2 live buffers at the same time. How?

[MiamiMan]

Surely tinkerers have already figured this out, but I have two basic questions about the S3 that's been bugging me for a long time. No doubt this has been asked before so please excuse me for bringing it up again, but I can't find any information on it.

1) I was under the impression (incorrectly it seems) that the S3 uses some kind of RAM memory for the two 30 minute buffers, but I have been told that, instead, the S3 records the buffers live, directly to the hard drive.

I'm trying to understand how it can do this, especially considering that the S3 can buffer two different shows at once and the user can watch a recorded show, all at the same time. Without using RAM, it would seem that the S3 would have to have 3 separate read/write heads (two recording the buffers and one reading the prerecorded show) on three different drive platters! I doubt that the S3 uses RAID drives, so if it doesn't use multiple platters, or RAM, how in the world does it do this trick?

2) Even though the S3 is slightly faster than the S2, and I love it regardless, it is still slow (IMO). Is this because of an inexpensive CPU? What CPU does it use and what is its clock speed?

[SeanC]

1. It's well within standard hard drive technology for the S3 to write 2 streams while reading a third, no trick there.

[MiamiMan]

1. It's well within standard hard drive technology for the S3 to write 2 streams while reading a third, no trick there.

Well that doesn't explain anything. :confused:

Of course standard hard drive technology with RAM can do it, but how can it do it without RAM? It would seem the drive head would have to be in thee locations at once, otherwise the time taken to move from one track to another would result in loss of data. With RAM, however, the data can be continuously stored while the read/write head reads and writes in non-continuous fashion (as long as the hard drive is fast enough.)

[Brad Eleven]

Of course RAM is required. Just not the system RAM.

There are multiple buffers on the drive itself for I/O. Maybe they're not RAM, per se. They're certainly faster than access to/from the drive.

[russwong]

Drives today can do even more. On my HTPC, I'm able to record 3 streams, watch locally AND stream to an xbox360 all at the same time. That's technically 5 streams. You know, my pioneer plasma is capable of streaming from my HTPC too, maybe I'll try and add that to the loop and see if there is any problems with 6 streams....

[abadan]

The only thing wrong with Tivo buffer is that it is only for 30 minutes. Please increase it to at least an hour.

[phox_mulder]

Of course RAM is required. Just not the system RAM.

There are multiple buffers on the drive itself for I/O. Maybe they're not RAM, per se. They're certainly faster than access to/from the drive.

The MPEG encoders and decoders have to have some kind of buffer as well.


phox

[pauldy]

At their highest bit rates HD content is less than 3MB/s. Some of the "tricks" that have been used to make the drive behave more consistent and less prone to stuttering for random reads and writes is to use very large block sizes. Like 1MB or more per block, most people using NTFS have a maximum block size of 32k(maybe a bit higher). So what is probably happening is some of the data is buffered into system ram using some type of direct memory access the data is then written to the drive in large blocks and read from the drive into a memory buffer in large blocks. Doing this puts more of the stress on moving the data around than the random access of the disk to maintain these multiple data streams. I would imagine the same type of DMA is used to pump the data to a hardware video decoder as well pushing the total internal bandwidth for 2 writes and 1 read to around 18MB/s. That takes two streams from the decoder to the drive and one from the drive to the decoder. So based off the block size there may between 12 and 36MB of RAM being used for the live buffers and up to 8.5GB of disk space being used for the live buffer at 30 minutes per buffer.

So system RAM is used for the live stream but only as a buffer on its way to the disk.

[vman41]

If you put you head real close to the TiVo, you can usually hear the drive head moving several per second as hops between the different streams it is reading and writing (plus occasional I/O to OS and swap partions).

[MiamiMan]

So what is probably happening is some of the data is buffered into system ram using some type of direct memory access the data is then written to the drive in large blocks and read from the drive into a memory buffer in large blocks. Yes, that makes sense. My "expert" informants, however, say that it buffers data onto the hard drive directly, which doesn't make sense, unless they didn't explain it in detail.

Isn't it strange that the Tivo can handle video data so quickly yet it takes forever to move from one menu line to another. Even my old Apple II was able to scroll from one line to another much faster than my S3, and it was written in Apple BASIC! :o

[pauldy]

Yes, that makes sense. My "expert" informants, however, say that it buffers data onto the hard drive directly, which doesn't make sense, unless they didn't explain it in detail.

Isn't it strange that the Tivo can handle video data so quickly yet it takes forever to move from one menu line to another. Even my old Apple II was able to scroll from one line to another much faster than my S3, and it was written in Apple BASIC! :o

Which also makes sense as data coming for the ui might be coming in really small blocks off one of the ext partitions and it is competing with the system bus. Because it is true random access, meaning you can't predict what will be requested next as efficiently as you can with a video stream. Even if each request only takes an extra 30ms best case and you have 20 small requests then you are looking at 600ms delay in just getting the data which may only be 640kb of data vs 20MB of video moved. This is very noticeable and I'm sure future software updates can address this by using various caching techniques and queuing the requests.

[Royster]

Yes, that makes sense. My "expert" informants, however, say that it buffers data onto the hard drive directly, which doesn't make sense, unless they didn't explain it in detail.

I think we're confusing two duifferent uses of the word "buffer". Yes, the TiVo uses in-memory buffers to hold data coming in from the tuners. The operating system of the box (Linux) coordinates writing the contents of those buffers onto the disk.

In addition, the TiVo interface has a live "buffer" of up to the last 30 minutes. This buffer is not the same as the buffer used to temporarily hold data before writing it to disk.

Isn't it strange that the Tivo can handle video data so quickly yet it takes forever to move from one menu line to another. Even my old Apple II was able to scroll from one line to another much faster than my S3, and it was written in Apple BASIC! :o

It's all about priorities. The system is designed to make sure that data moves from the tuner to memory to the disk and from the disk to memory the screen in a reliable fashion. This is the first priority. Moving the menu bar on an on screen UI is a much lower priority task. Yes, the system could have be designed so that the screen response is always snappy, but you would be peeved if it occasionally dropped a second of video or audio track.

[MiamiMan]

It's all about priorities. The system is designed to make sure that data moves from the tuner to memory to the disk and from the disk to memory the screen in a reliable fashion. This is the first priority. Moving the menu bar on an on screen UI is a much lower priority task. Yes, the system could have be designed so that the screen response is always snappy, but you would be peeved if it occasionally dropped a second of video or audio track. No doubt it has priorities but that can't (or shoudn't) explain menu slowness, unless it's really poor programming (which of course, could very well be the problem). Consider that while in the menus, you're not watching a prerecorded program. If priorities had anything to do with it, then take the prerecorded priority and apply it to the menu display.

Moreover, put an inexpensive CPU just to handle the menus. Many computers today utilize several CPUs for just these kind of things.

[CrispyCritter]

No doubt it has priorities but that can't (or shoudn't) explain menu slowness, unless it's really poor programming (which of course, could very well be the problem). Consider that while in the menus, you're not watching a prerecorded program. If priorities had anything to do with it, then take the prerecorded priority and apply it to the menu display.

Moreover, put an inexpensive CPU just to handle the menus. Many computers today utilize several CPUs for just these kind of things.It's not a question of CPU speed at all;any slow CPU can handle it. It's purely a question of disk access priority. The data on most menus cannot be accessed without disk accesses. There's too much data to hold in memory. The disk accesses for menu items has to be interspersed with the disk accesses for shows, which you certainly want to take priority. Perhaps TiVo can tune things a bit better now, but remember, once the S3 gets full MRV, it will be reading/writing 5 disk streams at once at times, with other networking activity (TiVoCast?) possibly also affecting the disks. TiVo has to be sure the UI does not access a lot of data at once and thus lock out these other disk activities, some of which can't wait.

[MiamiMan]

It's not a question of CPU speed at all;any slow CPU can handle it. It's purely a question of disk access priority. The data on most menus cannot be accessed without disk accesses. There's too much data to hold in memory. Well then there's not enough memory! Moreover, if it can handle 5 streams at once, then there's no excuse to cripple the menu when its only doing two streams.

[CrispyCritter]

Well then there's not enough memory! Moreover, if it can handle 5 streams at once, then there's no excuse to cripple the menu when its only doing two streams.More memory won't help much (well, several GBytes would :) ), and you're wrong about wanting to worry about the effect of streams. You're making absolute pronouncements about things you don't know enough about!

[MiamiMan]

More memory won't help much (well, several GBytes would :) ), and you're wrong about wanting to worry about the effect of streams. You're making absolute pronouncements about things you don't know enough about! I'm not worrying about streams; I'm only responding to the "information" you've giving. The priority due to streaming to a hard drive doesn't explain the slowness of the menus. It doesn't make sense.

Again, the S3 is able to record at least two programs and read a prerecorded show at the same time. But while in the menus, it's NOT streaming a prerecorded show. The saved time that was used during streaming a prerecorded show could be spent for disk access for the menus (if your priority for disk access theory holds). Do you see what I mean? It wouldn't account for the slow menus unless its just poor programming.

[ashu]

Sheesh, another 'diss TiVo' thread under the pretext of intelligent (and very weak at that) inquisitiveness. Give it up!

[CrispyCritter]

I'm not worrying about streams; I'm only responding to the "information" you've giving. The priority due to streaming to a hard drive doesn't explain the slowness of the menus. It doesn't make sense.

Again, the S3 is able to record at least two programs and read a prerecorded show at the same time. But while in the menus, it's NOT streaming a prerecorded show. The saved time that was used during streaming a prerecorded show could be spent for disk access for the menus (if your priority for disk access theory holds). Do you see what I mean? It wouldn't account for the slow menus unless its just poor programming.I learned long ago that I can't teach someone who insists they know what the answer is while knowing nothing about the subject. I'm dropping this: you've been absolutely wrong several times in this thread so far, and I see no point in continuing.

[Brad Eleven]

I think I understand your question/gripe. If you're talking about the delay in bringing up menus (can't remember the text in the pop-up, is it "Please Wait..."?), then it has nothing to do with the CPU or memory.

You did mention programming, and that's it.

Heard of the Cache Card? 9th Tee makes it, I think? It's RAM that you can plug into the back of the unit, hack the unit to store the directory information for Now Playing and other lists that might need refreshing. All TiVos fetch this from disk, so it can be really slow while recording one or more streams. I think the 30-second buffer impacts this, too.

That is, the fetch of whichever list is needed has to queue up along with other drive tasks.

Whether or not I guessed what you're getting at, I totally understand how this discussion got heated up, especially if anyone involved was watching his/her TiVo and had to look at "Please Wait...", or "Just a Second...", or "Our operators are busy serving other customers..."

[MiamiMan]

Sheesh, another 'diss TiVo' thread under the pretext of intelligent (and very weak at that) inquisitiveness. Give it up! If you're talking about me, then you have it wrong. I'm not dissing Tivo. I LOVE Tivo. I have an S1, and S2, and an S3. However, I am curious about some of its features and I am honestly trying to find answers in order to make suggestions to Tivo. I don't think I am the only loyal Tivo fan who has disappointments about the speed of the S3's menus.

I think a lot of Tivo users would spend the dollars for a faster machine. I'm asking questions and some of the responders have given good answers about the buffers but so far, I don't understand the technical reasons why the S3 has slow menus. CrispyCritter's response didn't make sense, and I said so. I'm not attacking him, only the theory.

But so far, I'm being accused of making absolute pronouncements about things I don't know enough about (I don't even believe in absolutes), and for pretending a pretext of intelligence, and a weak intelligence at that. Well perhaps I'm not intelligent compared to you all, but answering me by making false accusations about me doesn't answer the question either.

My question remains.

[MiamiMan]

I learned long ago that I can't teach someone who insists they know what the answer is while knowing nothing about the subject. I'm dropping this: you've been absolutely wrong several times in this thread so far, and I see no point in continuing. I don't know the answer and I although I made a few guesses (poor programming, etc.), I don't claim to know. That's why I'm asking questions. But your response did not answer the question either. Your answer doesn't make sense, and your personal attack does nothing here but produce anger.

I have also learned something a long time ago. If a person can't explain it, then he or she probably doesn't know.

Brad Eleven: Thanks for understanding. Indeed, poor programming gives a possible explanation. Perhaps it's something else, but at least poor programming makes sense.

[haysdb]

It's definitely a programming issue. Not "bad code" per se, but something that can and will be optimized in a subsequent release of the firmware. As a programmer you worry about getting it to WORK first, and worry about making it FAST later. Aren't we still on the first release of the Tivo software?

BTW, I'm finding the menus to be quite snappy. Compared to the Moxi they run at light speed.

[ashu]

I am honestly trying to find answers in order to make suggestions to Tivo.

Rrriiggghhhhttt.

Good luck with the "Dear TiVo, your programmers suck. here are a few demandssuggestions ... "-Approach

It's easy to armchair-software-design with negligible understanding of the issues.

:rolleyes:

[ashu]

Brad, you added a piece of hardware. That costs mucho dinero.
Miami Man insists everyone would pay extra and desires greated snappiness.

I argue that that is a ridiculous presumption, and I find the menus just as fast as I need them to be. It's not a Seti@Home number cruncher - it is ONLY a DVR.

I'm not arguing that it could be improved upon - but only by throwing hardware (thatc ostc money) at the solution.

At a certain level, I'm sure some design decisions have been made to not overly fork/separate the code trees for S2 & S3 units, and it isn't unlikely that the S3 hardware is not necessarily being utilized to its fullest. I'd give them some time before repeatedly, ascerbically attacking their programming ability and pretending to understand the issue better than you actually do!

Peace out.
Is that still used, by anyone, in this millenium? :)

[MiamiMan]

It's definitely a programming issue. Not "bad code" per se, but something that can and will be optimized in a subsequent release of the firmware. As a programmer you worry about getting it to WORK first, and worry about making it FAST later. Aren't we still on the first release of the Tivo software?

BTW, I'm finding the menus to be quite snappy. Compared to the Moxi they run at light speed. That would be great if they fixed this in the next release! Do you have any information for the next firmware release?



ashu: Your personal attacks are way out of line here. Why don't you take your own advice and "Give it up"?

[haysdb]

1) I was under the impression (incorrectly it seems) that the S3 uses some kind of RAM memory for the two 30 minute buffers, but I have been told that, instead, the S3 records the buffers live, directly to the hard drive.All video, including the 30 minute circular buffer, is stored on disk. To store that much video in memory would require 4 to 5 GB of free RAM.

I'm trying to understand how it can do this, especially considering that the S3 can buffer two different shows at once and the user can watch a recorded show, all at the same time. Without using RAM, it would seem that the S3 would have to have 3 separate read/write heads (two recording the buffers and one reading the prerecorded show) on three different drive platters! I doubt that the S3 uses RAID drives, so if it doesn't use multiple platters, or RAM, how in the world does it do this trick?Simply by doing things extremely fast! A cpu can run exactly ONE set of instructions at a time. A hard drive can read or write only one block of data at a time. Computers give the impression of doing many things at once by allocating very small slices of time to each task. In the space of a second, a modern CPU can execute many MILLIONS of instructions, and read and write many MANY megabytes of data.

At 10 GB/hr (?), the hard drive must read/write "only" 3 MB/sec. The SATA hard drive interface used in the S3 is theoretically capable of 150MB/sec transfer rates, or FIFTY TIMES more than is required by a single video stream. Even 6 streams will not begin to task a single hard drive.

2) Even though the S3 is slightly faster than the S2, and I love it regardless, it is still slow (IMO). Is this because of an inexpensive CPU? What CPU does it use and what is its clock speed?There are multiple processors in the S3. How many I can't tell you. Only one "CPU", but many MANY processors imbedded in encoders, decoders and such. The cpu isn't necessarily called on to do that much, because dedicated/embedded processors are doing most of the heavy lifting. I wouldn't blame the cpu for slow menu's. It's partly a disk I/O issue, and partly an issue of code that just needs to be optimized to take better advantage of the available hardware.

[MiamiMan]

All video, including the 30 minute circular buffer, is stored on disk. To store that much video in memory would require in the neighborhood of 4 GB of free RAM. . . . Yes I understand this now if it uses some RAM. What I couldn't understand was doing it without any RAM.

Needing 4 GB for a complete RAM buffer makes sense (250 GB hard drive, 32 hours of HDTV programing), and 8 GB for 2 buffers. If it had this much RAM, I suspect that this would allow the S3 to spin down the hard drive during Standby. How about a flash disk with 8 Gigs? <hee-hee>. I wish Tivo would build a DVR with expandability just like computer manufacturers do.

Thanks.

[ashu]

Some incredible & ridiculous pipe dreams there - 8GB RAMdisk? More memory? More CPU?

Well, how about more $$$ for a custom one-off solution :p or buy one of those 4000+ Euro Balupunkt or Pioneer units that CAN record 4-6 HD streams (expandable) etc. That (and likely, YOU) is not the market TiVo is aimin to completely satisfy. Given the constraints of the hardware price/budget they've kept the S3 under (compare to greedy 'ole Apple and their $600 iPhone - that's ridiculously overpriced!), you're really expecting & asking for too much from what is a nearly-universally-affordable (to anyone that wants to enjoy their HD experience, and can afford an HDTV already) HiDef recording solution.

Wait for Moore's law to help ... although by then it is likely 2160p with 3840x2160 will be commonplace tranmission standards, demainding ever more processing & bus/bandwidth horsepower!

I'm sorry that you are sensitive enough to interpret disagreement with your approach and far-fetched expectations as a personal attack :)

[haysdb]

I actually understated the case with the hard drive. Apparently the S3 uses SATA-II hard drives, which have theoretical transfer rates of 300GB/sec, which is 100 times greater than what's needed by a single HD video stream. Boggles the mind, doesn't it?

Yes, some RAM is used. The video data is written to memory first and then written to the hard drive.

I don't think flash memory would work very well because I believe it has limits on how many times it can be written to. Besides, what's the point? Modern drives are very quiet, and can be expected to run continuously for many many years.

[drowe]

ashu: Your personal attacks are way out of line here. Why don't you take your own advice and "Give it up"?
I agree. Let the man ask his question.

[MiamiMan]

I don't think flash memory would work very well because I believe it has limits on how many times it can be written to. Besides, what's the point? Modern drives are very quiet, and can be expected to run continuously for many many years. I was thinking of cost here and not for hard drives but to use in place of RAM, but I suspect they will replace hard drives too in the near future. There's already laptops that use RAND memory instead of disk platters and they claim they are fast and cool. Samsung also now makes a 32 GB solid state disk.

[jautor]

Yes I understand this now if it uses some RAM. What I couldn't understand was doing it without any RAM.

This was actually one of TiVo's primary inventions, and the patented tech. that Echostar was sued for copying... The "Media Switch" chip (ASIC) allowed data to stream between the MPEG encoder/decoder and the hard drive (using DMA techniques) without ever being buffered in main system RAM. This meant that the CPU wasn't actually involved in the data stream, which allowed TiVo to use a much, much cheaper CPU way back in 1999... I don't know if the same technique is used any more in the S3, given the progression of Moore's law, but it wouldn't surprise me. Since, after all, it was, and still is, a very, very cool idea! :)

Jeff

PS - as to expandability - you can do this yourself (see TiVo Upgrade forum), and in the S3, there's an external SATA port there we're hoping will be lit up soon... A search will give you plenty of threads on that topic, though. :rolleyes:

[haysdb]

I suspect they will replace hard drives too in the near future. There's already laptops that use RAND memory instead of disk platters and they claim they are fast and cool. Samsung also now makes a 32 GB solid state disk.32 GB would store less than 4 hours of HD. Flash memory is also prohibitively expensive:Flash memory costs around $30 per gigabyte; the memory needed for the 32GB SSD drive works out to about $960, before any other costs are taken into account.I don't think we'll see the demise of hard drives any time soon.

[MiamiMan]

jautor: Surely the encoders use some kind of RAM, no?

haysdb: 32 GB now, 500 GB tomorrow. As jautor alluded to, the progression of Moore's law could very well produce large GBs and lowered costs in the near future. ;)

[doraemon]

Don't forget that flash drives have fairly lousy write speeds. They excel at reading, though.

No HDD out there saturates the SATA150 interface, let alone the SATA300 one.

[CrispyCritter]

I agree. Let the man ask his question.I have absolutely no problem with his questions. They are good questions. I have problems with his answers: They need a more powerful CPU; they need more CPUs; they need more memory; there is no excuse, it has to be poor programming.

If you have questions, ask questions, don't state controversial answers which you don't really believe (because you have insufficient knowledge to support) in the hopes of getting people to correct you. I choose not to respond to him.

[MiamiMan]

I have absolutely no problem with his questions. They are good questions. I have problems with his answers: They need a more powerful CPU; they need more CPUs; they need more memory; there is no excuse, it has to be poor programming.

If you have questions, ask questions, don't state controversial answers which you don't really believe (because you have insufficient knowledge to support) in the hopes of getting people to correct you. I choose not to respond to him. I never claimed to have THE answers. I do however speculate. It's YOU who are making amazing claims that don't make sense. Examples: "It's not a question of CPU speed at all; any slow CPU can handle it". Oh really, ANY CPU? "It's purely a question of disk access priority." Purely? Nothing else? "There's too much data to hold in memory." Really?

And when I respond to your answers (pretending these are right answers) but showing the inconstancy of your theory, you go ballistic and accuse me of making absolute pronouncements. Wow!

[ashu]

jautor: Surely the encoders use some kind of RAM, no?

haysdb: 32 GB now, 500 GB tomorrow. As jautor alluded to, the progression of Moore's law could very well produce large GBs and lowered costs in the near future. ;)

Are you sure jautor was the first person to allude to Moore's law? :)
Perhaps folks'd feel less insulted if they actually made the effort to read answers posted in response to their questions. :p

No offense meant, at any stage, Miami man - I sensed the same hostility & holier-than-thou attitude that Crispy Critter described (more eloquently than I did), and if it wasn't your intention to sound that way - I apologize for my interpretation :)

[Forcelite]

There are a few good questions here.
First does anyone know the S3 specs hardware wise.

Harddrive = 300 gb (I believe)
Ram= XXX
HD interface = Sata
CPU speed = XXX
Is there a branded Video card = ATI or Nvidia?


Afterall it is a Linux box so there have to be all of the components

Force

[ashu]

250GB, custom video solution by Broadcom (c3 or someone mentioned the precise specs and capabilities - including resolution transcoding - in another ancient thread).

[MiamiMan]

There are a few good questions here.
First does anyone know the S3 specs hardware wise.
I've been wondering the same thing.

After doing further digging on Google, I finally found something on Wikipedia:

The Series 2 Standalones:

CPU:
1xx/2xx-series: NEC 5432 (MIPS), 166 MHz
5xx-series: Broadcom BCM7317[1], 242 MHz
6xx-series: Broadcom BCM7318[2], 266 MHz

RAM:
1xx/2xx/5xx-series: 32 MB of 133 MHz 16-bit DDR
6xx-series: 64 MB of 133 MHz 16-bit DDR

MPEG Encoder
1xx/2xx/5xx-series: Broadcom BCM7040[3]
6xx-series: Two Broadcom BCM7042

The Series 3 uses a Broadcom BCM7038 CPU, with 128 MB RAM for general CPU use- more for the encoders (doesn't say how much)

It didn't cite the MHz for the CPU so I went to Broadcom's site and found this:

"The BCM7038 also incorporates a 300 MHz 64-bit MIPS® CPU, along with a floating-point processor, which has a very fast path to 400 MHz DDR system memory to support the high-performance needs required by advanced applications. The device also includes Broadcom's 3D graphics engine."

And I found this interesting bit of information about PIP:

"features in the BCM7038 offer new advanced capabilities not yet available on HDTV systems, such as support for the delivery of HD broadcasting to multiple-televisions in the home, separate HD programs with picture-in-picture (PIP), and HD program watch and record PVR capabilities."

I would love to have PIP on Tivo.

[jautor]

jautor: Surely the encoders use some kind of RAM, no?

haysdb: 32 GB now, 500 GB tomorrow. As jautor alluded to, the progression of Moore's law could very well produce large GBs and lowered costs in the near future. ;)

They use "buffers" on board the chip, which is RAM, but not the same thing as "system memory", which is what we generally mean when talking about RAM. The difference being that the buffer memory inside a chip is, well, part of the chip, whereas system memory is added cost on top of the price of the chip. If you can creatively build a device that can get by just using the buffers, you've cut out a lot of cost. And in the consumer electronics market, a few dollars is "a LOT of cost". But this is somewhat of a history lesson, as these designs are moving to System-On-Chip type architectures, which will have the CPU and encoders all on one chip, with a single external memory.

And regarding flash memory - I, along with Al Gore, *INVENTED* Moore's law... :D :D :D

But using Moore's law on flash without also applying it to magnetic storage (which in that context I think is actually Shugart's law - quick, check Google!) skews your results. Even if 500GB of flash was, let's say, $100 in 3 years. Neat. But a 4TB hard drive could also be $100... I know which one I'd pick for my Series5. That, or maybe the refurb 1TB Series4 which is free after rebate...

Jeff

[goldfndr]

Consider that while in the menus, you're not watching a prerecorded program. If priorities had anything to do with it, then take the prerecorded priority and apply it to the menu display.
I haven't seen this misinformation corrected, maybe I missed it.

MiamiMan, have you really not noticed the animation in the background of the menus? That's a prerecorded program. If you aren't seeing animation in the background of the menus[1] then your TiVo DVR is defective.

[1]HME apps in particular tend to be designed with static backgrounds, since sending an animated background over the network can be bandwidth intensive.

[MiamiMan]

But using Moore's law on flash without also applying it to magnetic storage (which in that context I think is actually Shugart's law - quick, check Google!) skews your results. Even if 500GB of flash was, let's say, $100 in 3 years. Neat. But a 4TB hard drive could also be $100... I know which one I'd pick for my Series5. That, or maybe the refurb 1TB Series4 which is free after rebate... Jeff

Perhaps platter hard drives may very well fall under the prices of flash, but that doesn't mean most people will be using them. With all the hard drive failures I've had in the past, I'll happily see them go. No doubt, some people will continue to use them, but not me. :) Most people today are using ATA or SATA drives yet there are those who claim that SCSI is superior. Maybe so, but I have no desire to deal with them again either.

I've yet to have a flash drive failure on my cameras, flash drives, etc. Flash drives are not as fast as platter drives (yet) but some are already using flash for DV video. The future's so bright, I gotta wear shades. :cool:

goldfndr: the prerecorded program I was referring to was watching HD program content. Moreover this does not remotely compare to the relatively tiny amount of processing power needed for the looped background animation in the menus. But even here, I'm not complaining about the menus during displays with the background animation; I'm referring to the slow menus after pressing the guide button. There is no background animation there.

I don't have a defective S3 and the same kind of slowness occurs on all the Tivos I've ever owned, including the Philips S1, Toshiba, and my S2. Every time I purchae a new Tivo, I always hope it will be faster, but it never is (actually my S3 menus are a little faster than my S2, but it is still too slow). Now some people don't find this a problem at all, but I do, and many others have the same complaint.

[aaronwt]

Multiple streams is easy for an ATA100 5400 rpm drive. I was doing it in 2001 when I first started recording HD. I could record and also send a couple of streams accross my netwrok simultaneously
I have an HR10-250 with two 5400 rpm drives. The performance is identical to the two HR10-250 boxes I have with 7200rpm drives.
The S3 has an SATA drive and all newer drives are much faster than the drives I had in 2001. Five streams should be easy.

[haysdb]

jautor: Surely the encoders use some kind of RAM, no?Sure.

haysdb: 32 GB now, 500 GB tomorrow. As jautor alluded to, the progression of Moore's law could very well produce large GBs and lowered costs in the near future. ;)The size of these devices has doubled in the past year. At this rate, in just 4 years they will be available in 500 GB capacities, probably for about the same $1000 cost for just the memory as todays 32GB drives (which don't actually exist yet BTW). By that time, 500 GB hard drives will be available for $60, and 2 Terabyte drives will be available for $300. I'm pulling these numbers out of my butt, but think about it. Will solid state mass storage ever be economical to use in a DVR? Not for a very Very VERY long time.

[ashu]

Sure.

The size of these devices has doubled in the past year. At this rate, in just 4 years they will be available in 500 GB capacities, probably for about the same $1000 cost for just the memory as todays 32GB drives (which don't actually exist yet BTW). By that time, 500 GB hard drives will be available for $60, and 2 Terabyte drives will be available for $300. I'm pulling these numbers out of my butt, but think about it. Will solid state mass storage ever be economical to use in a DVR? Not for a very Very VERY long time.

Especially when you consider the possibility that video/broadcast resolutions will also increase,a s will display capabilities. Just how much 2160p60 video do you think that 500GB flash drive will hold? :) (Only ~7.5 hours, by my quick calculation)

[MiamiMan]

Sure.

The size of these devices has doubled in the past year. At this rate, in just 4 years they will be available in 500 GB capacities, probably for about the same $1000 cost for just the memory as todays 32GB drives (which don't actually exist yet BTW). By that time, 500 GB hard drives will be available for $60, and 2 Terabyte drives will be available for $300. I'm pulling these numbers out of my butt, but think about it. Will solid state mass storage ever be economical to use in a DVR? Not for a very Very VERY long time.
Actually the 32GB flash drives do exist (although maybe you mean no one can purchase them yet). I'd give you the url page about it, but I don't know how to include it in this forum.

According to Samsung, the SSD will read and write data at 57 MB/s and 32 MB/s, respectively. Not too shabby!

A very Very VERY long time to me is about 3 years (one year equaling one "very") :)

[haysdb]

Yes, I mean you cannot actually buy one.

Newegg offers 16GB flash drives for $180 after rebate. This is ~$11/GB. A 500 GB hard drive costs 20 cents per gigabyte (and is STILL too expensive to put in even a $700 DVR.). The flash drive is more than fifty times more expensive per GB, and has a maximum capacity 1/50th that of the largest currently shipping hard drive. As the cost of flash memory comes down, so too will the cost per MB of magnetic storage.

People have been predicting the demise of magnetic media for a very long time. A DVR is about the LAST place I would expect to see solid state memory employed. Low power requirements? Don't need 'em, we're plugged into AC. Silent operation? Can you actually hear the hard drive in your S3? Having something even quieter doesn't add any value. Reliability? We're not talking mission critical application here, but recorded TV. Hard drives are more than reliable enough for the application.

Comparing flash memory in a camera to the 24x7 operation of a DVR hard drive is hardly fair.

[haysdb]

According to WikipediaNote: flash memory supports only a limited number of erase/write cycles (about 300,000) before a particular "sector" can no longer be written. Typically the controller in a CompactFlash attempts to prevent premature wearout of a sector by mapping the writes to various other sectors in the card - a process referred to as wear leveling.

[MiamiMan]

haysdb: I've just been informed that this year's CES, the company A-DATA presented SSDs in 32 GB, 64 GB, and 128 GB. The U.S. military has been using 256 flash drives for several years now. SanDisk also has 32 GB flash drives and are now available to manufacturers.

I suspect that the Wikipedia article about erase/write cycles refers to the key chain type flash drives which do not have the demand as do computer hard drives.

According to SanDisk, their flash drive is rated at "two million hours mean time between failure (MTBF)."

[jautor]

haysdb: I've just been informed that this year's CES, the company A-DATA presented SSDs in 32 GB, 64 GB, and 128 GB. The U.S. military has been using 256 flash drives for several years now. SanDisk also has 32 GB flash drives and are now available to manufacturers.

I suspect that the Wikipedia article about erase/write cycles refers to the key chain type flash drives which do not have the demand as do computer hard drives.

According to SanDisk, their flash drive is rated at "two million hours mean time between failure (MTBF)."

Nope, two different things completely. Flash is limited in its erase/write cycles. Depending on the wear leveling techniques and the manufacturer, you'll get different numbers, but 100k-200k cycles is the top end of the range. MTBF is an operating statistic, and isn't the same thing.

As was mentioned earlier in the thread, folks have been predicting the demise of hard drives since, well, shortly after they were invented. But they continue to provide the most storage per dollar, and the highest capacity, usually by an order of magnitude each...

The benefits of solid state storage (low power, quiet, shock/vibe tolerance, very fast random read performance) for a DVR application just don't apply well enough to overcome the 50-100x cost. Yes, it's nifty, but very, very, very few people would pay for it. If one was really concerned about hard drive reliability in a DVR device, a simple mirrored drive pair would solve that concern a lot cheaper than flash...

Jeff

[haysdb]

MiamiMan, I should imagine you will be able to swap out your hard drive for a smaller flash drive. One should look just like a hard drive to the Tivo OS. Meanwhile Tivo can continue to cater to the masses who want their DVR's cheap with massive capacity.

[MiamiMan]

jautor: Actually MTBF is an all encompassing estimate that includes media integrity as well as all the components in the drive.

The flash drive manufactures are also giving warranties equal to platter hard drives. That says a lot.

Even aircraft are now using flash drives in data recording (black boxes). These have to be reliable. Moreover, not many people will accept a laptop, iPod, etc. with poor reliable flash drives.

haysdb: Once the flash drives hit the consumer market, you can bet that the prices will drop like a rock on Saturn. Platter drives require precision made platters, bearings, ball bearings, etc. Flash drives should be far more economical to manufacture than platter hard drives, by far.

[hoyty]

Lets use an example that is currently selling close to a million units a year already, the iPod Video and Nano. For $250 you can get a 30 GB hard drive with a larger screen attached or 8 GB of flash memory with a smaller screen. If apple could make a 32 GB flash based player for less than a 30 GB magnetic hard drive don't you think they would have switched already.

The TiVo version of linux is built with a couple of realtime priorities. First, never fail to record properly. Second, display live TV as best as possible. Third, deal with UI. It has to deal with these priorities in order. That is one of the unique features of the tivo OS.

This is why the UI lags some time, because it is busy doing other things.

[MiamiMan]

Lets use an example that is currently selling close to a million units a year already, the iPod Video and Nano. For $250 you can get a 30 GB hard drive with a larger screen attached or 8 GB of flash memory with a smaller screen. If apple could make a 32 GB flash based player for less than a 30 GB magnetic hard drive don't you think they would have switched already.

The 32 GB flash drives are just being released to manufacturers. You can bet that Apple is looking into it. However, because of Apple's secrecy, you probably won't here about it until its released.

[haysdb]

Once the flash drives hit the consumer market, you can bet that the prices will drop like a rock on Saturn. Platter drives require precision made platters, bearings, ball bearings, etc. Flash drives should be far more economical to manufacture than platter hard drives, by far.What is your prediction for when solid state memory will achieve price parity with hard drives?

My own prediction is "Maybe someday," but I'll wager that it's not within the next 10 years, at least not in the larger sizes. For devices which don't have massive storage requirements, where low power requirements and shock resistance bring genuine value (basically any portable battery operated device), flash memory will find a market. The DVR market will be one of the very last consumer products to find a need for it.

[hefe]

jautor: Actually MTBF is an all encompassing estimate that includes media integrity as well as all the components in the drive.

The flash drive manufactures are also giving warranties equal to platter hard drives. That says a lot.
It says that people are not using them the same way as hard drives.


Even aircraft are now using flash drives in data recording (black boxes). These have to be reliable. Moreover, not many people will accept a laptop, iPod, etc. with poor reliable flash drives.
They are. Up until you "run out" of writes.

There isn't much that will push flash memory to exhaustion faster than constantly writing video streams to it, 24x7.

[haysdb]

Actually hefe, I don't think the limited writes are necessarily a problem for a DVR application given that any one block would be written to no more than once every 30 minutes, and even that is a worst case scenario. 100k cycles would be 50K hours, which is almost 6 years. With any half-intelligent "wear leveling," the life of a 250 GB device would be on the order of a hundred years.

And then again, my math could be off by a decimal point or two.

[hefe]

I see your point, but I suppose you have to account for the OS space as well. Maybe you'd have to move that around too. In any case, the write speed of current flash technology would certainly be inadequate compared to what current hard drives can do.

[MiamiMan]

What is your prediction for when solid state memory will achieve price parity with hard drives?

My own prediction is "Maybe someday," but I'll wager that it's not within the next 10 years, at least not in the larger sizes. I don't know but it would surprise me if it took 10 years. I'm only saying that because of past technology releases. First announced products are almost always extremely high in price, especially those products that are in high demand. These SSDs are already in high demand. People want high capacity SSDs in MP3 players, cell phones, TVs and computers. SSDs are also energy savers.

The SSD manufacturers have to recoup the cost for R&D, tooling, etc. Once they ramp up production and manufacturing, I see nothing to prevent dramatic price drops.

[haysdb]

These SSDs are already in high demand. People want high capacity SSDs in MP3 players, cell phones, TVs and computers. SSDs are also energy savers.Please tell me how these devices are similar to a Tivo. Other than being consumer electronics devices. The Tivo isn't portable, it's not battery powered, it doesn't need to be as small and light as possible.

The SSD manufacturers have to recoup the cost for R&D, tooling, etc. Once they ramp up production and manufacturing, I see nothing to prevent dramatic price drops.I don't either, and I haven't argued that the technology wouldn't see dramatic price drops. But the cost per GB for magnetic storage continues to drop rapidly also, so I maintain that magnetic storage will ALSO remain in high demand any place high capacity and low price are both high priorities, e.g. DVR's.

A few years from now, when 500 GB hard drives cost $60, will SSD drives of comparable capacity be available for less? You say yes, I say no. Let's get back together in 3 years and compare notes, eh?

[Brad Eleven]

Brad, you added a piece of hardware. That costs mucho dinero.

(don't know why I'm piling on more to this thing, guess I just hate my job :-)

No, I haven't added any hardware, although I did replace the first drive cuz it was broken.

I agree, the CacheCard costs more than I want to pay for 3rd-party solution that mayn't work.

How did this thing get to be three pages long, FFS? The menus are slow b/c the database is on disk, and has to contend with other disk requests. Even if you're not recording, the unit could be housekeeping, and it's always buffering, anyway.

IMHO, it's inconvenient, but after trying other [DP]VRs, TiVo rules.

Hey, I know!! If you want control over programming, get MythTV and hack the source. Discussing how well/poorly TiVo programmers did IYHO is like arguing on the internet: Even if you win, hardly anyone will know.

PEACE OUT. :->