USB 3.0 demonstrations dazzle: uncompressed 1080p transfer proves simple
By Darren Murph 
posted Nov 19th 2008 5:47PM
posted Nov 19th 2008 5:47PM
Please note: If you're a current Engadget commenter, you do not need to re-register. Simply login and you will be prompted to select a username.
humph. still not as good as firewire
/fanboi
Not when Engadget writes about it at 450Mbps. However, when properly referenced at 450MBps, yes, yes it's better than firewire...
How on earth does a tech blog get bits/bytes mixed up, when it quite clearly states "PLDA demos 450 Mbytes per second". Even if the capitalisation is confusing, it actually says 'bytes' numerous times in the article.
Psst, Engadget - there's a difference, 8 times the difference.
Yeah really, I was very confused when I saw that..
Engadget made a pretty n00b mistake there, that's for sure. We should be able to rank posts. Jee, I wonder what ranking that kind of awesome fact reporting would get?
Bytes != Bits
Those must be some nice FPGAs they're using, looks awesome, I just want some USB 3.0 compatible gear already.
This makes me horny.
Is that weird?
Not at all man.... I'm freaking wet just reading this!
.....wait
well do you have a usb port instead of a Asshole ??
Yes. I would seek professional help.
Welcome to nerdgasm my friend.
Take a break from Engadget and join "Nerd Anonymous". ;)
I've got August 2298 marked on my calender.
Add it to this
Predictions for the 23rd century
http://en.wikipedia.org/wiki/2298
Just a question. What harddrive on the market can write at speeds close to 450Mbytes/sec???
Damm, pity we wont live to see an official release
You'ld think by Year 2298AC we would've made more progress..
But I want to see the day Easter falls on March 22nd again, and Google indexes EVERYTHING...
I want to see the Star Trek cast.
3.0 GB/s - http://gizmodo.com/382013/western-digital-velociraptor-is-new-fastest-hard-drive-ever
Sorry, that's Gb/s, the writers of the article effed up and by doing effed me up too . . lol. Still that's 375 MB/s, pretty close, and there's always USB-to-USB transfer or reading into memory from the USB device . . .
@ Homeboy
Not all data goes to/from the hard drive. A lot of stuff just goes to RAM which is plenty fast for USB3 and with RAM sizes constantly increasing I think even less will need to be written to the drive (swap/VM).
http://en.wikipedia.org/wiki/23rd_century#Other_predictions
An uncompressed 1080p stream at 30 fps is far more than 450 Mbps. The 1080p stream itself is compressed (as per the HDTV spec).
An uncompressed 30 fps stream at 1080 x 1920 is going to be far more than 450 Mbps.
It's a typo. The article states 450MBps. Bytes, not bits. Isn't the USB 2.0 spec 480Mbps?
August 2298 ??
Wireless, uncompressed 1080p HD will be out next year.
http://news.softpedia.com/news/Panasonic-Demonstrates-Uncompressed-Full-HD-Video-Wireless-Streaming-94661.shtml
Wish my Popcorn Hour A-110 had USB 3.0 =(
Soon the sun will rise on a beautiful day where we will never again see another component, HDMI, or DVI cable again, everything will be USB based! The Koreans will stop being evil, China and Russia will get along, the Jews and the Muslims will resolve their differences, and the world will be at peace!
Then again, maybe not.
"currently on track for an August 2298 release,"
Typo?
joke
I thought engadget readers didn't require the tags...
Hahaha, it was supposed to say sarcasm /sarcasm tags... Oh well
Your sarcasm tags were omitted because we don't require them.
If you aren't intelligent enough to detect sarcasm, you shouldn't be commenting here. :P
cant usb 2 theoretically handle 480Mb/s??? i know that in the real world that this speed would probably never be achieved, but it is still possible. If they really wanted to show off usb 3 they should at least get it running at around 3Gb/s because it has a theoretical max of 5Gb/s. But thats just me. Take esata for example it could already easily handle 450Mb/s, which mkes this "dazzle" not that impressive.
It was suppose to read 450Mbytes or 450MBps, it's bytes not bits. USB 2.0 can only handle 60MB theoretically.
oh ya, big difference. they should correct that to MB/s i hate when people do that. That makes a big difference because eSata only has a real life speed of around 300MB/s, and theoretical of 375. the only thing is that no harddrive could even handle that much bandwidth unless it was in raid 0, i believe, and even then it might not even reach that. Personally i think its great having this much bandwidth but what would the average consumer use it for as of right now, nothing because anything that can handle even a fraction of that bandwidth would cost a fortune. Just my 2 cents.
What about connecting USB mouse to the same hub and trying again?
USB remains USB...
Yeah, but at this speeds, your mouse would be moving on its own!
450MBps is 3600Mbps. with a theoretical speed of 5Gbps, this is a pretty good real life speed.
Ya, but you have to relize that the article has a miss print, Mbps is Mega BITS per second, and MBps is Mega BYTES per second. What he should have said is that it did 450MBs as pointed to me above by Kevinm.
Semicolon - I think he is aware of this, don't reiterate something somebody has said to you only a few lines up.
*Facepalm*
Also! Aside from the genuine mistake, I think that is a good throughput and although drives may not be able to write at the speed at the moment is besides the point. If they were announcing technologies which were only just coping with todays read/write speeds, then a year down the line it would be totally useless. You have to future proof these things as they last for years.
Brand new standards aren't based around the average consumers current use for them, if they were we wouldn't be getting anywhere.
Ummmmm.......No such thing as 1080p @ 30fps...try 60. Otherwise its not 1080p is it...Not 1080i either....More like 1080 L 8)
@topsitecourie
Yes there is 1080p @ 30fps. Most 1080p movies are 24fps. Just because it displays 30 frames doesn't mean it has to display them interlaced, does it?
Cool, they are running KDE 4.1
So I guess Linux users can expect support for USB 3.0 very soon :D
Before all other operating systems?
Of course.
Dammit, bigcow, you stole my thunder. I was hoping no one else would notice so I could be the king of the linux nerds :(
450Mbps? LOL, multiple standards of reference FTW.
Yes, I do believe they meant 450MBps, 1000x that of USB 2.0
1080p @ 60 fps is approximately 356MBps, so it's entirely feasible to get those speeds out of a port that theoretically goes more than half a gig per second...
30 fps at
186,624,000 so around 187 MBps
which translates to (at 60 fps)
373,248,000 or around 373 MBps
Also, we're frickin talking about a bitmap image bit depth for each frame.
should look pretty nice on my GDM-FW900. lol =]
1080p@30fps raw should be 1920x1080 pixels/frame * 24 bits/pixel * 30 frame/sec ~= 1.5Gbps
can you imagine the lulz when watching that? I'd be like LOOOLOLOLOLOLOL
=]
"1080p@30fps raw should be 1920x1080 pixels/frame * 24 bits/pixel * 30 frame/sec ~= 1.5Gbps"
HD is delivered as YUV or YCbCr at 16 bits per pixel, not 24. (4:2:2 sampling, not 4:4:4) so for your example its 995Mb/s or around 124MB/s
No idea where the extra 336MB/s would have went.
Ah.. Reading the article the reference makes more sense. They did not claim that they were transfering the video at 450MB/s they claim that the throughput of their USB 3 implementation was good for an effective data transfer rate of more than 450MB/s which makes sense since the theoretical transfer rate of USB 3 is 600MB/s.
I wish everyone would just use megabytes per second as the unit of measure. I know what a megabyte is. I never measure anything in megabits.
Manufacturers use megabits... People use megabytes.
Which makes more sense to you?
Crap, that iTunes update is 72 megabytes.
or
Crap, that iTunes update is 576 megabits.
Numbers in Megabits always look more impressive, and big numbers sell.
Although I would like the everyone to use bytes as well. +1
Well there really is some reason behind this:
If your storing a file, your storing it in Bytes in memory so saying 72megaBytes makes sense
If your transferring data over a wire, your sending it bit by bit so saying a theoretical speed of 5gigabits per second makes sense, people just shorten bits per second down to bits...
Also the throughput can be calculated in megabits per second, but it is much more difficult to calculate the throughput in mega bytes per second.
This is because you send your data (bytes) along the wire in packets containing other data such as parity bits for error checking. Hence the reason on a 100Mbps line you would never get 12.5MBps throughput.
And the CPU usage on the host was 50? 70? 90? 99%?
ha!
No kidding. Who cares if you can fly at 450 megs a second, if you are utilizing 80% of your cpu.
Further excitement towards S3200, in my opinion, when you're retaining relatively low (by comparison) host cpu demands.
*rumor: Firewire 800 only lacks software to run at 3200 speeds... read it somewhere... c'mon software update!
"*rumor: Firewire 800 only lacks software to run at 3200 speeds... read it somewhere... c'mon software update! "
I really doubt that is the case, as the higher frequency and bitrate probably requires newer hardware - then again I'm not an engineer so you may be correct. However, regardless of that point, S3200 does indeed use the same cable and connector as standard FW800 which is great to hear!
Am I dated, Daza? Or did we get rid of the parity bit for this type of thing? (I could easily be dated since it was ~45 tears ago when I learned this stuff). But if it still applies, then Daza's admonition should'a been...
Psst, Engadget - there's a difference, 9 times the difference. (???)
1 Byte = 8 bits
Mbps = Megabits per second
MBps = Megabytes per second
hope that clears it up
@Michael Scrip
8 bits in a byte
10 bytes = 80 bits
450 mbytes a second = 3600 bits
... in case you ever have to measure things in bits.
Yeah... I know that. I'm just asking why they measure stuff that way?
USB 2.0 is 480 Mbit/s... so then I have to convert it into 60 MB/s because megabytes mean more to me.
I guess it's just me.
Errr... No, that doesn't quite 'clear it up' in the context of the question. There's always been 8 bits to a byte (for hex systems) regarding data representation, even back 30 and 40 years ago. But internally for error checking (and largely not mentioned *unless* you cared about parity and error checks), there was an 'extra' bit called the parity bit. Parity bits were still in use at least throughout much of the 70s and I'm fairly sure into the 80s.
So the question wasn't really related to whether there are 8 bits to a byte regarding data, but have we gotten rid of the need for the 'extra' parity bit (first introduced in the early 1950s if I recall). Likewise, in normal use of the language, '10 bytes = 80 bits', but if you dealt at the hardware level where transmission over buses and say, disk I/O, there was indeed 90 bits transmitted (undetectable to higher level languages).
Hope that clears the question up.
USB is a serial bus that uses CRC (not parity) to protect the token and data packets.
Since it is a serial bus, the CRC fields (as well as some of the other non-data fields) go through the same channel as the rest of the data. Hence, you get a theoretical maximum (where everything is just data), to a practical throughput of the channel, which takes into account all the non-data related stuff.
Sounds cool but if its coming out in August 2298 like it says it the article then I don't think I'm ever going to use it and u aren't either. But it sounds interesting but by 2298 i would hope it would be much more advanced than that. I'm not waiting for it.
Welcome to the internet my friend, let me introduce you to my friend sarcasm over here...
By 2298 we'd be DEAD.
And yes, meet sarcasm.
Hmmm... I Wiki'd "Parity bit" and it seems like it's still used, but whether it's needed for USB is not mentioned. However, since the whole thing is specifically for detecting errors for transmission, I'd not be surprized if it is - it's certainly an appropriate application. Excerpt:
USAGE: "Because of its simplicity, parity is used in many hardware applications where an operation can be repeated in case of difficulty, or where simply detecting the error is helpful. For example, the SCSI and PCI buses use parity to detect transmission errors, and many microprocessor instruction caches include parity protection. Because the I-cache data is just a copy of main memory, it can be thrown away and re-fetched if it is found to be corrupted."
Thanks muchly, Genematcher. And I'll look that up a bit more! :-)
I'd like to see all connectors on a PC replaced by USB:
-USB monitors so you can easily connect as many as you want
-internal USB to connect Hard disks and DVD drives. Drives can be used internally or externally.
-who needs firewire or eSata when USB is faster and powered?
-USB speakers, head phones and microphones, so I can plug them in where I want (back, front, on the monitor's USB hub)
I was there at the DevCon.
The Synopsys demo:
They're actually padding the data to make it 450 MB/s. If you think about it, uncompressed 1080p30 is only 1920x1080 [resolution] x 3bytes [24bit RGB] * 30 [frames/sec] = 186624000 Bytes/s which is about 177.98 MBytes/s.
The PLDA demo:
This demo is pretty cool because you can modify the packet size & what-not to see the affects of them (larger packet size, higher throughput). But they're just basically pushing data back to back on the line, which is unrealistic in a real-life usage point of view (same with Synopsys' demo).
The Fresco demo:
Their demo is definitely not as flashy, but it was featured in the keynote speech Monday morning. What they're showing is the real data transfer speed at the software level, which should be closer to what we would see when it becomes a product. Intel wrote the software/drivers for the demo, and it complies with the new xHCI host specification that was released a couple months ago. The other 2 companies don't have xHCI host.
Oh, forgot to mention that the 2nd Fresco demo is showing some mass storage device running a Linux open-source driver (also written by some Intel person).
in all likely hood it's: 1920x1080x16bit (not 3, nor 24)x24 or 30fps under h.264 compression
I think people are getting confused here, or maybe it's just me (probably!), but there is a difference between bitrate streaming and transferring a file.
most devices compress with the free and awesome h.264 format (or the less free and less awesome avchd such as many cameras), h.264 ranges from 20Mbit to 200Mbit/s in streaming, depending on how the stream is setup. vc-1, usually for bluray format, is roughly 45Mbit and is not dynamic.
what makes this usb spec useful is if you have a file to transfer directly, not something to stream.
this is not to say that usb3 isn't great, we all want faster transfers! imagine transferring a 2 hour hd movie from a camera over usb 2.0. keep in mind though that bus speed is not all the factor, what media are you grabbing from? a 15000 rpm hdd in a RAID setup still maxes out near 350MB/s...I'm not sure what's even faster than this without actually generating the data on the fly.
but you have to ask your self, are you going to be transferring 2 hour hd movies from cameras or just watching the movie from a server? in that case, even ethernet would suffice, maybe even wifi...
oh yeah, one more thing: this is completely separate from video display transfer speeds, which use raw data for the video. eventually this will probably change to have chips to decode video feeds, but for now it's raw pixel data over cables. what is raw 1080p data for a monitor or tv across displayport or hdmi, you may ask? around 3-4Gbit/s
Our Synopsys demonstration was running between 463 to 464 Megabytes per second. The video only takes less than 30% of that bandwidth, so we added bytes to show the real throughput.
The bandwidth reading is in the lower right hand corner of the screen on the right, and is visible only when viewing it in person. I will ask our engineers to increase
Thank you to the the author of the article. I actually have to switch back and forth between bits and bytes a lot in my presentations so I understand how these things happen.
Eric Huang Product Marketing Manager for USB 3.0 at Synopsys
My blog is here for more info on USB 3.0: http://synopsysoc.org/tousbornottousb/index.php
http://synopsysoc.org/tousbornottousb/index.php
Please note that iamnotageek is most likely a Fresco employee named Dian. But that is okay because it's not biased or anything :)