browse through your PC-virtual-TIVO / movie-on-disk collection and suddenly
you're navigating in "real time" through hundreds of gigabytes of MP3s, 'DVDs',
DVR shows, digital family photos everyone in your family took over the last 10 years,
etc. Basically you need "real time" random access to lots more data than fits
in any affordable amount of RAM, even given the stupendous capacity of RAM per
the dollar.

* There are hybrid technologies like hydrid hard discs and flash storage and
emerging technologies like nanowire based memory, spintronics, ferroelectric memory,
nanotech hard disc drives, etc. that will not have the mechanical reliability problems,
fragility, or "high costs" of current rotating disc media. q.v. recent articles about
copper nanowire memory research and terabyte capacities cheap, etc.
You start to get "zero" or greatly reduced seek times, greater read bandwidth,
true random access, and long lifetime. Though it'll often still be an order of magnitude
SLOWER than "fast RAM" that might be the PC standard, it'll still be fast ENOUGH
and cheaper ENOUGH than expensive FAST RAM that it'll be a likely preferable
intermediate storage format for "quick random access" stuff as compared to whatever
kinds of true ARCHIVAL media end up replacing hard drives for terabyte level
cheap personal storage that's fast enough to spool video in real time (the killer
application for storage these days). Maybe the latter will be holographic storage
or nanomagnetics or whatever. So this doesn't make the cost effectiveness
of MERE gigabytes of RAM look good in the 3-10 year time frame.

* In the shorter term, who NEEDs *fast* access to 1-200 gigabytes of data?
Well consumers don't because the applications, OS, etc. aren't there to take advantage
of ANY of the stuff we need for making it all USABLE to average consumers for
media, documents, video, etc. TIVO hasn't replaced the VCR, but they're working on
it. Give it another 2 years, etc. and then information "appliances" will be ubiquitous
and people will start to think about more intelligent ways to architect the components
like OS, hard drives, RAM, applications, etc. to take advantage of it all.
Medium size business? Sure, ok, but any business over the size of about 20 people
probably ALREADY has 64-bit "enterprise class" computing OSs, file servers on
64 bit OS, file servers with doubly redundant RAID arrays, server class motherboards
which *DO* take many gigabytes of RAM and use them effectively for disk caches, etc.
So basically they'd say "yeah we could use it on our file server / web server, and,
hey, we already have it because we're using blade servers with 8 CPUs and 128GB of
RAM in front of RAID arrays, so anything that needs quick access is already in RAM or
is fast to pull from the disk array". Yes, it's hideously expensive hardware
for the servers and mainframes compared to the cost of PC parts, but it does
the job, the only job(s) they feel they need it for, and it's an acceptable
"capital equipment cost" to buy $50,000 server boxes with big RAM RAID caches etc.

* Should *everyone* have better use of their their RAM (e.g. ram disks, disk
caches that don't suck, operating systems that don't page/swap even though you're
JUST using a web browser and email application on a computer with MORE than enough
RAM, etc.)? SURE! It's one of the main areas Micro$haft claims is improved in
VI$TA, i.e. ReadyBoost, their "smart" prefetch / cacheing stuff, etc. Obviously
it totally sucks, but it sucks a lot less than XP, and they're CATCHING ON, SLOWLY.
Anyone CAN make a SSD or RAM disk software or whatever for Windows, it's
been done, and anyone can make a SAN drive / network attached storage unit,
etc. that lives outside the OS and has nice cacheing, fast response, intelligent
indexing, auto-defrag, auto-backup, security, whatever. It's been done and
small and large businesses buy this stuff all day long. The average USER
doesn't buy it or even install the ramdisk software because, well, they'd be
DELIGHTED with the speed of a PC that doesn't bog down, but it's just a 1/10th
solution to the OVERALL problems of managing / organizing / backing up / finding /
searching / storing / securing your files and media and whatever. Until the REST of the
OS / applications starts solving ALL those problems, it's going to be too "complicated"
(i.e. all solutions suck too bad in cost/usability to be bothered with them) for the
average user's uses. They'd BENEFIT from it just like they'd BENEFIT from
a 64bit OS, but nobody wants to be concerned with the DETAILS, it should JUST WORK,
transparently, and NOT depend on specific OS / APPLICATIONS / HARDWARE MODELS.
We're in a world where you can't even SYNC your PDA-phone nicely or sync files
on a thumb drive nicely or even have a half way decent photo-album application.
Nobody wants to talk drive letters or be worried about backups and bottlenecks and
accidental deletions, they want high performance access to the CONTENT, just
select a movie and hit PLAY, browse everything like a web site (or actually a lot
EASIER, hopefully), etc. So a faster CPU, more RAM, bigger DISK, whatever
can help you improve the speed at which everything that's a broken unmanagable
pain in the a$$ is a broken PITA, but speed/capacity (of CPU/DISK/MEMORY) hasn't
been the core ISSUE since 80386 days; it's all about seamless performance over
the use-cases of peoples' work-flows. There are MORE than enough gadgets /
components i.e. RAM, CPU, DISK, etc. to satisfy peoples' basic needs, but
the WAY in which it's all (FORCED -- by bad DESIGN) used is broken and
better "gadgets" won't fix it, it's a band aid where a tourniquet needs to be.
ENTERPRISE/BUSINESS customers DO appreciate the needs for PERFORMANCE
in their storage systems, but as I said they ALREADY pay for the (more or less)
right hardware for the job.

* Most of the RAM in PCs is ALREADY used about 90% for disk cacheing, given
that even on 32 bit OS versions your typical program still isn't more than about 50MB
in size, but its frequently used DATA is maybe up to 4GB in size. e.g. tiny bit for
CODE, ALL the rest for file data in a "most frequently used" or "most recently used"
access scheme. Pretty bad implementation in XP, better in VISTA, even better on
servers and the MAC. Sure you can multiply that out to 256GB RAM vs 2GB
of today's typical RAM, but it's not going to make a huge difference to MOST people
who email and browse the web (speed limited by the NETWORK, not the LOCAL DATA).

Gamers who are savy about laggy performance already know to buy 2GB, 4GB
or whatever lets their "load times" be basically zero because it's all in disk cache
after the first use. People with databases and such obviously tune this way too.
Usually 4GB (32 bit limit) or maybe 8GB (affodability limit) is pretty much enough
for a single-user PC since once your video game, your web browser, your email client,
and a few bloatware desktop applets all fit in RAM along with their most used data,
you don't benefit from more RAM or more disk cache whether that's ramdisk,
kernel level disk cache, or a network attached RAM cache / SSD.

What would make the need for more RAM cache a lot higher?

Multi-user systems; oops; beyond the "home computing net" serving the family
of 4, most people don't need it. Bigger oops, Microsoft is ADAMANTLY opposed
to "multi-headed" PCs i.e. hardware *OR* SOFTWARE that lets more than one person
get ANY use of a PC at once; they WANT you to buy 4 copies of the OS, 4
different PCs, 4 different monitors, keyboards, mice, 4 times the hard disk space
you need, etc. because they get paid that way. Windows Home Server and
"media center" are the ONLY concessions they've even made to encourage sharing
FILES within a home environment, and that's certainly NOT intended to create
a situation where more than one person's APPLICATION is running on one PC,
so single tasking single-user systems RULE while Micro$haft has ANYTHING to do
with it.

What else makes the need for more performant RAM cache a lot higher?
Multi-tasking (even single-user). Oops, 4GB is already enough to do all the
multitasking most people feel they want/need, web, email, multimedia, games,
office apps. It's hard to sell more than that into the consumer space, and
the business server space has it solved already.

So while I think your ideas are basically correct in identifying OPPORTUNITIES
to improve the DESIGN and PERFORMANCE of storage/computing systems
via engineering of the sub-components and low level software systems,
I don't think you or I or anyone has the PERSUASIVE CONTROL over the dominant
ROADBLOCKS to garning these seemingly "low hanging fruit" benefits.
Yes we all send in (politely) suggestions like "Hey, STUPID Microsoft,
DO IT THIS WAY, IT'LL WORK A LOT BETTER!!!", it's not ABOUT efficiency,
it's not ABOUT design, it's about PRESERVING THE STATUS QUO, POLITICS,
and MAKING A BUCK not *INNOVATING TO HELP YOUR CUSTOMER*, but
*SCREWING* your customer until they'll PAY you to stop (at least a little bit,
then maybe you screw them in a DIFFERENT way).

Look at WalMart, etc. Hear "race to the bottom"? Nobody wants to make
QUALITY DESIGNED products, they want to make everything as CRAPPY
as POSSIBLE just as long as you can sucker a few million people into buying
JUNK that'll break within 90 days (despite leaving the user profoundly UNSATISFIED
with the efficiency of the product), it's GREAT, because you saved $0.0001 / unit
and sold 10,000,000 units so now you can retire rich, and everyone gets to go
buy something 'new' and slightly DIFFERENTLY defective-by-design to replace
the previous cr*p.

New USEFUL features?
Better quality?
Integration?
Convergence?
Efficiency (energy, usability, or otherwise)?
You MUST be kidding. Save that $0.0001/unit and DAMN the rest.

Sorry, I wish it WAS all designed by people that CARED about
POSITIVE EVOLUTION and making things as good as they CAN BE, but
that's not who's in charge.

Write the 'killer application', make the next killer 'MP6 player' gadget,
whatever, you'll do FINE, and maybe cause a paradigm shift that'll make
people notice and emulate you. (look at MacOs vs VISTA).

But until you can get about 1,000,000 people to care/notice or
purchase your idea / meme / product, it's not going to become popularized
since Microsoft / Apple / Intel / Seagate / whoever will just value the status
quo they're happy with and see no reason to INNOVATE or be EFFICIENT.

Look at Firefox... started small, free software, now it's driven Microsoft
to redesign IE to emulate it because enough people realized how badly
the old alternative IE SUCKED and that there WAS a better choice.

They won't do that for a COMPONENT or a DRIVER, but write a better
APPLICATION or make a better APPLIANCE (e.g. media center / media player /
whatever), and then people will start to appreciate the usability advantages,
demand it, clone it, but they still won't CARE HOW or WHY it works, they'll
just like it THAT it works. Don't beg Seagate/Intel/etc. Make their CUSTOMERS
beg for your new and better "system", THEN they'll race to implement your ideas.

QuixoticOne is offline   Reply With Quote
Old 11-13-2007, 11:33 AM   #3
MVR
Junior Member
 
Join Date: Sep 2007
Posts: 17
Default Advanced Disk caching technology built into MB BIOS

Holy crap friend, I like your style.. verbose mode on eh?

Quote:
Originally posted by: QuixoticOne
Hmm In general I agree with you.In specific -- * I don't think most people will be "stuck" with a 32 bit OS that long.
Micro$haft will try hard to RAM


I know they will try. And heck, I'd do Vista 64 bit in a heartbeat, but I know for a fact serveral of my online games still are buggy with it . Also, I find it quite disturbing that even the most simple of application are now using rediculous amounts of memory. I run MIRC and it takes 4MB on its worst day. What happened to coders not building this wasteful crap? Here are my top 10 offenders at the moment: LOTR game.dat 832MB | iexplorer.exe 154MB | outlook.exe 101MB | MsMpEng.exe (win defender) 58MB (and CONSTANT cpu usage!) | winword.exe 55MB | MySpaceIM.exe (crap) 44MB | msnmgr.exe 42MB | Slype.exe 43MB | SkypePM.exe 24MB | googletalk.exe 19MB (shame on you google) | Sorry to rant that off, but I just don't understand why simple apps need this level of resources.

I'm sure when the motherboards are SO powerful and the RAM is SO cheap
as it is today, and this powerful technology percolates out into the consumer
desktop area for another year, Microsoft et. al. will have no choice BUT to
provide cheap (relative to MS's usual) upgrades to a next generation
OS since it will be painfully clear that 32 bit is a pathetic joke, incapable of
using more than a trivial amount of RAM, dealing with the
$200 TERABYTE drives people will have, dealing with the 3rd
generation DX10+ video cards, etc.


I agree. I think there is another awakening coming for hardware developers who are stuck with the ideas that 32MB cache on a TB drive is alot, or 1GB cache on a SATA raid controller is plenty. Everything tech (that works) becomes commodity in time. I'm hoping that with further mastery of virtualization, I can soon run Vista 64-bit and launch sandbox instances for things like my 32-bit games.

MS is already RUSHING "MS HOME SERVER" OS to market because they
realize they're pathetically unequipped to control / satisfy the home
file-server, media-center / media-server .....


I want a completely different approach to the home server. I've written up a fairly complex proposal for a hugely scalable architecture for a central home server that in its most basic form uses SBC (Server based Computing, otherwise remembered as the NC or Thin Client) to provide sessions to small "thin client" like devices connected to HDTVs, monitors, etc. Basically what SBC is missing today is the ability to send 3D graphics and video. What virtualization is missing is the ability to do something as wild as play multiple 3D games rendered on a GPGPU resource pool and send it to the clients.. Anyway, there are multiple companies working on the 3D over RDP issue, but it is in its infancy, this includes NEC, HP RGA, Citrix, and the 3rd party OpenGL RDP extenders. Part of my proposal is to develop a Mpeg4 H.264 / RDP Hybrid protocoll, where the host computer (home server, business server, etc) has a multicore H.264 real time hardware encoder (capable of 1920x1080p minimum) .. and the clients, of course, have a multicore H.264 decoder.

There is no reason (or there shouldn't be) for me (no family or roomates) should have 4 computers in my house. I desire to build a very scalable home server that has virtualized machines (PS3, Xbox, PCs, camera IP DVR, home automation contoller, DishNetwork Virtual receiver, Tivo Virtual Machine, etc), and is so simply scablable I can stack another unit, tie them together with (10GbE or PCIe/PCIe direct bridge), and have them instantly act as a single machine. unfortunately as vast as my idea is, I feel as though I'd need several big industry players to cooperate to make it happen (and still be a enthusiast level product).

Oops, I didn't mean to totally elaborate on the home server piece.


* Memory is dirt cheap now, true, and it's quite trivial to get 8-32 GB of
fast RAM if you want it; I'm already at the 32GB mark and climbing for just
some of the reasons you mentioned wrt. performance and memory / data hungry
computing.


Yes, memory is dirt cheap - ability to upgrade even enthusiast machines is stuck in the 4-8GB mark. Hence my idea for the memory expansion card with a multivendor standard. I remember reading an article where 3D artists, using Lightwave3D were rendering space battles for BSG, SG1, and SGA. On their first visit to 64bit land, which enabled them to jump from 4GB on their rendering boxes to 32GB (and of course 64bit calculations) and it dropped the render time on the given space battle scene from something like 6 hours to 1.2 hours. Kind of a nice boost eh.

* Clearly digital video, DVR uses,.....from some Intel guy about terabyte data stores being needed and "coming soon" for the average home to satisfy these consumer media needs.
I just dropped a 750GB USB drive on my Dishnetwork 722 DVR and it is filling up fast. Assuming Dish's interface improves for sorting thru media, I would easily store several TB a year. However I don't like the proprietary nature of storing all my stuff under DN's umbrella. I don't understand why shows without a "Do Not Copy" flag can't be copied to my PC. Home Director has a "Digital Life" media server that is a nice mix of a dedicated DVR with full compatibility with the PC. They advertise 300GB add-ons, but their system is somewhat unlimited in storage add on.

* Problem is even though RAM is DIRT cheap, whereas memory makers are lately selling onuncomfortably thin margins due to a glut in the market and they WANT RAM prices to climb,

Hmmm. Gee.. you'd think RAM makers would come up with a cheap (say $150) PCIe ram expansion board, develop some open source software for it, and push all those enthusiast markets to "drop an extra 16GB in your system".. Since RAM prices aren't going to go up (much), sounds like they need to sell more ;-)

* So given the 83:1 cost difference between disk and RAM, RAM "cacheing" of the
disk or replacement for the disk becomes RELATIVELY unaffordable and useless.
What good is a "cache" that's like 1% of the size that you would use in less than
2 hours -- chances are high the data you WANT won't be in RAM!


Exactly.. which is why I am asking for intelligent cache software, and the ability to add massive amounts of whatever the cheapest ram is. For Promise Technology to tell me that "only EMC/IBM/etc will ever do such levels of caching" probably ranks up at the same level of the forward looking statement BG made "No one will need more than 640k" Sorry Bill, call me ok? I need to run my Panacea computing architecture by you...


*I.e. 1 DVD = 4.7GB = $100 of cheap RAM, or $7 worth of DISK.
Browse through your PC-virtual-TIVO / movie-on-disk collection and suddenly
..... Basically you need "real time" random access to lots more data than fits
in any affordable amount of RAM, even given the stupendous capacity of RAM per
the dollar.


AMEN brother. But there is at least a performance bonus of caching initial chucks of each dataset, indexes, etc. I still don't understand why, with 4GB of ram in my comp, I can click on my media folder (25k files, but only 200 directories) and windows explorer freezes for 15 seconds to (I assume) calculate and show the directory list. sigh.

* In the shorter term, who NEEDs *fast* access to 1-200 gigabytes of data?
Well consumers don't because the applications, OS, etc. aren't there to take advantage


This is the issue that bothers me the most tho. No one writes apps to use large amounts of memory (enterprise apps yes, consumer apps no), so no hardware vendors are building solutions for massive amounts of memory. I think Violin Memory Systems is missing a boat in the consumer/enthusiast market.

Even with NVidia, AMD, and Intel's GPGPU solutions coming, which will scale to multiple external PCIe expansion chasis (I've been suggesting this since 2003!), there is still no industry standard to encourage main stream adoption across HPC and other various operating systems and VM hosts.


* Medium size business? Sure, ok, but any business over the size of about 20 people probably ALREADY has 64-bit "enterprise class" computing OSs, file servers on 64 bit OS, <snip> Yes, it's hideously expensive hardware for the servers and mainframes

Just as the mainframe world brought many of it's goodies to the workstation and x86 business server, I feel it is time commodity high performance desktop boxes to advance a combination of virtualization and clustering to eventually become the mainframe of future home, sm business, and large enterprise. In a couple years, 10GbE (on copper) will be as cheap as 1GbE is today. Hopefully something like this can become a common simplified cluster interconnect.

Give me your opinion. As an IT administrator, would you rather buy a single $50k server, or 10 $5k servers that can each have their own specs, can be swapped out, taken on/offline - and act as a single resource for an advanced Virtual Machine host controller?


* Should *everyone* have better use of their their RAM (e.g. ram disks, disk
caches that don't suck, operating systems that don't page/swap even though you're JUST using a web browser and email application on a computer with MORE than enough RAM, etc.)? SURE!


Don't get me started. Oh wait, I already griped about that. I agree with you completely.. and why the heck can't I at least drop that damn paging file onto a secondary memory device.

*It's one of the main areas Micro$haft claims is improved in
VI$TA, i.e. ReadyBoost, their "smart" prefetch / cacheing stuff, etc. Obviously
it totally sucks, but it sucks a lot less than XP, and they're CATCHING ON, SLOWLY.


Heh. It is nice to suck less. I was quite dissapointed when Vista arrived and upon review of the prefetch, cache, readyboost I discovered it wasn't so hot

* Anyone CAN make a SSD or RAM disk software or whatever for Windows, it's
been done, and anyone can make a SAN drive / network attached storage unit,
etc. that lives outside the OS and has nice cacheing, fast response, intelligent
indexing, auto-defrag, auto-backup, security, whatever.


Yea.. There are several System on a Chip (SoC) designs for these devices. Dlink even has a low end iSCSI server. What is it missing? The ability for a mass of programmable cache. I don't understand how a $300 server motherboard can have 64GB worth of memory capability, yet its impossible to build a iSCSI or Etherdrive box that would allow me to serve my comps with a insanely fast bit of storage...

* We're in a world where you can't even SYNC your PDA-phone nicely or sync files on a thumb drive nicely or even have a half way decent photo-album application.

Interoperability only happens as a final move if something has caught on from another vendor or technology. For example, I spent a great deal of time converting several hundred gigabytes of AVI video from a digital video camera to .MP4 (the AVC or H.264 version).. I sent it out to family and low and behold, no one's Windows Media Player would recognize the file.. Quicktime, Nero Media, and other players thought the .mp4 extention was just great. But M$ wants us playing windows media extensions, so this is their harassing way of nudging people into their standards.. It took me a week to finally find a set of codecs that made everything work. What the hell is the reason for this?

* Gamers who are savy about laggy performance already know to buy 2GB, 4GB
or whatever lets their "load times" be basically zero because it's all in disk cache
after the first use.


Too bad so many aren't savy. Now that fat code reuse has become the standard, and game companies aren't optimizing code modules they've ported from their last games (EA !) the games are getting worse about data load and ram usage. In a 4v4 on LOTR BFME2, with medium settings, the game will take up to 1.2GB on its own. So many players in our games say "Dude, I've got a gig of ram - nothing will use that up" I just had a debate with a friend who insisted that if I bought a quad-core processor I'd be wasting my time because "No game will use all 4 cores for another 2+ years" .. I had to explain to him that after you boot your OS, load all the chat clients, OneCare, etc, you don't have have a "game.exe process and a single WindowsOS.exe process." heh.. sigh.. <frustrated head pounding on desk>

* What would make the need for more RAM cache a lot higher? Multi-user systems; oops; beyond the "home computing net" serving the family of 4, most people don't need it. Bigger oops, Microsoft is ADAMANTLY opposed to "multi-headed" PCs i.e. hardware *OR* SOFTWARE that lets more than one person get ANY use of a PC at once;

Bingo.. and (I have to confirm this), I'm assuming the Win Home Server doesn't allow multiple terminal sessions does it?

* They WANT you to buy 4 copies of the OS, 4 different PCs, 4 different monitors, keyboards, mice, 4 times the hard disk space you need, etc. because they get paid that way.

Hmm, I guess I'm not getting MS behind my scalable central home server architecture idea.. Your conspiracy theory just ruined my morning, SIR.

So while I think your ideas are basically correct in identifying OPPORTUNITIES
to improve the DESIGN and PERFORMANCE of storage/computing systems
via engineering of the sub-components and low level software systems, I don't think you or I or anyone has the PERSUASIVE CONTROL over the dominant
ROADBLOCKS to garning these seemingly "low hanging fruit" benefits.


I just peeled off the storage and secondary memory pool idea from my document, I guess this means the rest of it will take more work too.. damn it.
I agree with the rest of your reply. That was some great input.
MVR is offline   Reply With Quote
Old 11-13-2007, 11:36 AM   #4
Peter
Elite Member
 
Join Date: Oct 1999
Posts: 9,640
Default Advanced Disk caching technology built into MB BIOS

When you're down on the mainboard already, there's very little point in adding a lot of dedicated RAM for disk caching - just add more system RAM and do it there.
Peter is offline   Reply With Quote
Old 11-13-2007, 08:14 PM   #5
Cerb
Elite Member
 
Cerb's Avatar
 
Join Date: Aug 2000
Posts: 14,648
Default Advanced Disk caching technology built into MB BIOS

Alas, there is one solution to this that would totally rock: Registered (but not ECC) RAM support for common desktop CPUs and chipsets. Done. Everybody's happy. Market Registered+ECC for servers, so that using it on the desktop leaves ECC unused, and you can double, or maybe even quadruple, your system RAM for the number of slots. Meanwhile, unbuffered RAM stays cheaper, just as it is now.

But, as QuixoticOne pointed out...there's just not the incentive to do so, when it could mean encouraging technologies and PC uses that could mean less $ in some cases.

I do little enough I'll be fine with 4GB for now...but I want to be sure 8GB is an option. But then, I'll need a whole new system, and probably have to wait 2+ years, to get 16-32GB support outside of using server parts.

It's really annoying given that over the last platform I've used, I went from 256MB to1.5GB (SDR to DDR, but my first Duron board supported 1.5GB, all after it 3GB)...but can't make near that kind of jump again this time around.

So, meh, 8GB as a limit will just have to do...though, sadly have to do. I'm sure by the time I can justify 8GB I'll be able to use far, far more.
__________________
"The computer can't tell you the emotional story. It can give you the exact mathematical design, but what's missing is the eyebrows." - Frank Zappa
Cerb is offline   Reply With Quote
Old 11-15-2007, 03:40 AM   #6
QuixoticOne
Golden Member
 
Join Date: Nov 2005
Posts: 1,855
Default Advanced Disk caching technology built into MB BIOS

I found this that I thought you'd be interested to see.

http://www.extremetech.com/art.../0,1697,2216973,00.asp

Good response above; I'll respond in more detail shortly.

QuixoticOne is offline   Reply With Quote
Old 11-18-2007, 05:31 PM   #7
MVR
Junior Member
 
Join Date: Sep 2007
Posts: 17
Default Advanced Disk caching technology built into MB BIOS

Quote:
Originally posted by: Peter
When you're down on the mainboard already, there's very little point in adding a lot of dedicated RAM for disk caching - just add more system RAM and do it there.
You didn't really address the issue that even an enthusiast class motherboard is limited to 8-16GB ram. Thus it may be required to have a secondary memory expansion system.

Quote:
Originally posted by: Cerb
So, meh, 8GB as a limit will just have to do...though, sadly have to do. I'm sure by the time I can justify 8GB I'll be able to use far, far more.
Unfortunately true. I wish the motherboard makers (or I guess chipset makers) would keep on top of, or even have support for future, ram module sizes. i.e. 4GB modules are now actually available, some 8GB modules now. Yet most only support 2GB. 8 slots supporting 4GB/ea begins to get us where we need for an intelligent cache system. That being said, I still want a solution as I outlined before that uses the cheapest memory modules. $1000 for 24GB is a number that I believe a large number of enthusiasts would jump on. You could fit all your OS files and some core program files in 10GB-15GB could you not?

Quote:
Originally posted by: QuixoticOne
I found this that I thought you'd be interested to see. Link to Sandisk's Vault
This had me excited for a moment. I thought it was a ram based disk cache, but unfortunately it is still based on Sandisk's rather poorly performing flash memory. Still not giving a convincing enough arguement over a pair of high end RAID 0 configured drives. Assuming flash memory can double or quadruple its performance in the coming years, this is a step in the right direction for sure.


MVR is offline   Reply With Quote
Old 11-19-2007, 12:14 PM   #8
tcsenter
Lifer
 
tcsenter's Avatar
 
Join Date: Sep 2001
Location: Central California
Posts: 16,904
Default Advanced Disk caching technology built into MB BIOS

Quote:
I've had an idea running through my head for a couple years now.
Cool, that's about 1/6 as long as variations of your idea and its primary objective have been the subject of interest and investigation by numerous others. This idea dates back to the early 1990s (if not longer). Several companies have actually tried to market RAM disks, solid state disks, caching and prefetching software that uses system RAM, dedicated RAM device, hard drives, or a combination of them.

Millions upon millions of dollars have been spent on related R&D efforts by storage technology vendors, core logic vendors, technology IP firms, ASIC vendors, operating system developers, software developers, industry working and standards groups (e.g. ATA/T13, SATA-IO, PCI, et. al.), and university-sponsored projects.

The fruits of such efforts, like caching, prefetching, and read-ahead functions (just to name but a few), have found their way into every operating system since DOS, storage drivers, controller ASICs and firmware, storage hardware, I/O interfaces...pretty much where ever it is productive to transfer data faster rather than slower, more data rather than less, with greater efficiency or lesser costs/barriers. What is the RAM-based write cache found on high-end RAID controllers if not but one particular application of this?

I don't want to poop on your parade, but you ain't the only one who is or has been exploring this. Attend a few computer-centric industry trade shows. You will likely encounter some small company with a very similar product in hand, hoping to hear those words that give every budding entrepreneur a case of the happy pants - "Here's my business card." And that product will almost certainly be marketed for enterprise segments with an enterprise price tag.

Quote:
I've approached various SATA raid card manufacturers with it and they've all replied that they just don't see a market for it. That or they've told me that only the enterprise market (EMC, IBM Storage, etc) will ever feature such technology.
They are sorta right, but I don't think they really meant to say that other segments of the market will 'never feature such technology'. Instead, I think they meant something more like, 'We looked at this last year, and again two years before that. We did not find enough potential for other segments to justify it. Check back in a couple years.'

There is a reason the market for this has been limited to segments that have many many thousands of dollars to spend on I/O boosting goodies. The high-end enthusiast market that is willing to spend $1000 on a graphics card, processor, or I/O boosting device is very far from "commodity" level hardware. Not quite as far as $50,000 rackmount RAM appliances, but far enough.
tcsenter is offline   Reply With Quote
Old 11-19-2007, 04:17 PM   #9
MVR
Junior Member
 
Join Date: Sep 2007
Posts: 17
Default Advanced Disk caching technology built into MB BIOS

Quote:
Originally posted by: tcsenter
What is the RAM-based write cache found on high-end RAID controllers if not but one particular application of this?


Well, to start, it is an extremely limited application of this. Show me a consumer RAID controller that implements what I mentioned previously.

I don't want to poop on your parade, but you ain't the only one who is or has been exploring this.
I don't want to poop on your efforts to thoroughly show me my idea isn't new, but perhaps it would help make your point better if you understood that not only am I not trying to re-invent a wheel, I am not trying to grab some credit for proposing a minor twist on what people are already working on. I only want to see solutions come to the enthusiast market, I'm not trying to invent the product.

And that product will almost certainly be marketed for enterprise segments with an enterprise price tag.
Ahem... and I was hoping some motherboard or chipset vendor would take the leap to bring it to we enthusiast nerds. I was merely suggesting a few methods I felt were practical.

There is a reason the market for this has been limited to segments that have many many thousands of dollars to spend on I/O boosting goodies. The high-end enthusiast market that is willing to spend $1000 on a graphics card, processor, or I/O boosting device is very far from "commodity" level hardware. Not quite as far as $50,000 rackmount RAM appliances, but far enough.
Well, as you said, give it a few years. I'm sorry you seem so passionate about showing this idea is impractical vs contributing a constructive idea. Since everyone is focusing on current SSD technology, which to me has miles to go, I thought it would be interesting to bring back up the "old" idea of caching since various pieces of the industry seem to be falling into place that make it practical.

I just came across this site: Fusionio.com this morning, but haven't had time to review. They do say they are aiming for $30/GB and have the speeds that sound more attractive to me. I'll read up when I have some time.
MVR is offline   Reply With Quote
Old 11-21-2007, 10:07 PM   #10
tcsenter
Lifer
 
tcsenter's Avatar
 
Join Date: Sep 2001
Location: Central California
Posts: 16,904
Default Advanced Disk caching technology built into MB BIOS

Quote:
Originally posted by: MVR
Well, to start, it is an extremely limited application of this.
Err...an extremely limited application is the same thing as one particular application. You didn't add anything there, just changed the wording and implied it meant something different.

Quote:
Show me a consumer RAID controller that implements what I mentioned previously.
Why would I show you something that doesn't exist? The fact it does not exist is precisely the point.

RAID/SAS controllers don't have an integrated DRAM controller capable of supporting the large memory capacities you are targeting. They require separate silicon such as Intel's IOP for the DRAM controller, and Intel's IOP family only supports 2GB max. This is why even multi-thousand dollar RAID/SAS controllers don't support more than 2GB max per controller.

Alternative IOP designs such as AMCC's PowerPC440 support up to 16GB max. Xilinx's Spartan and Virtex FPGA cores support up to 8GB max. None of which support more than four chip-select banks (four DIMMs), nor are the MCU/MMU robust enough to handle 16GB ~ 32GB with much efficiency.

Its not as easy as slapping eight DIMM slots onto a RAID controller card.
tcsenter is offline   Reply With Quote
Old 11-25-2007, 07:39 AM   #11
MVR
Junior Member
 
Join Date: Sep 2007
Posts: 17
Default Advanced Disk caching technology built into MB BIOS

Quote:
Originally posted by: tcsenter
Why would I show you something that doesn't exist? The fact it does not exist is precisely the point.
Sigh. I'll skip the explanation.

RAID/SAS controllers don't have an integrated DRAM controller capable of supporting the large memory capacities you are targeting. They require separate silicon such as Intel's IOP for the DRAM controller, and Intel's IOP family only supports 2GB max.
Good info, you've actually contributed to the thread ;-) heh. If I'm reading this correctly, Intel's IOP315 supported 12GB. But the IOP348 only has 2GB as you stated. It is also unfortunately limited to ECC memory.

nor are the MCU/MMU robust enough to handle 16GB ~ 32GB with much efficiency. Its not as easy as slapping eight DIMM slots onto a RAID controller card.

Bummer. Well thank you for your input. I guess I'll just have to wait and see when it will happen.
MVR is offline   Reply With Quote
Old 11-26-2007, 12:17 AM   #12
VirtualLarry
Lifer
 
VirtualLarry's Avatar
 
Join Date: Aug 2001
Posts: 24,481
Default Advanced Disk caching technology built into MB BIOS

Quote:
Originally posted by: Cerb
Alas, there is one solution to this that would totally rock: Registered (but not ECC) RAM support for common desktop CPUs and chipsets. Done. Everybody's happy. Market Registered+ECC for servers, so that using it on the desktop leaves ECC unused, and you can double, or maybe even quadruple, your system RAM for the number of slots. Meanwhile, unbuffered RAM stays cheaper, just as it is now.
The truth is, with the massive amounts of RAM on today's consumer motherboards, there is no reason why EVERYONE shouldn't already be running ECC RAM today.

Bit errors happen. It's a statistical likelyhood. Wouldn't you prefer that they be detectable and correctable?

Given the low cost of memory, it would cost almost nothing for some additional chips for each module.
__________________
Rig(s) not listed, because I change computers, like some people change their socks.
ATX is for poor people. And 'gamers.' - phucheneh
haswell is bulldozer... - aigomorla
"DON'T BUY INTEL, they will send secret signals down the internet, which
will considerably slow down your computer". - SOFTengCOMPelec
VirtualLarry is online now   Reply With Quote
Old 11-26-2007, 02:31 AM   #13
MVR
Junior Member
 
Join Date: Sep 2007
Posts: 17
Default Advanced Disk caching technology built into MB BIOS

Quote:
Originally posted by: VirtualLarry
The truth is, with the massive amounts of RAM on today's consumer motherboards, there is no reason why EVERYONE shouldn't already be running ECC RAM today. Bit errors happen. It's a statistical likelyhood. Wouldn't you prefer that they be detectable and correctable? Given the low cost of memory, it would cost almost nothing for some additional chips for each module.
Well I just learned something. As far as I understood FB-DIMMs were to become the next big thing. Now it looks like Registered ECC DDR3is the hip thing all the kids at school are excited about.

To address your note, I never knew I was having any issues with data corruption using standard DDR2 memory.. The thing I don't get tho, is why after all these years, ECC is priced so much higher than non-ECC.

As of day (11.26.2007) quality brand sticks are
PC2-6400 DDR2-800 1GB ....................... $35
PC2-6400 DDR2-800 1GB - ECC...............$50
PC2-5300 DDR2-667 1GB - ECC FBDIMM.. $67




MVR is offline   Reply With Quote
Old 11-26-2007, 09:03 PM   #14
tcsenter
Lifer
 
tcsenter's Avatar
 
Join Date: Sep 2001
Location: Central California
Posts: 16,904
Default Advanced Disk caching technology built into MB BIOS

Quote:
Originally posted by: MVR
To address your note, I never knew I was having any issues with data corruption using standard DDR2 memory.. The thing I don't get tho, is why ECC memory is why after all these years, ECC is priced so much higher than non-ECC.
ECC brings more complexity on the PCB, adds one more chip per rank (typically two more chips per module), and has lower production volumes than non-ECC due to lower demand. The market for ECC tends to be more discriminating and less influenced by price alone. In response, companies tend to exercise better revision control on server product lines.
tcsenter is offline   Reply With Quote
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -5. The time now is 08:23 AM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2014, vBulletin Solutions, Inc.