Nforce 4, Q4 2004, Dual Opteron, SLI, details leak
NForce 4 details leak
Four different versions emerge
By Fuad Abazovic: Monday 23 August 2004, 12:08
EVEN THOUGH THE CHIPSET won't materialise before December, a French site already has some details about it. We should use the plural and say chipsets as there will be four different versions of Nforce 4, according to the source.
The first, Nforce 4 will be an entry level chipset and it won't feature high end features but should be a good performer.
The Nforce 4 Ultra will add some additional functionality such as a S-ATA II and will add 3 Gbit/s transfer. Hard drives that support this option should be ready around them. Nforce 4 Ultra will have firewall options - something that Nvidia is pushing hard.
The third chipset, Nforce 4 SLI, will, as the name suggests have SLI support and will finally offer a dual slot PCI Express motherboard where you can potentially use two graphic cards. Two Nvidia cards, as SLI is Nvidia only at this time, or should we say ex-3DFX.
The last, the Nforce 4 Pro, will have support for two Opteron CPUs and will be used for workstation and server market and AMD is definitely picking up on this market segment.
The first silicon should be ready as we speak and manufacturers are playing with them in Taiwan.
You can find more details here.
What we can add is that at least some of these versions are expected before Yule 2004. µ
[edit translated web page]
Détails! 19-08-2004 10:47:49 - Samuel D. Whereas the chipset successor of the nForce 3 at nVidia would owe s'appeller nForce 4 logically, one speaks yet currently only about his code name, the CK8-04. A Russian source wishing to keep l'anonymat forwarded to us of information on this next chipset. One learn qu'nVidia thus will not announce one, but four versions of the nForce4, of which here some details: nForce 4: Standard version of the chipset, which is intended to equip the GCV d'entrée with range and which will allow the construction of cheap mother chart without sacrificing the performances nForce 4 Ultra: Version improved of the precedent having additional functionalities. It will support the SATA-II (3 GBit/s) as well as advanced options of firewall. Its price will remain reasonable nForce 4 SLI: As one can s'en doubt, this chipset intended for the workstations and the fortunate gamers will be the only able one to support mode SLI of new charts NCV Express train of nVidia nForce 4 Pro: Intended for the waiters and thus for solutions Bi-Opteron, the nForce 4 Pro will be thus the first chipset of nVidia only intended for this market. It will be capable to support a controller NCV additional Hypertransport Express train of nVidia baptized IO4, but c'est another history... In short, the new awaited range trés of chipset nVidia seems promising. With current l'heure, the final first samples start to be available in the laboratories of the manufacturers, l'annonce official is currently planned for the second quizaine of September... Wait & See.
[edit translated web page]
There they go. Off to a blazing start. I wonder if AMD will have their dual core processors readied by XMAS?
Four different versions emerge
By Fuad Abazovic: Monday 23 August 2004, 12:08
EVEN THOUGH THE CHIPSET won't materialise before December, a French site already has some details about it. We should use the plural and say chipsets as there will be four different versions of Nforce 4, according to the source.
The first, Nforce 4 will be an entry level chipset and it won't feature high end features but should be a good performer.
The Nforce 4 Ultra will add some additional functionality such as a S-ATA II and will add 3 Gbit/s transfer. Hard drives that support this option should be ready around them. Nforce 4 Ultra will have firewall options - something that Nvidia is pushing hard.
The third chipset, Nforce 4 SLI, will, as the name suggests have SLI support and will finally offer a dual slot PCI Express motherboard where you can potentially use two graphic cards. Two Nvidia cards, as SLI is Nvidia only at this time, or should we say ex-3DFX.
The last, the Nforce 4 Pro, will have support for two Opteron CPUs and will be used for workstation and server market and AMD is definitely picking up on this market segment.
The first silicon should be ready as we speak and manufacturers are playing with them in Taiwan.
You can find more details here.
What we can add is that at least some of these versions are expected before Yule 2004. µ
[edit translated web page]
Détails! 19-08-2004 10:47:49 - Samuel D. Whereas the chipset successor of the nForce 3 at nVidia would owe s'appeller nForce 4 logically, one speaks yet currently only about his code name, the CK8-04. A Russian source wishing to keep l'anonymat forwarded to us of information on this next chipset. One learn qu'nVidia thus will not announce one, but four versions of the nForce4, of which here some details: nForce 4: Standard version of the chipset, which is intended to equip the GCV d'entrée with range and which will allow the construction of cheap mother chart without sacrificing the performances nForce 4 Ultra: Version improved of the precedent having additional functionalities. It will support the SATA-II (3 GBit/s) as well as advanced options of firewall. Its price will remain reasonable nForce 4 SLI: As one can s'en doubt, this chipset intended for the workstations and the fortunate gamers will be the only able one to support mode SLI of new charts NCV Express train of nVidia nForce 4 Pro: Intended for the waiters and thus for solutions Bi-Opteron, the nForce 4 Pro will be thus the first chipset of nVidia only intended for this market. It will be capable to support a controller NCV additional Hypertransport Express train of nVidia baptized IO4, but c'est another history... In short, the new awaited range trés of chipset nVidia seems promising. With current l'heure, the final first samples start to be available in the laboratories of the manufacturers, l'annonce official is currently planned for the second quizaine of September... Wait & See.
[edit translated web page]
There they go. Off to a blazing start. I wonder if AMD will have their dual core processors readied by XMAS?
Comments
The last, the Nforce 4 Pro, will have support for two Opteron CPUs and will be used for workstation and server market and AMD is definitely picking up on this market segment.
Yummy. Even better when dual core Optys come down the pipe in 2005.
Not a lot of info from the register article. I'd expect to see dual gigabit on the boards like they have today. Maybe some built in wireless features if they want to get aggressive. Faster Hyper transport links. I still need confirmation about the SLI. I've heard that it takes dual 16x slots and I've also heard that you can do 16x and 8x slots for the SLI configs. I'm hoping it's the latter because that gives mobo designers more flexibility.
I doubt AMD has their DC ready that early. Their press releases seem to be more about mindshare right now. I'm guessing they have DC Opterons by March-April of 2005 with PowerPC G5 DC systems shipping late summer.
the nForce motherboards have been nice for Nvidia but they need something special to make'em stand out a bit. The competition is pretty tough with Via making nice chipsets as well. It'll take a bit more than SLI to get sales popping right.
Of course PowerPCs don't utilize Nvidia chipsets aimed at x86 but I think we'll see some suprises in the next major Powermac Motherboard architecture. I'm hoping for Dual Gigabit in the whole Powermac line built in wireless, better audio I/O and more FW busses. PCI Express of course is mandatory. I think we get this in January with a small refresh of the PM lineup. SATA II should be there as well. I'm eager to see how these new NCQ enabled SATA drives do with 16MB buffers. They kick arse right now in multitasking benchmarks.
Apple needs a boost to their performance image in the PC world, and using something as familiar, and significant as a NForce4 would certainly give the appearance of more balanced graphics, and have a subtle, but leveling image of cross-platform equality should they choose to do so.
Tigers demo at WWDC had much talk about leveraging the GPU with Core Image, and one would expect if Apple were to continue with their GPU leveraging demonstrations, and discussions they would also be addressing the OpenGL Firmware, and driver issues they have been plagued with, and succumbed to vs. their x86 adversaries.
I had more to say, but I suddenly just got too busy. Sorry if that sounds incomplete. But it is.
Apple is going to move to PCI Express it's just a matter of when.
The Nforce chipset is aimed at X86. There really isn't a way to license it and even if that was possible Apple wouldn't do it because their current chipset is already more sophisticated than the current Nforce3 mobos.
We've had faster hypertransport links, Dual Sockets, the U3 controller and PCI-X for over a year now.
Now it's time to move to
PCI Express
SATA II
More FW busses 2x of FW 400 and 800
Improved Controllers.
Dual Gigabit
There is just not enough info on the Nforce4 to lead me to believe it is better than what Apple/IBM can design themselves.
Originally posted by hmurchison
their current chipset is already more sophisticated than the current Nforce3 mobos.
I was only referencing the Nforce4 designs which are probably going to out shine what Apple has in the pipe currently for implementation in January at MWSF.
In brief the Nforce 4 will support the SATA-II (3 GBit/s) as well as advanced options of firewall, Dual Processors, and Dual PCI-E.
Chances off Apple being on top of all the above, and undisclosed features are slim, but I suppose we will see.
But as you were saying -
Of course they can use the same technologies but. My point in licensing the Nforce branding was merely for name recognition, and improved hardware, (driver/firmware) card compatibility.
If Apple were to use the same features, and also include what they intend to use of their own.
Rather than comparing them as going head to head vs. the x86 - showing that they are playing with the x86 plus adding their own features could be beneficial on many levels.
If they brought in the recognized technology of Nvidia Nforce motherboards, and add Mac compatibility to them to increase their performance Instead of comparing differences in performance they would do better showing design similarities, identical parts, and identical performance out of the Nforce4 running under OS X Panther.
After that - then continue on past the licensed board, and into the Apple additions such as FireWire (which they can get via PCI card anyway), but continue with improved statistics leveraging the GPU with Tiger, and other additions with Core Image. I think you get my point so I'll stop explaining it here.<--[edit 5 hours later after re-reading] I got lost in my own mind while writing all that, but I'm not changing it because I think it's funny.
Next to the Processor - the Motherboard, and Graphics cards are what separates PC, and Mac boxes apart from each other in the minds of most. Appearing to bring in 2/3 of that equation, and supplying great performance out of your existing processor would be a huge step in changing the way the Mac is perceived by the computing community as whole.
AFAIK from what people have told me is that OS X has never been a problem for PC users, and most are delighted by the thought of using it on a regular basis. Changing all their platform software has been a concern, and Macintosh computers themselves have been a concern also, but introducing familiar PC hardware even if it is for appearance only is still a major step, and would improve the overall appearance of the Macintosh to the majority of computer users. Even if IBM is still having CPU troubles as of now.
Plus it also adds an additional element , and brings in a closer working relationship with NVidia, and IBM.
I can toss in a Blackmagic uncompressed card in a PowerMac G5 with external RAID and capture 280+ MBps video.
PCI-X allows for this PCI does not. So when we're talking about the relative merits of the architectures it becomes clear that Powermacs, despite some of the flashy acoutrements of X86 brethren, offer features that common user doesn't need but the Pro requires in many cases.
Frankly I've always been impressed with the Powermac from an architecture standpoint. Dual Point to Point FSB, Hyper Transport links connecting some of the controllers. PCI-X and FW800. It was/is a nice mobo and I'm hoping Apple takes it to the next level soon.
I like the nForce as well. I'll base my next homebuilt off of the best AMD chipset I can get at that time. Should be fun. The waiting part is the hardest isn't it?
Dual core Opteron? Not anytime this year. It's mindshare talk. But the SLI? Sampling September and arriving in time for Christmas?
If I was an AMD buyer (and I will be...) I'd the nine months until both PPC and AMD make the transition to dual core.
Current set ups will age less gracefully in the context of what is coming.
The next level of machines from Apple and AMD (ie in 9 months time...) should really be impressive performers.
Memory, hard drive, bandwidth, cpux2 cores, dual GPU or a single GPU with 512 megs on it?
It all adds up to some scream worthy machines.
I'll keep my money until my pocket to then.
It's Antares of bust for the Lemon. I've waited 5 years to update my tower. 7 since I bought my first Mac.
I'm confident IBM can meet whatever Nvidia comes up with.
1.5 gig of bandwidth for a dual core Antares at 3 gig...my, it doesn't bare drooling over...
Lemon Bon Bon
PS. I can't wait to see some reviews of SLI set ups with 6800 chips on Doom 3 etc. Should be stunning...leaving ATI shell shocked!
Originally posted by onlooker
Apple needs a boost to their performance image in the PC world, and using something as familiar, and significant as a NForce4 would certainly give the appearance of more balanced graphics, and have a subtle, but leveling image of cross-platform equality should they choose to do so.
This is all well and good if Apple could convince either Freescale or IBM to adopt HT on the CPU chip. As yet thate has been nothing in the wind to indicate either is going this way. IBM is part of the HT team, but I've seen little indication that they have intentions to use this as I/O on a CPU.
Like wise for better or worst Freescale is tied into another bus and does not appear to be willing to support HT. It is an open question right now if either Freescale or IBM could support HT as a special IP module for a chip specific to Apple. It is unfortunate that niether Freescale nor IBM really seem to be tuned into industry standard I/O buses.
It is in the use of standard parts that Apple has opportunity to focus more of its engineering skills on innovation. Working with odd buses is not really innovation especially when there is little advantage in their use.
I had more to say, but I suddenly just got too busy. Sorry if that sounds incomplete. But it is.
Hey be happy that the work is there.
Dave
Originally posted by PBG4 Dude
IIRC, Apple is part of the HT consortium as well. I just figured the G5 buss came from their involvement with HT.
Your not alone I thought it did too.
Originally posted by PBG4 Dude
Apple is a member of the consortium; a complete membership roster can be found here --> http://www.hypertransport.org/org_members.html
Well yes which is why i thought they were one of the first if not the first to get this. I thought that was the way the first G5 was described at the keynote.
Now I'm at a loss for words. Why didn't they?
Originally posted by onlooker
Well yes which is why i thought they were one of the first if not the first to get this. I thought that was the way the first G5 was described at the keynote.
Now I'm at a loss for words. Why didn't they?
Not only is Apple a member, it is a founding member of the Hypertransport consortium!
Apple uses Hypertransport between its two main core logic chips, the system ASIC ("northbridge") and the I/O ASIC ("southbridge"). Some versions of the PowerMac G5 use a PCI/PCI-X Hypertransport bridge that is between the two.
The reason Apple can't use the Nforce 4 is that the PPC 970 uses a different memory and bus architecture than the Opteron. PPC 970 has all of its memory and I/O traffic go through its elastic bus, so a system ASIC support chip would need to have a memory controller in it. Opteron has an on-die memory controller with Hypertransport links for I/O (graphics, disks, PCI, etc.) Since memory traffic goes through the on-die memory controller, the Nforce 4 doesn't have a memory controller in it, and therefore cannot be used in a 970 system. Not to mention it doesn't support the 970 elastic bus either.
For Apple to support PCI Express and SLI, they would just have to add support for it in the Power Mac G5 system ASIC. Also update their I/O chipset for the new I/O goodies and it'll be just as good as an Nforce 4 system.
PowerMac G5 architecture diagram:
Nforce 4 architecture (CK08-4):
That is an interesting piece of information but we have yet to see any hints that IBM will be producing a chip with this as the primary interface. It is always a possibility that they may go HT once they implement on die memory controllers.
Freescale on the other hand is going another route, which personally I think is a huge mistake. It does appear that Freescale is following prety much in Motorolas footsteps. Things like Rapid I/O may be well engineered but it will never achieve the critical mass needed for cost effective implementations.
The only possible save here is that both IBM and Freescale have discussed the possibility of custom chips. Freescale prehaps has the best opportunity here to offer swappable bus IP. I just don't see them (Freescale) involved in HT much at all. Must be that NIH sickness.
Dave
Originally posted by onlooker
Your not alone I thought it did too.
Originally posted by THT
Not only is Apple a member, it is a founding member of the Hypertransport consortium!
Apple uses Hypertransport between its two main core logic chips, the system ASIC ("northbridge") and the I/O ASIC ("southbridge"). Some versions of the PowerMac G5 use a PCI/PCI-X Hypertransport bridge that is between the two.
The reason Apple can't use the Nforce 4 is that the PPC 970 uses a different memory and bus architecture than the Opteron. PPC 970 has all of its memory and I/O traffic go through its elastic bus, so a system ASIC support chip would need to have a memory controller in it. Opteron has an on-die memory controller with Hypertransport links for I/O (graphics, disks, PCI, etc.) Since memory traffic goes through the on-die memory controller, the Nforce 4 doesn't have a memory controller in it, and therefore cannot be used in a 970 system. Not to mention it doesn't support the 970 elastic bus either.
For Apple to support PCI Express and SLI, they would just have to add support for it in the Power Mac G5 system ASIC. Also update their I/O chipset for the new I/O goodies and it'll be just as good as an Nforce 4 system.
(1) "Apple uses Hypertransport between its two main core logic chips, the system ASIC ("northbridge") and the I/O ASIC ("southbridge"). Some versions of the PowerMac G5 use a PCI/PCI-X Hypertransport bridge that is between the two."
(2) "For Apple to support PCI Express and SLI, they would just have to add support for it in the Power Mac G5 system ASIC. Also update their I/O chipset for the new I/O goodies and it'll be just as good as an Nforce 4 system."
It almost looks like you contradict your self, but I have no idea what this sh*t is anyway - so please take no offense.
So basically can we conclude that the Apple motherboard design has little, or a big change in store for it to get PCI-E, SLI? Is this a huge design overhaul, or minor adjustments, or replacing to a few smaller components to get the graphics on par with what many expect from them?
Originally posted by onlooker
[B]So basically can we conclude that the Apple motherboard design has little, or a big change in store for it to get PCI-E, SLI? Is this a huge design overhaul, or minor adjustments, or replacing to a few smaller components to get the graphics on par with what many expect from them?
Apple would have to replace the 8x AGP Pro graphics bus in the U3H system ASIC with two x16 PCI Express channels and the Nvidia SLI logic between them. That's it. One would hope they would space the slots "double-wide" as well, to account for the extra thickness in modern graphics cards. In the grand scheme of things, that's a small effort. The architecture they have in the Power Mac G5 is good for 4, maybe 5 years.
The SATA-2 support is even less of a deal. That's just upgrading the SATA support in the I/O ASIC (K2). They already got 3 different versions of the I/O ASIC for PowerMacs, Xserve and iMac G5, so they make small changes all the time.
The PCIe dual slot with SLI support will trickle into a slight form factor change too. The Power Mac G5 case will likely have to grow if they want to have 2 PCIe slots that occupy 4 slots of space plus 3 PCI/PCI-X slots. That'll be about 1.5 inches more in height.
Now whether they choose to do it is a different thing all together. They've got bean counters counting every dollar. So who knows.