i remember that motorolas roadmap spoke of dual core g4's less than 25 watts at 2 ghz.
that would be amazing......
it would have to be REAL good for apple to swithc back to moto.
by the way im a raider fan and feel we have a pretty good draft.
i cant wait to see what those receivers can do.....
I don't know what they are because I don't know anyone over at FreeScale/Moto - just IBM. Just seems they are implying something is coming. They've publicly released documents indicating that the 7400 will go beyond 2GHz - and I think the memory controller may be on the map, as well as rapid i/o. Not too sure about hypertransport.
I'm a big time Browns fan - and I liked our draft, even with giving up the 2nd rounder. Looking forward to seeing Winslow going in under coverage.
They've publicly released documents indicating that the 7400 will go beyond 2GHz - and I think the memory controller may be on the map, as well as rapid i/o.
Hey ass, do you have any idea who Pat Tillman is? He played for the Cardinals, but after Sept 11 he quit football, giving up his huge contract, to join the military. He recently died over in Afganistan. Show some respect.
Hey ass, do you have any idea who Pat Tillman is? He played for the Cardinals, but after Sept 11 he quit football, giving up his huge contract, to join the military. He recently died over in Afganistan. Show some respect.
My appologies, yes I did hear the story, since we were talking about draft picks, I thought there was some other Pat that you were making reference to, as in he was drafted over the weekend.
Going back and reading the posts in reference, it makes more sense now. I am not good with name recognition, as a matter of fact, I suck at it. So, I was not being an Ass, I just didn't recognize the name. I am sorry if you think I have no repect for what he did, I do very much so. I myself am a veteran of the Navy.
Rag on MaxBus all you want, but its real bandwidth reached 80-90% of theoretical on a sustained basis. Most PC busses (and the 60x bus) are lucky to reach 50-60%. That's efficiency.
The G4's caches are incredibly aggressive. The core is lean (if anything, it's a bit too lean). The 745x AltiVec unit is at the top of the SIMD heap.
Put a fast bus on a 7457, and you'll have something that will give you incredibly high efficiency and very high performance, on a power budget that would stall a 970fx.
I know it's fashionable to trash the G4 now because it's not the best chip to put in a $3000 tower (gee, whodathunkit?) but you're way, way, off here. It's a nice design, and we haven't seen the last of it.
... I know it's fashionable to trash the G4 now because it's not the best chip to put in a $3000 tower (gee, whodathunkit?) but you're way, way, off here. It's a nice design, and we haven't seen the last of it.
I would say that it is fashionable to trash the G4 because of Motorolla's history with the chip as well as a rather slow moving roadmap for future advances on the archetecture. This may change with the new management and structure and partnerships for the company, but I don't think that is evedent based on "real" information that we have today, such as speeds on current shipping 7457 chips. Powerlogix is shipping 1.33 Ghz processor upgrades, and did ship them at speeds of up to 1.4 Ghz, but Moto was already shipping 1.4 Ghz 7455's so the new processor had a net speed gain of 0. Now, 18 months after the introduction of Apple's 1.42 Ghz PowerMac we are finally seeing the G4 (7447) break that speed, and by a whopping 100 mhz. It is hard to be enthusiastic about the company based on recent history.
Product transitions are always a pain in the neck for any IT vendor. As soon as they know a kicker product is imminent, all customers, except those desperate to buy a new piece of hardware or software, sit tight, because they want to see what the new product will be, and how it will be priced. This makes perfect sense to IBM midrange shops that may have one, two, or even three decades of experience in acquiring computers.
If this is a problem for customers, it is, in many ways, a much more troublesome situation for IBM Corp. During the past several decades, about the only way to get System/3X, AS/400, or iSeries customers to buy ahead of a future product launch has been to cut prices, and this is exactly what IBM did last week with the iSeries line.
IBM learned a long time ago, way back when it was first selling System/360 computers, and even earlier with punch card tabulating equipment, that there is almost a perfectly elastic relationship between the price of an IT product and the ability of companies to absorb more and more capacity. However, IBM's goal has never been the proliferation of a technology, so much as the maintenance of the absolute maximum profits that can be wrung out of what amounts to captive customer bases locked into products by their legacy applications. IBM is all about profits, but it has to maintain a revenue stream to keep its top brass, sales reps, partner channel, and Wall Street happy. With iSeries sales down 7% in the first quarter and new Power5-based iSeries machines (or whatever they will be called) not expected until later in the second quarter, IBM had to do something to stimulate sales until the Power5 "Squadron" machines arrive on the scene, sometime before the end of the second quarter.
It is interesting to see what IBM did. First, it cut the list prices on base configurations of Model 810 servers (including iSeries for High Availability models) by between 9% and 35%. IBM also cut main memory prices on the selected Model 270s and on all Model 810 servers by 40%. The Model 810 machines use many of the same main memory cards that are used in the Model 270s, so it did not make much sense to not make the 40% price cut effective on these Model 270s that share at least some memory cards with the Model 810s. After the price cuts, Model 810 and Model 270 main memory prices range from $1.89 per MB, on relatively dense cards with 512MB or 1GB capacities, up to $4.32 per MB on cards with only 128MB of capacity. And I wouldn't be me if I didn't write the following sentence: While these price cuts are good, IBM is still charging about twice as much for ECC main memory on entry iSeries servers as it does for its own xSeries line and its competitors do for their X86-based servers.
IBM also cut list prices on the base configurations of Model 870 (16-way capable) and Model 890 (32-way capable) iSeries servers. Sources at IBM say that, generally speaking, list prices on these base boxes were dropped by approximately 20%. The actual price cuts vary on the processor performance and the OS/400 edition (Standard Edition with little 5250 green-screen capability or Enterprise Edition with fully enabled 5250 processing). Memory prices on the Model 870 and Model 890 were chopped by 60%. You heard that right. This brings iSeries memory, which can account for 25% of the cost of a configured system, more in line with other high-end SMP servers available from Big Blue and others. I said "more in line," not actually "in line." While IBM has cut prices from $12.15 to $4.86 per MB on the Model 870 and the Model 890, that is still about twice the cost (at list price), compared with ECC main memory for high-end Unix servers. Then again, big Unix boxes are sold at 35% to 50% of list these days, so the disparity at the street price level is larger than it might otherwise seem.
IBM cut prices on 10K RPM and 15K RPM SCSI disk drives used across the iSeries product line by between 20% and 50%, depending on the disk capacity and rotational speed. The price cuts are roughly in line with the disk price cuts that IBM announced almost exactly two years ago. After the cuts, IBM is charging $1,000, or 6 cents a MB, for a 17.5GB, 10K RPM disk. A 35.2GB disk spinning at 10K costs $1,359, or 4 cents per MB, while a 15K version of the 35.2GB disk costs $1,875, or 5 cents per MB. IBM's fattest 70.6GB, 15K RPM disk now costs $2,875, or 4 cents per MB.
What IBM did not do, you might notice, is cut prices on the entry Model 800 or on the midrange Model 825. This is the best indication that IBM does not intend to dramatically improve the price/performance of machines in this class with the impending Power5-based Squadron servers. What it probably means is that IBM will launch a two-way and four-way capable Squadron box that competes against the Model 810 and which will have similar bang for the buck. It also seems to imply that IBM will not immediately ship an eight-way Squadron box, or if it does, it will not offer much better price/performance than the current Power4-based Model 825 does.
That doesn't mean that Model 800 component prices have not changed. The main memory in the Model 800 had its price cut by 40%, and it uses the same disk drives that had their prices cut. Because IBM wants to keep the prices of the Model 800 machines constant, it actually raised the price of the base Model 800 servers to offset the cuts on the memory and disk components included in the base Model 800 packages. This is IBM and its iSeries channel trying to maintain a profit margin on a base box. There is no other explanation as to why IBM would not let the price on the entry iSeries Model 800 drop below $8,795. Obviously, adding more memory and disks to a Model 800 above and beyond the base configuration is cheaper than it was before the price cut, so it is not all bad. But it would have probably helped Model 800 shipments considerably if IBM actually cut the prices on these boxes by 25% to 30%. The resellers we have talked to in recent weeks say that the Model 800 is not selling well, and there is no surprise to that. It is too expensive, and especially so compared to second-hand Model 250 and Model 270 servers.
There is one more price change that Big Blue announced last week that might show how IBM will be positioning the future Power5 machines. On the Model 825, Model 870, and Model 890 servers, IBM lowered the price of permanently activating excess processors in a box by $15,000 while raising the price of activating an OS/400 license on those processors. Big deal, right? It's a net wash to activate OS/400 on a processor, so what is there to be excited about. Well, don't forget about Linux on the Power4 machines and Linux and AIX on the Power5 machines. Before this price change, it cost $20,000 to activate a 1.1GHz Power4 processor beyond the base three processors that come with a Model 825; after the cut, it only costs $5,000. Given that a single Power4 processor can do about twice the work of the fastest Xeon processor, this is the first time that the extra processors in the Model 825, which scales to six processors, are in price parity with the X86 Xeon and Opteron alternatives in the server market.
You heard that right - so now, adding a few extra processors to a Model 825 and slapping on 64-bit Linux on partitions running on these extra processors makes economic and technical sense, and now IBM can really make a credible argument for moving Web infrastructure workloads back onto the iSeries and inside its logical partitions.
The processor activation on the Model 870s and Model 890s is not nearly as attractive, which is a shame. Before the cuts, it cost $25,000 to activate a processor on a Model 870 (there are two versions of this machine, one that scales from 5 to 8 processors and another that scales from 8 to 16 processors); now it costs $10,000. This Model 870 uses a 1.3GHz Power4 processor, and why IBM thinks it is worth twice as much as a 1.1GHz processor is unknown. It just isn't. The price of activating a 1.3GHz Power4 processor on a high-end Model 890 used to cost $30,000, and now it costs $15,000. While it would have been great if IBM had standardized pricing across the iSeries line in such a way that bears some resemblance to performance - a 1.1GHz Power4 processor costs $5,000, and a 1.3GHz processor costs $6,000 - these lower prices are nonetheless a good thing. I also think that IBM will package the Squadron machines in a similar way, such that it is relatively inexpensive to add Linux or AIX, but OS/400 processors will cost about what they do now. IBM wants to encourage server consolidation on the iSeries, but it wants to rake in those profits on OS/400.
IBM Corp stopped offering upgrades from AS/400 Model 720, 730, and 740 servers to the iSeries 8XX machines in October 2003, in the hope that doing so would convince customers to upgrade before the end of the third quarter of 2003. The lack of upgrades probably did help IBM sell iSeries boxes in that quarter, but the lack of upgrades from 7XX to 8XX models has hurt sales since then.
That's why IBM instituted a promotion last week aimed at Model 7XX and is giving decent trade-in rebates for upgrading to an iSeries Model 810 or Model 825 server running OS/400 Enterprise Edition. These are not traditional upgrades, where IBM protects the customer's depreciation schedule on an existing machine by preserving that machine's serial number after the upgrade. This is a push-pull deal, where a 7XX customer will acquire a Model 810 or Model 825 and then ship back the 7XX box to IBM. Big Blue will then do as it pleases with the 7XX box, perhaps reselling it through its Global Financing unit or putting in the metal recycling bin to keep it off the open market.
Last week IBM dropped prices substantially on the more powerful versions of the Model 810 server (price cuts were above 60%), and even the entry box (Model 810-2465) got a 9% discount. When you throw the 7XX upgrade promotion on top of these Model 810 price cuts, the effective price of a Model 810 for a 7XX shop is about half, compared with acquiring a Model 810 before April 20. On the smaller Model 810 box, the discount and the rebate amounts to a 27% discount.
IBM did not cut prices on the Model 825 server, but it is offering a $50,000 trade-in credit to 7XX customers who trade up to a Model 825. This works out to a 15% discount off the acquisition costs.
Other discounts and deals can be combined with the 7XX to 8XX trade-in promotion, which will run until June 30. That is the end of IBM's second quarter, and it is also the date by which IBM has promised it will start shipping its Power5-based "Squadron" servers. They could, of course, ship earlier. Resellers and other business partners cannot upgrade their machines as part of this deal. It is intended for end user customers only. IBM is not offering trade-ins on Model 810 and Model 825 servers running OS/400 Standard Edition, and it similarly did not cut prices on any iSeries machines running Standard Edition on April 20.
What I'm wondering is if Apple would implement the extended addressing capability already in the G4? This would give them a laptop capable of handling processes with a full 4 gigs of memory. Sure you are still talking about 32 bit processes but for many applications this would more than double the real memory they could address. You could very easily have a laptop that would carry around 8 gigs of real memory.
With the current version of OS/X this should be very easy to implement. After all, all OS/X does is handle 32 bit processes in a larger address space. Granted the G4's ability to quickly get to all that memory is an issue with the current front side bus but I have to imagine that a solution for that has to be in the works.
Dave
Quote:
Originally posted by Matsu
I wonder if moto/freescale will have a 64 bit variant suitable for laptop use. It isn't critical at this time. There ain't no 64 bit "mobile" chips from intel or AMD. There are 64 bit "luggables" based on AMD chips but nothing we'd call a true "laptop".
I ask because if there is something with that huge address space, we might never see an IBM based Powerbook at all, what would be the point, so long as the moto offering is small/cool/frugal, and offers industry standard level performance. If there's no need for 64 bits, or there's a way to cover that need with a more mobile friendly chip, then there's no need for IBM in a laptop.
If, however, there is a pressing need for a mobile professional 64 bit platform, and no small/cool solution, I do see that we may have a third tire of Apple laptops added.
iBook G4
Powerbook G4+ ?
Powerbook G5
All to exist simultaneously, 32 bit machines with an emphasis on battery life, 64 bit machines with an emphasis on number crunching.
It is interesting your focus on the bus efficiency, it reminds me of American management of production facilities. They cry and whine all day long about efficiency of a plant here in the USA when competeing with the rest of the world that is focused on through put. Geuss whos been loosing that battle lately?
Frankly a bus hitting 90% of its theoretical is maxed out. There is no other way to look at it. This is why Motorolas G4 improvements look so good, they will hopefully deal with that congestion. The combination of DDR memory interfaces, faster buses (even if they are Rapid I/O), and lower power usage should provide a very nice boost to the G4.
The biggest problem with the G4 as I see it is the chips 32 bit nature. Its life is extremely limited. Apple has let the 64 bit genie out of the latern, in a couple of years they will not be able to put him back in at all.
All that being said if Apple comes out with one of these chips in a laptop I'd go for it. We I'd go for it if the speed jump was a little more than the pathetic 100MHz we just got.
Dave
Quote:
Originally posted by Amorph
You're way off.
Rag on MaxBus all you want, but its real bandwidth reached 80-90% of theoretical on a sustained basis. Most PC busses (and the 60x bus) are lucky to reach 50-60%. That's efficiency.
The G4's caches are incredibly aggressive. The core is lean (if anything, it's a bit too lean). The 745x AltiVec unit is at the top of the SIMD heap.
Put a fast bus on a 7457, and you'll have something that will give you incredibly high efficiency and very high performance, on a power budget that would stall a 970fx.
I know it's fashionable to trash the G4 now because it's not the best chip to put in a $3000 tower (gee, whodathunkit?) but you're way, way, off here. It's a nice design, and we haven't seen the last of it.
Here is an update on IBM's productions issues. Nothing substantially different from what we already know. There is one interesting comment:
Quote:
Fears that IBM's chip production problems are serious are seriously overblown, IBM Senior Vice-President and Chief Financial Officer John Joyce said in a recent conference call. And IBM has started to see much higher yields at its Fishkill plant lately, which means IBM's engineers are pretty close to ironing out the kinks, Joyce suggested.
The one thing that worries me is that they are showing 3 steps to get to 3Ghz and give no time frame.
Getting to 3 GHz should not be the emphasis here -- instead performance/clock and power/heat efficiency should be. More important is that the delivery of the e600 chip(s) needs to come within a year, preferably in time for Apple's next portable product update (i.e. 6-8 months from now).
More important is that the delivery of the e600 chip(s) needs to come within a year, preferably in time for Apple's next portable product update (i.e. 6-8 months from now).
After this Motorola update, I think it is *confirmed* : the next Powerbooks (end 2004-beginning 2005) will have enhanced G4s.
Getting to 3 GHz should not be the emphasis here -- instead performance/clock and power/heat efficiency should be. More important is that the delivery of the e600 chip(s) needs to come within a year, preferably in time for Apple's next portable product update (i.e. 6-8 months from now).
I understand all that, but there still has to be something for the consumer to go, WOW!
Comments
Originally posted by geekmeet
give us ome specifics about these "intriguing alternatives".
lets say the get the g4 to 2 ghz..........thats great but so what.
built-in memory controller/hypertansport/rapid i/o.......thats intrigue.
i remember that motorolas roadmap spoke of dual core g4's less than 25 watts at 2 ghz.
that would be amazing......
it would have to be REAL good for apple to swithc back to moto.
by the way im a raider fan and feel we have a pretty good draft.
i cant wait to see what those receivers can do.....
I don't know what they are because I don't know anyone over at FreeScale/Moto - just IBM. Just seems they are implying something is coming. They've publicly released documents indicating that the 7400 will go beyond 2GHz - and I think the memory controller may be on the map, as well as rapid i/o. Not too sure about hypertransport.
I'm a big time Browns fan - and I liked our draft, even with giving up the 2nd rounder. Looking forward to seeing Winslow going in under coverage.
Originally posted by Mr. MacPhisto
They've publicly released documents indicating that the 7400 will go beyond 2GHz - and I think the memory controller may be on the map, as well as rapid i/o.
Any link to these documents ?
EDIT: well, apart this one.
Originally posted by oldmacfan
Did you mean the Pats? Those guys from Boston don't no jack about future hardware.
Umm, no, I meant Pat Tillman.
Originally posted by Rhumgod
Umm, no, I meant Pat Tillman.
Tillman don't know jack either...
Originally posted by oldmacfan
Tillman don't know jack either...
Hey ass, do you have any idea who Pat Tillman is? He played for the Cardinals, but after Sept 11 he quit football, giving up his huge contract, to join the military. He recently died over in Afganistan. Show some respect.
Originally posted by kupan787
Hey ass, do you have any idea who Pat Tillman is? He played for the Cardinals, but after Sept 11 he quit football, giving up his huge contract, to join the military. He recently died over in Afganistan. Show some respect.
My appologies, yes I did hear the story, since we were talking about draft picks, I thought there was some other Pat that you were making reference to, as in he was drafted over the weekend.
Going back and reading the posts in reference, it makes more sense now. I am not good with name recognition, as a matter of fact, I suck at it. So, I was not being an Ass, I just didn't recognize the name. I am sorry if you think I have no repect for what he did, I do very much so. I myself am a veteran of the Navy.
Originally posted by emig647
Nah... I *think* the g4 is
You're way off.
Rag on MaxBus all you want, but its real bandwidth reached 80-90% of theoretical on a sustained basis. Most PC busses (and the 60x bus) are lucky to reach 50-60%. That's efficiency.
The G4's caches are incredibly aggressive. The core is lean (if anything, it's a bit too lean). The 745x AltiVec unit is at the top of the SIMD heap.
Put a fast bus on a 7457, and you'll have something that will give you incredibly high efficiency and very high performance, on a power budget that would stall a 970fx.
I know it's fashionable to trash the G4 now because it's not the best chip to put in a $3000 tower (gee, whodathunkit?) but you're way, way, off here. It's a nice design, and we haven't seen the last of it.
Originally posted by Amorph
... I know it's fashionable to trash the G4 now because it's not the best chip to put in a $3000 tower (gee, whodathunkit?) but you're way, way, off here. It's a nice design, and we haven't seen the last of it.
I would say that it is fashionable to trash the G4 because of Motorolla's history with the chip as well as a rather slow moving roadmap for future advances on the archetecture. This may change with the new management and structure and partnerships for the company, but I don't think that is evedent based on "real" information that we have today, such as speeds on current shipping 7457 chips. Powerlogix is shipping 1.33 Ghz processor upgrades, and did ship them at speeds of up to 1.4 Ghz, but Moto was already shipping 1.4 Ghz 7455's so the new processor had a net speed gain of 0. Now, 18 months after the introduction of Apple's 1.42 Ghz PowerMac we are finally seeing the G4 (7447) break that speed, and by a whopping 100 mhz. It is hard to be enthusiastic about the company based on recent history.
Originally posted by emig647
Nah... I *think* the g4 is
No,its prescott...unless you count certain Via designs.But the winky face shows you know it already,right?
http://www.cbronline.com/currentnews...256e830032dd99
IBM Cuts iSeries Prices Ahead of Power5 Launch
By Timothy Prickett Morgan
Product transitions are always a pain in the neck for any IT vendor. As soon as they know a kicker product is imminent, all customers, except those desperate to buy a new piece of hardware or software, sit tight, because they want to see what the new product will be, and how it will be priced. This makes perfect sense to IBM midrange shops that may have one, two, or even three decades of experience in acquiring computers.
If this is a problem for customers, it is, in many ways, a much more troublesome situation for IBM Corp. During the past several decades, about the only way to get System/3X, AS/400, or iSeries customers to buy ahead of a future product launch has been to cut prices, and this is exactly what IBM did last week with the iSeries line.
IBM learned a long time ago, way back when it was first selling System/360 computers, and even earlier with punch card tabulating equipment, that there is almost a perfectly elastic relationship between the price of an IT product and the ability of companies to absorb more and more capacity. However, IBM's goal has never been the proliferation of a technology, so much as the maintenance of the absolute maximum profits that can be wrung out of what amounts to captive customer bases locked into products by their legacy applications. IBM is all about profits, but it has to maintain a revenue stream to keep its top brass, sales reps, partner channel, and Wall Street happy. With iSeries sales down 7% in the first quarter and new Power5-based iSeries machines (or whatever they will be called) not expected until later in the second quarter, IBM had to do something to stimulate sales until the Power5 "Squadron" machines arrive on the scene, sometime before the end of the second quarter.
It is interesting to see what IBM did. First, it cut the list prices on base configurations of Model 810 servers (including iSeries for High Availability models) by between 9% and 35%. IBM also cut main memory prices on the selected Model 270s and on all Model 810 servers by 40%. The Model 810 machines use many of the same main memory cards that are used in the Model 270s, so it did not make much sense to not make the 40% price cut effective on these Model 270s that share at least some memory cards with the Model 810s. After the price cuts, Model 810 and Model 270 main memory prices range from $1.89 per MB, on relatively dense cards with 512MB or 1GB capacities, up to $4.32 per MB on cards with only 128MB of capacity. And I wouldn't be me if I didn't write the following sentence: While these price cuts are good, IBM is still charging about twice as much for ECC main memory on entry iSeries servers as it does for its own xSeries line and its competitors do for their X86-based servers.
IBM also cut list prices on the base configurations of Model 870 (16-way capable) and Model 890 (32-way capable) iSeries servers. Sources at IBM say that, generally speaking, list prices on these base boxes were dropped by approximately 20%. The actual price cuts vary on the processor performance and the OS/400 edition (Standard Edition with little 5250 green-screen capability or Enterprise Edition with fully enabled 5250 processing). Memory prices on the Model 870 and Model 890 were chopped by 60%. You heard that right. This brings iSeries memory, which can account for 25% of the cost of a configured system, more in line with other high-end SMP servers available from Big Blue and others. I said "more in line," not actually "in line." While IBM has cut prices from $12.15 to $4.86 per MB on the Model 870 and the Model 890, that is still about twice the cost (at list price), compared with ECC main memory for high-end Unix servers. Then again, big Unix boxes are sold at 35% to 50% of list these days, so the disparity at the street price level is larger than it might otherwise seem.
IBM cut prices on 10K RPM and 15K RPM SCSI disk drives used across the iSeries product line by between 20% and 50%, depending on the disk capacity and rotational speed. The price cuts are roughly in line with the disk price cuts that IBM announced almost exactly two years ago. After the cuts, IBM is charging $1,000, or 6 cents a MB, for a 17.5GB, 10K RPM disk. A 35.2GB disk spinning at 10K costs $1,359, or 4 cents per MB, while a 15K version of the 35.2GB disk costs $1,875, or 5 cents per MB. IBM's fattest 70.6GB, 15K RPM disk now costs $2,875, or 4 cents per MB.
What IBM did not do, you might notice, is cut prices on the entry Model 800 or on the midrange Model 825. This is the best indication that IBM does not intend to dramatically improve the price/performance of machines in this class with the impending Power5-based Squadron servers. What it probably means is that IBM will launch a two-way and four-way capable Squadron box that competes against the Model 810 and which will have similar bang for the buck. It also seems to imply that IBM will not immediately ship an eight-way Squadron box, or if it does, it will not offer much better price/performance than the current Power4-based Model 825 does.
That doesn't mean that Model 800 component prices have not changed. The main memory in the Model 800 had its price cut by 40%, and it uses the same disk drives that had their prices cut. Because IBM wants to keep the prices of the Model 800 machines constant, it actually raised the price of the base Model 800 servers to offset the cuts on the memory and disk components included in the base Model 800 packages. This is IBM and its iSeries channel trying to maintain a profit margin on a base box. There is no other explanation as to why IBM would not let the price on the entry iSeries Model 800 drop below $8,795. Obviously, adding more memory and disks to a Model 800 above and beyond the base configuration is cheaper than it was before the price cut, so it is not all bad. But it would have probably helped Model 800 shipments considerably if IBM actually cut the prices on these boxes by 25% to 30%. The resellers we have talked to in recent weeks say that the Model 800 is not selling well, and there is no surprise to that. It is too expensive, and especially so compared to second-hand Model 250 and Model 270 servers.
There is one more price change that Big Blue announced last week that might show how IBM will be positioning the future Power5 machines. On the Model 825, Model 870, and Model 890 servers, IBM lowered the price of permanently activating excess processors in a box by $15,000 while raising the price of activating an OS/400 license on those processors. Big deal, right? It's a net wash to activate OS/400 on a processor, so what is there to be excited about. Well, don't forget about Linux on the Power4 machines and Linux and AIX on the Power5 machines. Before this price change, it cost $20,000 to activate a 1.1GHz Power4 processor beyond the base three processors that come with a Model 825; after the cut, it only costs $5,000. Given that a single Power4 processor can do about twice the work of the fastest Xeon processor, this is the first time that the extra processors in the Model 825, which scales to six processors, are in price parity with the X86 Xeon and Opteron alternatives in the server market.
You heard that right - so now, adding a few extra processors to a Model 825 and slapping on 64-bit Linux on partitions running on these extra processors makes economic and technical sense, and now IBM can really make a credible argument for moving Web infrastructure workloads back onto the iSeries and inside its logical partitions.
The processor activation on the Model 870s and Model 890s is not nearly as attractive, which is a shame. Before the cuts, it cost $25,000 to activate a processor on a Model 870 (there are two versions of this machine, one that scales from 5 to 8 processors and another that scales from 8 to 16 processors); now it costs $10,000. This Model 870 uses a 1.3GHz Power4 processor, and why IBM thinks it is worth twice as much as a 1.1GHz processor is unknown. It just isn't. The price of activating a 1.3GHz Power4 processor on a high-end Model 890 used to cost $30,000, and now it costs $15,000. While it would have been great if IBM had standardized pricing across the iSeries line in such a way that bears some resemblance to performance - a 1.1GHz Power4 processor costs $5,000, and a 1.3GHz processor costs $6,000 - these lower prices are nonetheless a good thing. I also think that IBM will package the Squadron machines in a similar way, such that it is relatively inexpensive to add Linux or AIX, but OS/400 processors will cost about what they do now. IBM wants to encourage server consolidation on the iSeries, but it wants to rake in those profits on OS/400.
http://www.cbronline.com/currentnews...256e830032dd9b
AS/400 7XX Shops Get Rebates on iSeries 8XX Boxes
By Timothy Prickett Morgan
IBM Corp stopped offering upgrades from AS/400 Model 720, 730, and 740 servers to the iSeries 8XX machines in October 2003, in the hope that doing so would convince customers to upgrade before the end of the third quarter of 2003. The lack of upgrades probably did help IBM sell iSeries boxes in that quarter, but the lack of upgrades from 7XX to 8XX models has hurt sales since then.
That's why IBM instituted a promotion last week aimed at Model 7XX and is giving decent trade-in rebates for upgrading to an iSeries Model 810 or Model 825 server running OS/400 Enterprise Edition. These are not traditional upgrades, where IBM protects the customer's depreciation schedule on an existing machine by preserving that machine's serial number after the upgrade. This is a push-pull deal, where a 7XX customer will acquire a Model 810 or Model 825 and then ship back the 7XX box to IBM. Big Blue will then do as it pleases with the 7XX box, perhaps reselling it through its Global Financing unit or putting in the metal recycling bin to keep it off the open market.
Last week IBM dropped prices substantially on the more powerful versions of the Model 810 server (price cuts were above 60%), and even the entry box (Model 810-2465) got a 9% discount. When you throw the 7XX upgrade promotion on top of these Model 810 price cuts, the effective price of a Model 810 for a 7XX shop is about half, compared with acquiring a Model 810 before April 20. On the smaller Model 810 box, the discount and the rebate amounts to a 27% discount.
IBM did not cut prices on the Model 825 server, but it is offering a $50,000 trade-in credit to 7XX customers who trade up to a Model 825. This works out to a 15% discount off the acquisition costs.
Other discounts and deals can be combined with the 7XX to 8XX trade-in promotion, which will run until June 30. That is the end of IBM's second quarter, and it is also the date by which IBM has promised it will start shipping its Power5-based "Squadron" servers. They could, of course, ship earlier. Resellers and other business partners cannot upgrade their machines as part of this deal. It is intended for end user customers only. IBM is not offering trade-ins on Model 810 and Model 825 servers running OS/400 Standard Edition, and it similarly did not cut prices on any iSeries machines running Standard Edition on April 20.
With the current version of OS/X this should be very easy to implement. After all, all OS/X does is handle 32 bit processes in a larger address space. Granted the G4's ability to quickly get to all that memory is an issue with the current front side bus but I have to imagine that a solution for that has to be in the works.
Dave
Originally posted by Matsu
I wonder if moto/freescale will have a 64 bit variant suitable for laptop use. It isn't critical at this time. There ain't no 64 bit "mobile" chips from intel or AMD. There are 64 bit "luggables" based on AMD chips but nothing we'd call a true "laptop".
I ask because if there is something with that huge address space, we might never see an IBM based Powerbook at all, what would be the point, so long as the moto offering is small/cool/frugal, and offers industry standard level performance. If there's no need for 64 bits, or there's a way to cover that need with a more mobile friendly chip, then there's no need for IBM in a laptop.
If, however, there is a pressing need for a mobile professional 64 bit platform, and no small/cool solution, I do see that we may have a third tire of Apple laptops added.
iBook G4
Powerbook G4+ ?
Powerbook G5
All to exist simultaneously, 32 bit machines with an emphasis on battery life, 64 bit machines with an emphasis on number crunching.
Frankly a bus hitting 90% of its theoretical is maxed out. There is no other way to look at it. This is why Motorolas G4 improvements look so good, they will hopefully deal with that congestion. The combination of DDR memory interfaces, faster buses (even if they are Rapid I/O), and lower power usage should provide a very nice boost to the G4.
The biggest problem with the G4 as I see it is the chips 32 bit nature. Its life is extremely limited. Apple has let the 64 bit genie out of the latern, in a couple of years they will not be able to put him back in at all.
All that being said if Apple comes out with one of these chips in a laptop I'd go for it. We I'd go for it if the speed jump was a little more than the pathetic 100MHz we just got.
Dave
Originally posted by Amorph
You're way off.
Rag on MaxBus all you want, but its real bandwidth reached 80-90% of theoretical on a sustained basis. Most PC busses (and the 60x bus) are lucky to reach 50-60%. That's efficiency.
The G4's caches are incredibly aggressive. The core is lean (if anything, it's a bit too lean). The 745x AltiVec unit is at the top of the SIMD heap.
Put a fast bus on a 7457, and you'll have something that will give you incredibly high efficiency and very high performance, on a power budget that would stall a 970fx.
I know it's fashionable to trash the G4 now because it's not the best chip to put in a $3000 tower (gee, whodathunkit?) but you're way, way, off here. It's a nice design, and we haven't seen the last of it.
Originally posted by PB
Any link to these documents ?
EDIT: well, apart this one.
OK, it seems that Motorola actually reads these boards and delivers information on command :
Freescale Semiconductor reveals PowerPC® core roadmap and scalable system-on-chip platforms
High-Performance PowerPC Processors
Roadmap
The one thing that worries me is that they are showing 3 steps to get to 3Ghz and give no time frame.
Fears that IBM's chip production problems are serious are seriously overblown, IBM Senior Vice-President and Chief Financial Officer John Joyce said in a recent conference call. And IBM has started to see much higher yields at its Fishkill plant lately, which means IBM's engineers are pretty close to ironing out the kinks, Joyce suggested.
Originally posted by oldmacfan
The one thing that worries me is that they are showing 3 steps to get to 3Ghz and give no time frame.
Getting to 3 GHz should not be the emphasis here -- instead performance/clock and power/heat efficiency should be. More important is that the delivery of the e600 chip(s) needs to come within a year, preferably in time for Apple's next portable product update (i.e. 6-8 months from now).
Originally posted by Programmer
More important is that the delivery of the e600 chip(s) needs to come within a year, preferably in time for Apple's next portable product update (i.e. 6-8 months from now).
After this Motorola update, I think it is *confirmed* : the next Powerbooks (end 2004-beginning 2005) will have enhanced G4s.
Originally posted by PB
OK, it seems that Motorola actually reads these boards and delivers information on command :
Freescale Semiconductor reveals PowerPC® core roadmap and scalable system-on-chip platforms
High-Performance PowerPC Processors
Roadmap
this is impressive.
the next generation e600(im assuming "e" means "book e" remember that?) has a dual-core option!
and will hit 2 ghz!
i will make a big assumption and say this is the chip people have been droping hints about.
oh boy .......i hate to say it but it looks like moto is back in the race and in a big way!
it will be interesting to see how things shake out.
Originally posted by Programmer
Getting to 3 GHz should not be the emphasis here -- instead performance/clock and power/heat efficiency should be. More important is that the delivery of the e600 chip(s) needs to come within a year, preferably in time for Apple's next portable product update (i.e. 6-8 months from now).
I understand all that, but there still has to be something for the consumer to go, WOW!