[NEWS] - Canterwood stumbles at the gate.

[CougTek]

Guess it'll be a while longer for an 800MHz-bus P4. Intel apparently delayed its new P4 chipsets about two hours after the product launch due to a "product flaw". Specs look sweet(RAID0, DDR400, SATA, Gbit LAN) if you swing that way, but rumors are pointing to logistical issues for the real reason for the delay.

Link

Submitted by The Zoloft Avenger

 

[Tea]

For a minute there, I got really confused! I thought you wrote: "Intel apparently delayed its new P4 chipsets for about two hours". Are we getting down to the point where we criticise companies for being two hours late with a new product launch now, I thought?!

 

[Howell]

For a minute there, I got really confused! I thought you wrote: "Intel apparently delayed its new P4 chipsets for about two hours". Are we getting down to the point where we criticise companies for being two hours late with a new product launch now, I thought?!

Three cheers for ambiguous grammer. And these people call themselves professionals.

 

[Tea]

It wazn't ambiguous grammar, Clippy, it was an ambiguous brain function. I'm getting positively human theze days.

 

[Howell]

No, I had to read it three times as well to get the intended meaning. It would have been unambiguously writen as:

Guess it'll be a while longer for an 800MHz-bus P4. Apparently, about two hours after the product launch Intel delayed its new P4 chipsets due to a "product flaw". Specs look sweet(RAID0, DDR400, SATA, Gbit LAN) if you swing that way, but rumors are pointing to logistical issues for the real reason for the delay.

Or the "apparantly" could go at the end of the sentence.
No offence intended anyone, I'm just bored and nit-picky.

 

[mubs]

Back off, guys! Coug is perfectly understandable all the time. Take two aspirin and come back here in the morning.

 

[GIANT]

[NEWS] - Canterwood stumbles at the gate.

There's actually nothing wrong with "Canterwood," as this is only the chipset. Canterwood is perfectly fine.

The new 3.0 GHz (with 800 MHz FSB) P4 processor is the problem. The lower speed versions of this new 800 MHz FSB series -- 2.8 GHz, 2.6 GHz, and I believe there is a 2.4 GHz as well -- are all fine. It's only the 3.0 GHz that has a problem that has shown up during testing. A stepping will obviously have to occur once the problem is found and cured before the 3.0 GHz and the eventual 3.2 GHz speed (Q3, I believe, for 3.2 GHz) can be released.

The real excitement, however, will be with the 800 MHz FSB P4 will be with the release of the upcoming processors manufactured at 90 microns at the end of 2003. These will zoom up to around 5 GHz over the next couple of years, with the FSB speed zooming up beyond 1 GHz, with a 1 MB cache for standard P4 and up to 4 MB cache for P4 Xeon.

 

[Tea]

Back off, guys! Coug is perfectly understandable all the time. Take two aspirin and come back here in the morning.

In the morning? It waz morning that made me mangle Coug's perfectly understandable words in the first place! Mornings are responzible for more human misery than any other seven lettter word in the lexicon. Think of how many people would ztill be alive if George W Bush and Saddam Hussein spent a lot more time in bed drinking tea inztead of getting up early and plotting invasions!

No zensible person gets up before midday. Except for getting up to bar the door, take the phone off the hook, and make a nice cup of tea to take back to bed, of course. If it'z all the same to you, Mubs, I'll come back in the afternon. One or other of my braincells usually fires up by about lunchtime. Mostly, anyway.

 

[Pradeep]

GIANT: I wonder whether the so-called errata is purely an excuse to build up stocks and avoid another sham paper launch?

 

[Explorer]

GIANT: I wonder whether the so-called errata is purely an excuse to build up stocks and avoid another sham paper launch?

Now THAT would be an absurd business plan! No, I doubt it. They (Intel) have had this same sort of problem before. Remember the "original" 1.0 GHz Pentium III processor in the SECC package? Déjà vu all over again.

Mornings are responzible for more human misery than any other seven lettter word in the lexicon...

Yes, mornings can be bad, especially when you have to get up just before midnight to be at work at 1 AM. Now, let's see if I can correct the heinous grammar in my earlier post:

• The real excitement, however, will be with the 800 MHz FSB P4 will be with the release of the upcoming processors manufactured at 90 microns at the end of 2003. These will zoom up to around 5 GHz over the next couple of years, with the FSB speed zooming up beyond 1 GHz, with a 1 MB cache for standard P4 and up to 4 MB cache for P4 Xeon.

  
  The real excitement, however, will be with the upcoming 800 MHz FSB P4 processors manufactured at 90 microns towards the end of 2003. These will zoom up to around 5 GHz over the next couple of years, with the FSB speed zooming up beyond 1 GHz, with a 1 MB cache for standard P4 and up to 4 MB cache for P4 Xeon...

 

[Pradeep]

GIANT: I wonder whether the so-called errata is purely an excuse to build up stocks and avoid another sham paper launch?

Now THAT would be an absurd business plan! No, I doubt it. They (Intel) have had this same sort of problem before. Remember the "original" 1.0 GHz Pentium III processor in the SECC package? Déjà vu all over again.

Oh yes I remember it well. The infamous Pabster :mrgrn:

 

[Pradeep]

Whilst I have you here Explorer perhaps you could have a gander at this thread?

http://www.storageforum.net/forum/viewtopic.php?t=1920

Thanks

 

[Buck]

Mornings aren't that bad, or at least that is what the Bartender tells me. He gets up pretty early to get things setup for the day. Me, I don’t mind moseying out of bed at noon and into the bar shortly thereafter.

Nice way to links threads in order to get an answer Pradeep.

Giant, nice catch with Canterwood being the chipset.

By the way, who is Zoloft the Avenger? It sounds like antacid medication.

 

[Clocker]

looks like it is back on track

Looks like the problem is fixed, possibly. per Ars:

Posted 4/16/2003 - 10:57AM, by Ken "Caesar" Fisher
Intel Taiwan is saying that the 800MHz FSB-supporting Pentium 4 3.0 GHz will resume normal shipments next week "at the earliest." While somewhat guarded ("at the earliest"), it appears that whatever errata the company uncovered this past weekend before the launch of the Canterwood chipset, it's nothing tremendously serious. While Intel has been slim with the details, it has emerged that the problem only occurred in a segment of their chips.

 

[mubs]

You're right on, Tea. Couldn't have said it better meself.

 

[mubs]

The Inquirer thinks that the screwup is because the new processor will be priced less than the current 3.06 GHz P4. Read about it here.

 

[CougTek]

Three cheers for ambiguous grammer. And these people call themselves professionals.

Back off, guys! Coug is perfectly understandable all the time.

It waz morning that made me mangle Coug's perfectly understandable words in the first place!

Coug just wants to tell everyone that Coug didn't compose the ambiguous or non-ambiguous message above, he merely copied/pasted it from a submitted news. Coug isn't the Zoloft Avenger. That said, both Coug and the one behind "The Zoloft Avenger"'s mask are professionals and deserve to be called professionals.

 

[Tea]

I'm sure you are right, Coug. But I'm not allowed to zay zo becauze Tannin sayz I make too many mistakez in the mornings.

 

[Dïscfärm]

Whilst I have you here Explorer perhaps you could have a gander at this thread?

http://www.storageforum.net/forum/viewtopic.php?t=1920

Thanks

Do you have a mobo with a 133 MHz PCI-X slot? If not, it won't make any difference unless you will own such a mobo within a year. Besides, PCI Express (serial PCI) will be out in a year or so, which will make even 133 MHz PCI-X look like a Sunday stroll through the park in comparison. So, lusting for a future SCSI RAID controller with the highest host bus connector speed should be directed towards a model with features such as PCI Express and Serial Attached SCSI (SAS).

As for going for either "-2" model: If you are there at home with the idea of configuring a 2 or 3 drive array, a single channel model will suffice. A 4 ~ 6 drive array will be a more appropriate fit for a 2 channel RAID controller.

I really don't know your situation (requirements), so I'm strictly making guesses.

A LSI Logic MegaRAID 320-1 goes for around US$350 with the 320-2 going for around US$650. The 320-1X, 320-2X, and 320-4X are all currently being sold to "partners" at the moment. Around June/July, they will be available on a widespread basis. The new "X" series MegaRAID controllers have been shown around in public. The reports from the somewhat-recent CeBIT show in Hannover, was that the MegaRAID X series RAID controllers -- when plugged into a capable PCI-X/133 Xeon mobo -- were capable of astounding I/O rates (benchmarks).

 

[Pradeep]

Thanks for the reply farm. Right now I have the TigerMP with measly 64bit/33MHz slots. My thoughts were of the RAID 5 performance of the current 320-2, which uses a different (slower?) processor compared to the 400MHz one used in the 320-2X. As it is for the home, massive IO is not a major consideration, but the best possible read/write STR is important for me.

I'll be using 3 x15s, and prob 2 or 3 of the new Fujitsu 15Ks. I guess I'll wait for PCI Express and SA-SCSI. Or use the PCI-X 100MHz slots in the forthcoming Opteron mobos.

 

[mubs]

Pradeep, if it's ok to tell, why do you need such high-end storage capabilities at home? Lots of video cap?

 

[Pradeep]

Yup plenty of video.

 

[Howell]

Three cheers for ambiguous grammer. And these people call themselves professionals.

Coug just wants to tell everyone that Coug didn't compose the ambiguous or non-ambiguous message above, he merely copied/pasted it from a submitted news. Coug isn't the Zoloft Avenger. That said, both Coug and the one behind "The Zoloft Avenger"'s mask are professionals and deserve to be called professionals.

I assumed that was a cut and paste from the source, ie. professional journalists. Sorry, I should learn to just keep my trap shut.

 

[Explorer]

...Right now I have the TigerMP with measly 64bit/33MHz slots. My thoughts were of the RAID 5 performance of the current 320-2, which uses a different (slower?) processor compared to the 400MHz one used in the 320-2X. As it is for the home, massive IO is not a major consideration, but the best possible read/write STR is important for me.

I'll be using 3 x15s, and prob 2 or 3 of the new Fujitsu 15Ks. I guess I'll wait for PCI Express and SA-SCSI. Or use the PCI-X 100MHz slots in the forthcoming Opteron mobos.

There are many things that I must say and ask. I’ll attempt to organize them in a meaningful way:

• Do you plan on writing and editing video on this array, as opposed to just storing video data files on it? Consider offline storage.

If you only need to store a LOT of video files, you would be far better off – in a fiscal sense -- purchasing the largest 5400 RPM ATA drive(s) available along with an ATA drive bay and the required number of plug-able ATA drive carriers (1 for each hard drive mechanism). With this simple system, you need only to plug the correct drive into the drive bay to restore or backup video files for editing – with the editing occurring, presumably, on a faster and smaller disc drive.

By the way, a “large” modern 5400 RPM ATA hard drive is quite adequate for recording video onto. However, a 5400 RPM ATA hard drive is not recommended for non-linear editing, due to its relatively sluggish seek/access performance.

• Do you want your video data online at all times?

Personally, I wouldn’t, but then again, I’m not you (or anyone else reading this message). If you want your video data files online, you can take the above and modify it a bit by adding a 2-channel ATA RAID controller and configuring a RAID-1 storage volume with 2 “large” 5400 RPM ATA drives. Of course, a RAID-1 storage volume will only give you the capacity of 1 drive, but you’ll have drive failure resilience. If the RAID-1 storage volume will not provide enough storage capacity, then RAID-3 or RAID-5 can provide the storage volume capacity along with some resilience against drive failure.

The only reason I mention RAID-3 is because if you plan on real-time editing of video (or audio) using a storage volume comprised of one of the parity RAID levels, then RAID 3 or RAID 30 is definitely the best choice for editing large files. If you only plan on simple storage and retrieval of video files, RAID-5 is fine. If you are now thinking that RAID-3 might be the way to go, then you’re definitely stuck with SCSI RAID.

...with measly 64bit/33MHz slots. …the current 320-2, which uses a different (slower?) processor compared to the 400MHz one used in the 320-2X. …but the best possible read/write STR is important for me. …3 x15s, and prob 2 or 3 of the new Fujitsu 15Ks...

OK, how should I say this? Unfortunately, you are suffering from Common RAID Misconception Syndrome. I’ve seen a LOT of it for years at the old Storage Review. I used to speak up about some of these RAID misconceptions, but all I usually got was scorn -- if anyone paid attention.

The throughput of the RAID co-processor has to be precisely matched to the host adaptor, host bus type (width x speed), number of drive data channels, and the channel speed. In fact, all block components of a RAID storage subsystem should ideally match up closely in throughput characteristics. If you install a MegaRAID 320-2X controller in a 64-bit / 33 MHz PCI slot in place of a MegaRAID 320-2, you likely will see little if any performance improvement.

…wait for PCI Express and SA-SCSI. Or use the PCI-X 100MHz slots in the forthcoming Opteron mobos.

Nobody actually is being told (yet) how fast third party vendors (like Adaptec, LSI, 3Ware, etc) will be ramping up PCI Express expansion bus card products. In a lot of people’s opinion, Intel is pushing PCI Express out the door about 9 months to a year early. PCI-X hasn’t even ramped up to 266 MHz yet (or 533 MHz). Allegedly, Intel’s first PCI Express chipset will arrive at the end of THIS year at about the same time as the new 90nm Pentium 4 (Pentium V ?) processor. Intel wants PCI Express everywhere, not just in servers and advanced deskside workstations, though it’ll likely start off there. I’ve heard ATI and Nvidia are already working on PCI Express graphics adaptors, since PCI Express will completely replace AGP.

 

[Pradeep]

For mass storage I use large 5400rpm IDE drives. I also burn this data onto DVD as time permits.

The reason I am pondering SCSI RAID is that when editing, right now I go from one X15-36LP to another (booting off another X15). Using a Tekram U3D U160 controller. The x15.2s are in an SCA enclosure. I use IfoEdit a lot, and it is very disk bound.

So in the future I see myself running two arrays, perhaps in RAID 3 as you suggest. Certainly for my editing I will be remaining with SCSI, I already have a fair bit of it, I like the speedy access times, and I appreciate the peace of mind advantage when compared to IDE (I had the displeasure of owning a 45GB 75GXP). That said I will certainly be picking up some 250GB IDE drives for mass storage when they drop a bit in price.

Sounds like PCI-Express won't be around for a while, so when I get the Opteron board I'll see what controllers are around at that time. I'll have to get a new controller at that time because my Tekram is 5V only and won't fit in a 3.3V PCI-X slot. Easiest thing may be a GigEthernet connection and edit across the network, using the existing Tiger as a glorified storage box. I can get the 320-1 as an eval board for about $450, but it sounds like it would be a waste on my current PCI slots.

You've given me a lot to think about, thanks for the good info.

 

[Pradeep]

I should add that once I've finished with it, it may be used as a hotel reservations/back office server.

 

[Pradeep]

Ooops, I can get the 320-2 for $450. Damn no edits