Could Sun Microsystems use the G5 too?
Sun Microsystems has similar problems as Apple if you look at the processor technology. They're stuck at about 500 mhz, although their UltraSparc II/III cpus are already 64 bit, and they offer about the same performance/cycle level as the G3/G4 chips, partly even better.
Nevertheless - an UltraSparc III @ 500 mhz can no longer compete with a 2.2 ghz P4, even if this is "only" a 32 bit cpu. The UltraSparc III already came late, and it looks like SPARC isn't evolving much more in the last few years. SPARC for iA64 was planned, but it seems that neither Sun nor Intel are any more interested in this project, and in fact the Itanium isn't really faster than their current UltraSparc III machines.
I always though that sooner or later Sun would start to use x86 processors in their machines, starting with the entry level machines. But then I read that Sun plans to no longer continue the development of Solaris 9 for x86 (The Register had an article about this, most other computer news sites too).
Now here are the facts:
- Sun is developing Solaris 9, Sun does (definately) not plan to use an M$ OS
- Sun makes fast workstations and servers, so they will need fast processors
- Sun is developing Solaris9 for Sparc, but Sparc is stuck, it cannot compete with upcoming x86 or iA64 processors
- Sun has stopped development of Solaris/x86
- Sun has stopped development of Solaris/iA64
So what? There must be a processor they will use in the future, or not? Then I found something @ <a href="http://www.sun.com,"; target="_blank">www.sun.com,</a> a small article about a project called "Chameleon", very similar to Apple's "Star Trek". In fact Sun planned to release a Solaris edition for PowerPC in 1995 (yes, it's an older article), but this sounds more like a switchover plan to PPC (scroll down to read the article).
So here is some speculation: could it be that Sun has dropped all development for current and future Intel processors, and this while their own Sparc architecture is not progressing any more, because they have found another CPU they would like to support?
If yes, which architecture could be interesting for Sun? MIPS is not in a better position that Sparc, Alpha is dead. Maybe the PowerPC? Maybe the G5?
I know, very speculative... here's the article:
Chameleon
Robert Yung yung@sun.com
Dave Roberts droberts@sun.com
Neil Wilhelm wilhelm@sun.com
Overall objective:
To demonstrate the feasibility of a processor-independent Sun system. To provide a platform for heterogeneous multi-processor (MP) system research. To provide a platform for binary translation research.
Objective for FY95
Chameleon was started in the second half of FY95. The Chameleon project has three objectives: Build a PowerPC 604-based MBus card for the "PS-20" system, the chassis of the SPARCstation 20 product. Develop a P1275-compliant OpenBoot Prom 3.0 Implementation. Port a big-endian PPC Solaris to the "PS-20" system.
Description
Chameleon provides the hardware and software components of a non-SPARC based Sun workstation prototype that runs Solaris.
The hardware components of Chameleon are an MBus card containing a PowerPC (PPC) 604 and a IEEE P1275-compliant OpenBoot PROM (OBP) that can be used in an SS20 system to build a non-SPARC Sun workstation. The PowerPC (PPC) MBus card contains a PPC 604 CPU, an Altera PLD that controls the CPU, the external cache and Mbus interface, bus drivers, a PLL clock driver, and two sets of E-cache tag memories. The design is optimized for MP applications: both the internal data cache and the external unified E-cache are configured as write-back. This minimizes the peak bandwidth requirement at the E-cache and Mbus interfaces, and allows usage of slower E-cache SRAMs. Mbus snooping is filtered with two sets of E-cache tags: one for CPU accesses and the second for MBus snooping.
The software components of Chameleon are a big-endian Solaris port to this platform and a native C-compiler. The big-endian PPC Solaris port is descended from the little-endian PPC port SunSoft has done to the PReP platforms. The PReP I/O system is replaced by the Sun4m I/O system of the SS-20 platform.
Accomplishments
The PPC 604-based MBus card is complete. The prototype MBus card is fully functional and works at speed (40Mhz and 50 MHz MBus). The CPU can run at 2X, 3X, or 4X the Mbus frequency. The port of a P1275 OBP (version 3.0) is mostly complete. The big-endian PPC Solaris port is complete, with basic Sun4m driver integration.
Lessons Learned
There are several things we learned in the course of designing the prototype MBus card. In particular, rapid prototyping of full-speed, full-function processor cards is possible with off-the-shelf parts. The easy-to-use MBus and the generic address/data interface of the PPC 604 were very helpful. Some specifics are: The entire design was done in 3-person-months. All components used on the Mbus card except the CPU are off-the-shelf. A PLD/FPGA (plus discrete buffers) offers a realistic alternative to full-custom or gate-array processor interface chips. A single Altera PLD is used to control CPU interface, the external cache and the Mbus interface. This PLD (plus external parts) provides the same functionality as the MXCC in a SPARC Mbus module.
Nevertheless - an UltraSparc III @ 500 mhz can no longer compete with a 2.2 ghz P4, even if this is "only" a 32 bit cpu. The UltraSparc III already came late, and it looks like SPARC isn't evolving much more in the last few years. SPARC for iA64 was planned, but it seems that neither Sun nor Intel are any more interested in this project, and in fact the Itanium isn't really faster than their current UltraSparc III machines.
I always though that sooner or later Sun would start to use x86 processors in their machines, starting with the entry level machines. But then I read that Sun plans to no longer continue the development of Solaris 9 for x86 (The Register had an article about this, most other computer news sites too).
Now here are the facts:
- Sun is developing Solaris 9, Sun does (definately) not plan to use an M$ OS
- Sun makes fast workstations and servers, so they will need fast processors
- Sun is developing Solaris9 for Sparc, but Sparc is stuck, it cannot compete with upcoming x86 or iA64 processors
- Sun has stopped development of Solaris/x86
- Sun has stopped development of Solaris/iA64
So what? There must be a processor they will use in the future, or not? Then I found something @ <a href="http://www.sun.com,"; target="_blank">www.sun.com,</a> a small article about a project called "Chameleon", very similar to Apple's "Star Trek". In fact Sun planned to release a Solaris edition for PowerPC in 1995 (yes, it's an older article), but this sounds more like a switchover plan to PPC (scroll down to read the article).
So here is some speculation: could it be that Sun has dropped all development for current and future Intel processors, and this while their own Sparc architecture is not progressing any more, because they have found another CPU they would like to support?
If yes, which architecture could be interesting for Sun? MIPS is not in a better position that Sparc, Alpha is dead. Maybe the PowerPC? Maybe the G5?
I know, very speculative... here's the article:
Chameleon
Robert Yung yung@sun.com
Dave Roberts droberts@sun.com
Neil Wilhelm wilhelm@sun.com
Overall objective:
To demonstrate the feasibility of a processor-independent Sun system. To provide a platform for heterogeneous multi-processor (MP) system research. To provide a platform for binary translation research.
Objective for FY95
Chameleon was started in the second half of FY95. The Chameleon project has three objectives: Build a PowerPC 604-based MBus card for the "PS-20" system, the chassis of the SPARCstation 20 product. Develop a P1275-compliant OpenBoot Prom 3.0 Implementation. Port a big-endian PPC Solaris to the "PS-20" system.
Description
Chameleon provides the hardware and software components of a non-SPARC based Sun workstation prototype that runs Solaris.
The hardware components of Chameleon are an MBus card containing a PowerPC (PPC) 604 and a IEEE P1275-compliant OpenBoot PROM (OBP) that can be used in an SS20 system to build a non-SPARC Sun workstation. The PowerPC (PPC) MBus card contains a PPC 604 CPU, an Altera PLD that controls the CPU, the external cache and Mbus interface, bus drivers, a PLL clock driver, and two sets of E-cache tag memories. The design is optimized for MP applications: both the internal data cache and the external unified E-cache are configured as write-back. This minimizes the peak bandwidth requirement at the E-cache and Mbus interfaces, and allows usage of slower E-cache SRAMs. Mbus snooping is filtered with two sets of E-cache tags: one for CPU accesses and the second for MBus snooping.
The software components of Chameleon are a big-endian Solaris port to this platform and a native C-compiler. The big-endian PPC Solaris port is descended from the little-endian PPC port SunSoft has done to the PReP platforms. The PReP I/O system is replaced by the Sun4m I/O system of the SS-20 platform.
Accomplishments
The PPC 604-based MBus card is complete. The prototype MBus card is fully functional and works at speed (40Mhz and 50 MHz MBus). The CPU can run at 2X, 3X, or 4X the Mbus frequency. The port of a P1275 OBP (version 3.0) is mostly complete. The big-endian PPC Solaris port is complete, with basic Sun4m driver integration.
Lessons Learned
There are several things we learned in the course of designing the prototype MBus card. In particular, rapid prototyping of full-speed, full-function processor cards is possible with off-the-shelf parts. The easy-to-use MBus and the generic address/data interface of the PPC 604 were very helpful. Some specifics are: The entire design was done in 3-person-months. All components used on the Mbus card except the CPU are off-the-shelf. A PLD/FPGA (plus discrete buffers) offers a realistic alternative to full-custom or gate-array processor interface chips. A single Altera PLD is used to control CPU interface, the external cache and the Mbus interface. This PLD (plus external parts) provides the same functionality as the MXCC in a SPARC Mbus module.
Comments
I really doubt that Sun would transition over to PPC now.
We know that IBM has a number of PPC-related projects stretching off through the next several years, such as the deal to produce future Playstation chips in conjunction with Toshiba and Sony. I can't think of any comparable details for SPARC, and I haven't heard anything about MAJC in ages.
MIPS seems to be dead, and we all know about Alpha.
If Sun moved to PPC, I'm pretty sure that there would be no attempt to revive CHRP and have standard PPC boxen running multiple unix OSs. But it would be a good thing all round if (say) Codewarrior could automatically produce OSX- and Solaris-compatible binaries.
sun's engineers are quite good, I doubt Sun will have problems go over 1ghz if they really, really wanted.
G-News
- Sun is the only user of Sparc, which means they alone are footing an enormous development bill. If they could share the bill they would be more profitable.
- The fast Sparcs just arrived recently and they are still at G4 clock rates. Eventually Sun will be able to push the clock rates higher, but its not any easier for them than it is for Motorola.
- If the G5 does exist, and it is 64-bit as rumoured, and it is fully MP capable (like the 7400 and 604e before it), HT or RapidIO, and it has a brand new core (as rumoured and documented by the eCore stuff and roadmaps)... then it has the potential to knock the socks off of the UltraSparc. Sun does care about single processor performance... look at the benchmark war they are having with IBM's POWER4 family.
- Apply all the "G5 does exist and will be fast" arguments here. See other threads for details.
- Apple would also want a desktop processor development partner, and having a workstation company with experience building fast machines would be a bonus.
- AltiVec is a better SIMD architecture and is already proving useful to network hardware companies like Cisco.
Not a completely outlandish idea. Personally I would like to see all the workstation companies that are using their own RISC processors jump on Apple's PPC bandwagon -- this would make them all a lot stronger and able to face the x86 market juggernaught. It would certainly strengthen the desktop PowerPC's future.
yes but not ad infinitum. obviously sun won't like mac os x, you'll see why in a year or so. it's not so much a threat to windows but also and foremost to unix vendors.
scott mcnealy's ego is bigger than the empire state building. but then again that maybe a good thing
oh sorry i forgot, the other one is the enemy, not scott...
<strong>last time I looked sun had chips running at 750-900mhz (750- or 900-MHz UltraSPARC-III). and they're 64bit already.
sun's engineers are quite good, I doubt Sun will have problems go over 1ghz if they really, really wanted.</strong><hr></blockquote>
We have already gone over 1Ghz with the new UltraSparc III Cu 1050 processor.
<a href="http://www.sun.com/smi/Press/sunflash/2001-11/sunflash.20011119.2.html"; target="_blank">http://www.sun.com/smi/Press/sunflash/2001-11/sunflash.20011119.2.html</a>;
BTW, Sun is *not* the only company that uses Sparc, Fujitsu does also. Cray had some projects internally which used Sparc, in the past.
-Moazam
<strong>Not a completely outlandish idea. Personally I would like to see all the workstation companies that are using their own RISC processors jump on Apple's PPC bandwagon -- this would make them all a lot stronger and able to face the x86 market juggernaught. It would certainly strengthen the desktop PowerPC's future.</strong><hr></blockquote>
The idea isn't bad by itself, but a lot of the high-end server/workstaion companies are forced into niches and where some want past single cpu, others want scalable multi-cpu environment, some don't care about heat and power drain others want laptops. And some have customers who spend millions whereas others have customers who spend $ 1800 and complain about the small LCD screen they got with that.
Besides, who'd want to put their faith into Steve's let-form-the-AIM-alliance hands. Rather make your own chip and know who to blame than be pissed off at Motorolla.
Apple just has to do some tweeks on the current motherboards and BABOOM. G5
[ 01-10-2002: Message edited by: Scott H. ]</p>
WHY did you let her die? WHY?
hehe
<strong>Yes they can use G5's, everyone can use G5 and everybody will want to use G5's because they are going to be Bloody fast. Apple just has to do some tweeks on the current motherboards and BABOOM. G5
Oh, and they have to make the G5.
The benchmark war between Suns Sparc and IBM's Power4 is so far based on the performance of 2 supercomputers: a 72 way Sparc using 900MHz chips and a 24 way Power4 using 700MHz or so chips, don't hang me on the details. Anyway, the Power4 beats teh Sparc machine, despite the singificantly lower CPU count. Sparc is fast, several times faster than any PPC chip ever used in a Mac, even at low speeds, but the Power4 is one of the worlds most powerful, if not THE most powerful chip to date. I seriously doubt the G5 will beat that, thus it's unlikely that Sun is going to adopt it INSTEAD of teh Sparc, maybe alongside it, but not instead.
G-news
<strong>If Sun decides to use the G5 I hope they make Solaris mac-compatible so I can have Solaris on one partition and OSX on another on my upcoming PowerMac G5. That would be cool!!</strong><hr></blockquote>
I find that odd. I never thought to use a Sun so that I could use Solaris. It seems to be people use Sun because of the hardware. Solaris is just tolerated.
I think an Apple/Sun merger would not be possible because of the egos of Steve an Scott (who would be the Apple/Sun CEO?).
I don't think Sun is afraid of MacOS X. Apple doesn't even offer rack mount servers, and we don't know how well OS X can be scaled (okay, it will run on MP systems, but has someone tested OS X on a 256 processor machine?). I would also say that Solaris is currently more stable than OS X (although this has nothing to do with the Darwin/UNIX layer, it's the user interface that sometimes crashes - ok, not often). Sun has a marketshare in design, construction and scientific areas, and I'm not sure Apple will be able to enter these markets (or is even interested in those).
On the other hand Apple seems to be very aggressive in the video production area. Therefore I don't believe Apple will buy SGI (or merge with them), as they are just trying to take over this market (but they'll need faster machines to do so).
Running Solaris on the PowerMac would be great, and it was even planned in 1994/1995, but never completed.
What else... oh yeah, did you know that Sun once planned to team with Commodore? They wanted to sell Amigas as consumer machines and Sparcs as pro models (Commodore was not interested, big mistake). Couldn't both companies help each other (both CEOs are M$ haters as you know, maybe this would be some motivation)?
I'm asking myself if it would be possible to make OSX and Solaris/PPC binary compatible. OS X is based on BSD while Solaris is based on UNIX (no, BSD and UNIX are not exactly the same).
<strong>Some very interesting replies, thanks.
I think an Apple/Sun merger would not be possible because of the egos of Steve an Scott (who would be the Apple/Sun CEO?).
I don't think Sun is afraid of MacOS X. Apple doesn't even offer rack mount servers, and we don't know how well OS X can be scaled (okay, it will run on MP systems, but has someone tested OS X on a 256 processor machine?). I would also say that Solaris is currently more stable than OS X (although this has nothing to do with the Darwin/UNIX layer, it's the user interface that sometimes crashes - ok, not often). Sun has a marketshare in design, construction and scientific areas, and I'm not sure Apple will be able to enter these markets (or is even interested in those).
On the other hand Apple seems to be very aggressive in the video production area. Therefore I don't believe Apple will buy SGI (or merge with them), as they are just trying to take over this market (but they'll need faster machines to do so).
Running Solaris on the PowerMac would be great, and it was even planned in 1994/1995, but never completed.
What else... oh yeah, did you know that Sun once planned to team with Commodore? They wanted to sell Amigas as consumer machines and Sparcs as pro models (Commodore was not interested, big mistake). Couldn't both companies help each other (both CEOs are M$ haters as you know, maybe this would be some motivation)?
I'm asking myself if it would be possible to make OSX and Solaris/PPC binary compatible. OS X is based on BSD while Solaris is based on UNIX (no, BSD and UNIX are not exactly the same).</strong><hr></blockquote>
It will be called SNAPPLE = Sun + Apple. OS X on the desktop and Sun Solaris on the Server. Simple.
Since Fujistu uses the SPARC, Sun clearly already has a partner... but who knows, maybe Fujistu isn't the best partner or perhaps the don't want to continue with SPARC either. This is all speculation, after all.