Does new SCSI card mean revamped hardware?
<a href="http://www.attotech.com/press/pressattolsi.html" target="_blank">http://www.attotech.com/press/pressattolsi.html</a>
It's interesting that ATTO would release something like this, considering it has features like PCI-X compatibility. I wonder if this is a clue that we're going to see some heavily pumped hardware in July?
(Hopefully )
It's interesting that ATTO would release something like this, considering it has features like PCI-X compatibility. I wonder if this is a clue that we're going to see some heavily pumped hardware in July?
(Hopefully )
Comments
TTO Technology, Inc., is to be commended for foresight. or is that foursight or forsight or foresite or fo....oh forget it.
<strong>I would assume PCI-X is needed to achieve full bandwidth. Isn't the stated bandwidth of the current PCI bus only 215 MBps? That means a 320 SCSI Card would definitely oversaturate the bus.</strong><hr></blockquote>
The G4's have 64bit 33MHz PCI slots, providing 264MBps of shared bandwidth. I don't know much about the current G4's, but I believe the internal ATA, ethernet, and firewire are on a separate bus than the PCI bus. This is unlike some x86 motherboards that channel all of the data through a 32bit 33MHz PCI bus. Now this is 132MBps shared by EVERYTHING (ATA controllers, ethernet, sound, etc)
<strong>
The G4's have 64bit 33MHz PCI slots, providing 264MBps of shared bandwidth. I don't know much about the current G4's, but I believe the internal ATA, ethernet, and firewire are on a separate bus than the PCI bus. This is unlike some x86 motherboards that channel all of the data through a 32bit 33MHz PCI bus. Now this is 132MBps shared by EVERYTHING (ATA controllers, ethernet, sound, etc)</strong><hr></blockquote>
I guess that proves my point that the current PCI bus has less than 320 MBps of throughput
<strong>
The G4's have 64bit 33MHz PCI slots, providing 264MBps of shared bandwidth.</strong><hr></blockquote>
That's 264MB/s of theoretical peak bandwidth. Apple claim 215MB/s sustained bandwidth somewhere on the G4 page.
[quote]<strong>I don't know much about the current G4's, but I believe the internal ATA, ethernet, and firewire are on a separate bus than the PCI bus.</strong><hr></blockquote>
You're right about ethernet and FireWire, but ATA is still inside KeyLargo (which is, as a whole, connected via the aforementioned PCI bus).
[quote]<strong>This is unlike some x86 motherboards that channel all of the data through a 32bit 33MHz PCI bus. Now this is 132MBps shared by EVERYTHING (ATA controllers, ethernet, sound, etc)</strong><hr></blockquote>
Well, to be fair, most current x86 mainboards do circumvent that bottleneck by employing a proprietary high-speed interface between north- and south-bridge (HubLink (Intel), V-Link (VIA), MuTIOL (SiS), HyperTransport (nVidia), ...).
Out of the current variety of chipsets, I think the AMD760 is the last one to just use PCI to interconnect all the stuff.
Bye,
RazzFazz
<strong>I guess that proves my point that the current PCI bus has less than 320 MBps of throughput</strong><hr></blockquote>
Not quite. there's another variant of PCI. known as 64-bit 66Mhz PCI. which tops out at 352Mbps.
[quote]Originally posted by Matsu:
<strong>PCI-X is an interesting development. It's been available for server type duties for a while</strong><hr></blockquote>
Not a very long while. it's only been out since the earlier part of this year. and at that. only in experimental form. I have yet to hear of _ANY_ shipping computers that use PCI-X on their motherboards. AMD's 12Gbps HyperTransport(Which has been on the market for well over two years) is the better bus. and even Motorola's 8Gbps RapidIO kicks the $&@% out of PCI-X.
[quote]Originally posted by Matsu:
<strong>and I think it'd make a nice update to the mac architecture. While even faster stuff is in the pipe, nothing else gives backwards compatibility to a LOT of PCI sound, video, effects, processing, and capture cards. Some of it, in the case of HD and SD video effects, pretty expensive stuff.</strong><hr></blockquote>
Wrong. all of the newer busses feature PCI adapters(And let's not forget slot-to-slot adapters. like Apple's old PDS->NuBus adapters or SGI's XIO->PCI adapter).
[quote]Originally posted by Matsu:
<strong>I think PCI-X, with about a GB on bandwidth, should be good enough for the next 4 or 5 years while letting people keep their expensive PCI cards.</strong><hr></blockquote>
Yeah. right. <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
[quote]Originally posted by Matsu:
<strong>PCI-X is probably the next interface for Apple and most PC vendors as well.</strong><hr></blockquote>
I certainly hope not. there's <a href="http://www.hypertransport.org" target="_blank">something much better</a> out there already.
PS: before posting a reply. I'd recommend reading <a href="http://forums.appleinsider.com/cgi-bin/ultimatebb.cgi?ubb=get_topic&f=1&t=001359&p=" target="_blank">this prior thread on why PCI-X sucks wind</a>.
Eric,
[ 04-10-2002: Message edited by: Eric D.V.H ]</p>
<strong>I already read it. Agree with a lot of it. I just don't have that much faith in PC vendors, including Apple. We're talking desktops here, right? The industry's gonna screw this up. I just have this feeling they'll screw it up, either we're going to get backwards compatibility and slower than ultimately possible speeds, or we're gettin a few avenues to lots of speed, without any backwards compatibility and a nice little standards war (scuffle?) to boot. I think we can sit back and watch as the delivery of the discussed technologies (at least initially) falls well short of the promise. Maybe 4 years is optimistic for any PCI variant, and Apple may be the worst off of the bunch as they await an actual emergent 'standard' (VHS-Beta type FUD at work perhaps) they might just end up sticking to PCI longer than anyone else. </strong><hr></blockquote>
My glass is half full. isn't your's that fine spattering of damp glass shards on the floor?
As for backwards compatability. _all_ of the new standards have full PCI support through bridge-chips. so that isn't a factor.
Eric,
When I said the "current PCI bus", I was referring to the bus implemented in current Apple computers.
<strong>
Not quite. there's another variant of PCI. known as 64-bit 66Mhz PCI. which tops out at 352Mbps.
</strong><hr></blockquote>
352MB/s?
Where does that number come from?
Seems a little low for sustained throughput, given the peak rate for PCI64/66 should be 64bits * 66MHz = 528MB/s...
Bye,
RazzFazz
<strong>When I said the "current PCI bus", I was referring to the bus implemented in current Apple computers.</strong><hr></blockquote>
This is one of my pet peeves. at the time I bought my shiny new Formac Proformance 3+ 32MB(A 66Mhz 32-bit card made with the Blue and White G3 in mind. although. it runs fine on mismatched versions of PCI. just like all high-end cards) for my Power Macintosh 6500/250(Which used 33Mhz 32-bit PCI). I assumed that when I replaced my 6500 with a G3/G4. it would boast full 66Mhz 64-bit PCI. But instead it just featured the exact opposite of what I was hoping for. 33Mhz 64-bit PCI. ugh.
Eric,
<strong>352MB/s?
Where does that number come from?
Seems a little low for sustained throughput, given the peak rate for PCI64/66 should be 64bits * 66MHz = 528MB/s...</strong><hr></blockquote>
Ahwoops! I just got that number by lazily doubling Majuki's 33Mhz 64-bit PCI figure(Which was supposed to be in <a href="http://www.webopedia.com/TERM/M/Mbps.html" target="_blank">Mbps</a> anyway. not <a href="http://www.webopedia.com/TERM/M/MBps_megabytes.html" target="_blank">MBps</a>):
[quote]Originally posted by Majuki:
<strong>The G4's have 64bit 33MHz PCI slots, providing 264MBps of shared bandwidth. I don't know much about the current G4's, but I believe the internal ATA, ethernet, and firewire are on a separate bus than the PCI bus. This is unlike some x86 motherboards that channel all of the data through a 32bit 33MHz PCI bus. Now this is 132MBps shared by EVERYTHING (ATA controllers, ethernet, sound, etc)</strong><hr></blockquote>
I know I was wrong. but I'm not sure whether you're exactly right. MacKido agrees with your figure. but The Webopedia says it should be 4MBps higher. either way. my assumption was WAY off.
Eric,
[ 04-12-2002: Message edited by: Eric D.V.H ]</p>
<strong>352MB/s?
Where does that number come from?
Seems a little low for sustained throughput, given the peak rate for PCI64/66 should be 64bits * 66MHz = 528MB/s...</strong><hr></blockquote>
Ahwoops! I just got that number by lazily doubling Majuki's 33Mhz 64-bit PCI figure(Which was supposed to be in <a href="http://www.webopedia.com/TERM/M/Mbps.html" target="_blank">Mbps</a> anyway. not <a href="http://www.webopedia.com/TERM/M/MBps_megabytes.html" target="_blank">MBps</a>):
[quote]Originally posted by Majuki:
<strong>The G4's have 64bit 33MHz PCI slots, providing 264MBps of shared bandwidth. I don't know much about the current G4's, but I believe the internal ATA, ethernet, and firewire are on a separate bus than the PCI bus. This is unlike some x86 motherboards that channel all of the data through a 32bit 33MHz PCI bus. Now this is 132MBps shared by EVERYTHING (ATA controllers, ethernet, sound, etc)</strong><hr></blockquote>
I know I was wrong. but I'm not sure whether you're exactly right. MacKido <a href="http://www.mackido.com/Hardware/bussed.html" target="_blank">agrees with</a> your figure. but The Webopedia <a href="http://www.webopedia.com/TERM/P/PCI.html" target="_blank">says</a> it should be 4MBps higher. either way. my assumption was WAY off.
Eric,
<strong>
Ahwoops! I just got that number by lazily doubling Majuki's 33Mhz 64-bit PCI figure(Which was supposed to be in Mbps anyway. not MBps):
</strong><hr></blockquote>
Um, may I suggest re-checking your integer unit?
Majuki stated 264MB/s for PCI64/33 and 132MB/s for PCI32/33, none of which really doubles to 325MB/s as far as I can tell...
[quote]<strong>
I know I was wrong. but I'm not sure whether you're exactly right. MacKido agrees with your figure. but The Webopedia says it should be 4MBps higher. either way. my assumption was WAY off.
</strong><hr></blockquote>
Well, depends on how you see 133MHz.
I guess 533MB/s would in fact be the correct number, as 133MHz usually really is 133.33333...MHz, which quadruples to 533,33333...Mhz.
Bye,
RazzFazz
[ 04-12-2002: Message edited by: RazzFazz ]</p>
<strong>Um, may I suggest re-checking your integer unit?
Majuki stated 264MB/s for PCI64/33 and 132MB/s for PCI32/33, none of which really doubles to 325MB/s as far as I can tell... </strong><hr></blockquote>
Argh. <img src="graemlins/embarrassed.gif" border="0" alt="[Embarrassed]" />
[quote]Originally posted by RazzFazz:
<strong>Well, depends on how you see 133MHz.
I guess 533MB/s would in fact be the correct number, as 133MHz usually really is 133.33333...MHz, which quadruples to 533,33333...Mhz.</strong><hr></blockquote>
Thanks for clearing that up.
Eric,