I'm pretty sure .... As PROGRAMMER said " The 970 will run 32-bit apps and "G4 code" just fine. The small amount of supervisor code in the kernel needed to make the OS function is (probably) the only exception and as you said that could be taken care of in a small-dot release. These processors all run 32-bit PowerPC code and optimizing it for one or the other doesn't limit which processors can run the code, it just affects their performance in doing so. The one current caveat is that AltiVec code cannot run on non-AltiVec enabled processors."
970 will be sold in single and dual config... *CONFIRMED!*
single (low-end) 970 will smoke current dual (high-end) G4... *CONFIRMED!*
Bryce 6 will be multi-threaded... *CONFIRMED!*
rough parity with Wintel world again... *CONFIRMED!*
-
*unCONFIRMED!* possibly sketchy news from MacBidoulle:
low-end 1.4, mid 1.6-1.8, high 2.0...
IBM and other apparent leaks suggest initial ranges from 1.6-2.5
unspecified environment/bus/cache config between models...
history suggests more than just the CPU speed changes between models.
Bus speed/RAM/Mobo/GFX card/L2-vs-L3(size and speed) may differ from single to dual, so logically, benchmarks may seem odd due to non-standard config between test units
--
if Apple Legal jumps all over MB to pull the "rumours", it might validate them a bit, but it might just be seen as jumping on the current "French-bashing bandwagon"
IMO, there is some truth to the view that preaching to the choir means some people will believe because they want it to be true, but given the preponderance of other information that seems to have either leaked out or, by analysis (here and elsewhere), has been theoretically dissected and reverse-engineered in terms of hypothetical pipeline length/FSB/clock/dualFPU/etc to the point of reasonable predictions...
the apparent convergence between these sources of info is a good sign that these predictions are in the ballpark.
it's also possible (and historically precedented) that the benchmark machines are early mules (of which dozens of variations and tweaks are produced), and that the actual shipping hardware is similar in some config, different in others.
(for example, these may have been benchmarked on current DDR RAM and current GFX card... publicly available models may have faster/better/wider/more fragrant options that cause the first widely verifiable benchmarks to differ from those posted here... this wouldn't mean MB has lied, merely that the iron they tested on isn't the same as what ships as "970 Rev. A")
unless we're all guilty of gleefully drinking the kool-aid,
i'm thinking this is damn tasty stuff... got a hankering for more.
as for the hypothesized XStation with 16 blades of 4x970 each, for a total of 64 processors, this is also entirely believable and has historical precedent
there was a NeXT model which supported multiple CPUs in multiple slots for, IIRC, a total of 16 x 68030 or 68040 CPUs in a single machine... threaded multiprocessing back in the early 90s.
given the number of former NeXT folks at Apple, it's a no-brainer that the hardware teams have had one of those on the pedestal with a "beat this" sticker on it for motivation.
If "Steve's Old Company" could do it more than 10 years ago, then Steve's Original (older) Company with similar staff should be able to do it now that the technology has caught up.
< rummages around trying to find specs of a multiple-Nitro-boarded MP Cube >
I was reading a post elsewhere that quoted an Ars post stating that the 970 had a weaker int and altivec units than the G4e ... And can the int performance really be worse?
Like always it's not much left to say after Programmer has answered
The G4e (745X family) with its 4 integer units and feeding 3 of them per beat should perform about 50% faster than a G4 (7400 & 7410) at the same clock speed. But it didn't in real life because the redesign didn't remove other bottlenecks of the G4.
Don't bother about how good or how bad a chip design looks on the papers. Only real life performance matters. And a 970 will outperform any G4 for shure (or even two of them at once ).
If IBM can get a POWER4 to 1.7GHz then why shouldn't a 970 reach 2GHz right from the start?
I don't pretend to be as versed in the nuances of Chip Voodoo that many around here are, but I keep getting this feeling that everyone is expecting a mini Power4 or Power5 (except a lot cheaper) in their next Mac. As it's been stated, a 970/980/whatever is a derivitive of their ultra-parents. What that ultimately means is anyones guess at this point.
To wit: I have two Volkswagons that have a derivitive of the most advanced Porsche engine of their time. Do they go as fast as a Porsche? Hell no. But they are clean, mean and lean (and upgradeable ). They kick ass as far as efficiency, ease of use and user friendliness. Sound familiar?
I remember as much as 2 or 3 months ago, 1 or 2 technology writers who (apparently under NDA) couldn't give specifics, mentioned no processors, specific platforms, etc., but indicated in very general terms that the performance of this summers new Macs would bring them into parity and possibly a small jump ahead of current Windows offerings.
That's not anything at all like what the MacDoobie benchmarks describe. The 'Doobie benchmarks show the low end, 1.4 GHz Mac keeping even with and at times surpassing a Pentium IV, while the high end, dual 1.8 GHz Mac decimates the Pentium IV. There is no "performance parity" with the Wintel machine, only a bloody stump ringed in pulpy splatters and driblets.
I sure hope that tech writer you speak of is wrong, because what Apple needs now is a shot in the mainline of 100% pure methamphetamine.
That's not anything at all like what the MacDoobie benchmarks describe. The 'Doobie benchmarks show the low end, 1.4 GHz Mac keeping even with and at times surpassing a Pentium IV, while the high end, dual 1.8 GHz Mac decimates the Pentium IV. There is no "performance parity" with the Wintel machine, only a bloody stump ringed in pulpy splatters and driblets.
Okay, I don't follow Wintel benchmarks that closely but it is my impression is that there was "only a bloody stump ringed in pulpy splatters" partly because they didn't put the dual 970 up against the fastest Wintel machines. How does a P4 stand up against a dual Xeon. Am I wrong in thinking that that would be a little closer?
That's not anything at all like what the MacDoobie benchmarks describe. The 'Doobie benchmarks show the low end, 1.4 GHz Mac keeping even with and at times surpassing a Pentium IV, while the high end, dual 1.8 GHz Mac decimates the Pentium IV. There is no "performance parity" with the Wintel machine, only a bloody stump ringed in pulpy splatters and driblets.
I sure hope that tech writer you speak of is wrong, because what Apple needs now is a shot in the mainline of 100% pure methamphetamine.
Nicely put.
However, it remains to be seem how much the 90nm Prescott P4 improves over 130nm Northwood P4 with the added instructions and, apparently, much better 'hyperthreading' (SMT). This may bring back 'parity' of the top end x86 and PPC in single processor configs.
The above assumes the MB data is roughly correct, which I hope it is.
That's not anything at all like what the MacDoobie benchmarks describe. The 'Doobie benchmarks show the low end, 1.4 GHz Mac keeping even with and at times surpassing a Pentium IV, while the high end, dual 1.8 GHz Mac decimates the Pentium IV. There is no "performance parity" with the Wintel machine, only a bloody stump ringed in pulpy splatters and driblets.
I sure hope that tech writer you speak of is wrong, because what Apple needs now is a shot in the mainline of 100% pure methamphetamine.
Very true. I still can't find those links, and will have to keep looking. My comments were deliberately understated until I can find the links, because I cannot remember the exact descriptive adjectives used by the writers, but they were a bit more optimistic although neccesarily vague at the time. I just don't wish to misquote them.
I was very surprised by the macbidouille benchmarks. I expected a big performance increase, but nothing that great. I really hope it is true. This is just the beginning, and the road map looks good for the future, no matter what Intel brings out. I actually look forward to seeing how a dual 1.8 970 does stack up against a dual 3.06 Xeon.
I actually look forward to seeing how a dual 1.8 970 does stack up against a dual 3.06 Xeon.
Want any tests run? I just got in a Dual Xeon 3.06 DP machine. I'm running 6Gb RAM, RAID 320 0+1 I don't have Windows for it but I do have RH Linux Ent AS 2.1.
Want any tests run? I just got in a Dual Xeon 3.06 DP machine. I'm running 6Gb RAM, RAID 320 0+1 I don't have Windows for it but I do have RH Linux Ent AS 2.1.
Could you please install Doom II or something? I like framerates!
Want any tests run? I just got in a Dual Xeon 3.06 DP machine. I'm running 6Gb RAM, RAID 320 0+1 I don't have Windows for it but I do have RH Linux Ent AS 2.1.
I was going to purchase a Dual Xeon for 3D right before IBM started talking about the 970, and they (Cebit folk) all figured out that it would go into a Mac. From Much Older Speculations on the speed of the 970's. I came to a conclusion that a dual 1.8 would match a Dual 3.GHZ Xeon machine in raw speed based on early cebit crowd Speculations. I was not taking into account for Altivec because I am not sure how to rate this new Altivec unit. nor do I know what else the motherboard will include. Needless to say I put off that purchase until I saw what these new Machines would be truly able to do.
So one of two things could happen. We could put together some good showdown #'s at a later date, or we will be talking about how nice my render speeds are on that Dual Xeon I was talking about getting.
Truth is I can put off this purchase only for so long before I have to get on a faster boat.
But my boat of choice has always been a Mac, and I'd like it to stay that way.
That is correct. All that has been announce is 1.2 to 1.8. Reliable word has slipped out that IBM is experiencing higher then expected yields of faster chips up to 2.5 ghz.
The key words are higher then expected. This does not mean that the yields are production level. If their expectation was to have a yield of one 2.5 ghz chip per 100 (random speculation), and they ended up with 2 per 100, that would be a higher then expected yield. It would still be far shy of producing enough chips in that range to meet demand.
Even MacBidouille indicates that the availability of the dual 2.0 is uncertain, because of chip yields. Expect 1.4 to 1.8 ghz machines in single and dual configurations. Speed bumps will come quickly when yields improve. I would expect at least one or 2 speed bumps before January.
onlooker, i just did a quick check and on pricewatch i found a no name xeon system for 2k which i do not think was dual?!?
What sort of prices are you seeing for dual?
If its in the $2500 range and a dual 1.6 970 system is in the $2200 range then things should get intresting soon.
Not only with the Xeon either. The Opteron and Athlon64 will make the next 12 months a very interesting time for Apple fans. Why? Because once Apple get 'parity' with Wintelon they will want to see that stay that way or improve but they will be aiming at a moving target since Intel and AMD aren't standing still.
Very interesting times to be thinking of an upgrade.
Comments
I'm pretty sure .... As PROGRAMMER said " The 970 will run 32-bit apps and "G4 code" just fine. The small amount of supervisor code in the kernel needed to make the OS function is (probably) the only exception and as you said that could be taken care of in a small-dot release. These processors all run 32-bit PowerPC code and optimizing it for one or the other doesn't limit which processors can run the code, it just affects their performance in doing so. The one current caveat is that AltiVec code cannot run on non-AltiVec enabled processors."
Think about it
8)
*CONFIRMED!* positive news from MacBidouille:
970 will be sold in single and dual config... *CONFIRMED!*
single (low-end) 970 will smoke current dual (high-end) G4... *CONFIRMED!*
Bryce 6 will be multi-threaded... *CONFIRMED!*
rough parity with Wintel world again... *CONFIRMED!*
-
*unCONFIRMED!* possibly sketchy news from MacBidoulle:
low-end 1.4, mid 1.6-1.8, high 2.0...
IBM and other apparent leaks suggest initial ranges from 1.6-2.5
unspecified environment/bus/cache config between models...
history suggests more than just the CPU speed changes between models.
Bus speed/RAM/Mobo/GFX card/L2-vs-L3(size and speed) may differ from single to dual, so logically, benchmarks may seem odd due to non-standard config between test units
--
if Apple Legal jumps all over MB to pull the "rumours", it might validate them a bit, but it might just be seen as jumping on the current "French-bashing bandwagon"
IMO, there is some truth to the view that preaching to the choir means some people will believe because they want it to be true, but given the preponderance of other information that seems to have either leaked out or, by analysis (here and elsewhere), has been theoretically dissected and reverse-engineered in terms of hypothetical pipeline length/FSB/clock/dualFPU/etc to the point of reasonable predictions...
the apparent convergence between these sources of info is a good sign that these predictions are in the ballpark.
it's also possible (and historically precedented) that the benchmark machines are early mules (of which dozens of variations and tweaks are produced), and that the actual shipping hardware is similar in some config, different in others.
(for example, these may have been benchmarked on current DDR RAM and current GFX card... publicly available models may have faster/better/wider/more fragrant options that cause the first widely verifiable benchmarks to differ from those posted here... this wouldn't mean MB has lied, merely that the iron they tested on isn't the same as what ships as "970 Rev. A")
unless we're all guilty of gleefully drinking the kool-aid,
i'm thinking this is damn tasty stuff... got a hankering for more.
ah, the unquenchable thirst for power.
there was a NeXT model which supported multiple CPUs in multiple slots for, IIRC, a total of 16 x 68030 or 68040 CPUs in a single machine... threaded multiprocessing back in the early 90s.
given the number of former NeXT folks at Apple, it's a no-brainer that the hardware teams have had one of those on the pedestal with a "beat this" sticker on it for motivation.
If "Steve's Old Company" could do it more than 10 years ago, then Steve's Original (older) Company with similar staff should be able to do it now that the technology has caught up.
< rummages around trying to find specs of a multiple-Nitro-boarded MP Cube >
Originally posted by kupan787
I was reading a post elsewhere that quoted an Ars post stating that the 970 had a weaker int and altivec units than the G4e ... And can the int performance really be worse?
Like always it's not much left to say after Programmer has answered
The G4e (745X family) with its 4 integer units and feeding 3 of them per beat should perform about 50% faster than a G4 (7400 & 7410) at the same clock speed. But it didn't in real life because the redesign didn't remove other bottlenecks of the G4.
Don't bother about how good or how bad a chip design looks on the papers. Only real life performance matters. And a 970 will outperform any G4 for shure (or even two of them at once
POWER4+ 1.7GHz (1 core):
SPECint2000 1077(base) 1133(peak)
SPECfp2000 1598(base) 1699(peak)
Available from july on.
If IBM can get a POWER4 to 1.7GHz then why shouldn't a 970 reach 2GHz right from the start?
Originally posted by smalM
IBM has announced new POWER4+ chips at 1.5 and 1.7 GHz.
POWER4+ 1.7GHz (1 core):
SPECint2000 1077(base) 1133(peak)
SPECfp2000 1598(base) 1699(peak)
Available from july on.
If IBM can get a POWER4 to 1.7GHz then why shouldn't a 970 reach 2GHz right from the start?
$$$$$$$$$$$$$$$$$$
Originally posted by smalM
IBM has announced new POWER4+ chips at 1.5 and 1.7 GHz.
POWER4+ 1.7GHz (1 core):
SPECint2000 1077(base) 1133(peak)
SPECfp2000 1598(base) 1699(peak)
Available from july on.
If IBM can get a POWER4 to 1.7GHz then why shouldn't a 970 reach 2GHz right from the start?
Goddamn! IBM impresses!
Originally posted by da Pr0grammah
$$$$$$$$$$$$$$$$$$
!
Originally posted by smalM
If IBM can get a POWER4 to 1.7GHz then why shouldn't a 970 reach 2GHz right from the start?
I don't pretend to be as versed in the nuances of Chip Voodoo that many around here are, but I keep getting this feeling that everyone is expecting a mini Power4 or Power5 (except a lot cheaper) in their next Mac. As it's been stated, a 970/980/whatever is a derivitive of their ultra-parents. What that ultimately means is anyones guess at this point.
To wit: I have two Volkswagons that have a derivitive of the most advanced Porsche engine of their time. Do they go as fast as a Porsche? Hell no. But they are clean, mean and lean (and upgradeable
Originally posted by Shaktai
I remember as much as 2 or 3 months ago, 1 or 2 technology writers who (apparently under NDA) couldn't give specifics, mentioned no processors, specific platforms, etc., but indicated in very general terms that the performance of this summers new Macs would bring them into parity and possibly a small jump ahead of current Windows offerings.
That's not anything at all like what the MacDoobie benchmarks describe. The 'Doobie benchmarks show the low end, 1.4 GHz Mac keeping even with and at times surpassing a Pentium IV, while the high end, dual 1.8 GHz Mac decimates the Pentium IV. There is no "performance parity" with the Wintel machine, only a bloody stump ringed in pulpy splatters and driblets.
I sure hope that tech writer you speak of is wrong, because what Apple needs now is a shot in the mainline of 100% pure methamphetamine.
Originally posted by Junkyard Dawg
That's not anything at all like what the MacDoobie benchmarks describe. The 'Doobie benchmarks show the low end, 1.4 GHz Mac keeping even with and at times surpassing a Pentium IV, while the high end, dual 1.8 GHz Mac decimates the Pentium IV. There is no "performance parity" with the Wintel machine, only a bloody stump ringed in pulpy splatters and driblets.
Okay, I don't follow Wintel benchmarks that closely but it is my impression is that there was "only a bloody stump ringed in pulpy splatters" partly because they didn't put the dual 970 up against the fastest Wintel machines. How does a P4 stand up against a dual Xeon. Am I wrong in thinking that that would be a little closer?
Originally posted by Junkyard Dawg
That's not anything at all like what the MacDoobie benchmarks describe. The 'Doobie benchmarks show the low end, 1.4 GHz Mac keeping even with and at times surpassing a Pentium IV, while the high end, dual 1.8 GHz Mac decimates the Pentium IV. There is no "performance parity" with the Wintel machine, only a bloody stump ringed in pulpy splatters and driblets.
I sure hope that tech writer you speak of is wrong, because what Apple needs now is a shot in the mainline of 100% pure methamphetamine.
Nicely put.
However, it remains to be seem how much the 90nm Prescott P4 improves over 130nm Northwood P4 with the added instructions and, apparently, much better 'hyperthreading' (SMT). This may bring back 'parity' of the top end x86 and PPC in single processor configs.
The above assumes the MB data is roughly correct, which I hope it is.
MM
Originally posted by Junkyard Dawg
That's not anything at all like what the MacDoobie benchmarks describe. The 'Doobie benchmarks show the low end, 1.4 GHz Mac keeping even with and at times surpassing a Pentium IV, while the high end, dual 1.8 GHz Mac decimates the Pentium IV. There is no "performance parity" with the Wintel machine, only a bloody stump ringed in pulpy splatters and driblets.
I sure hope that tech writer you speak of is wrong, because what Apple needs now is a shot in the mainline of 100% pure methamphetamine.
Very true. I still can't find those links, and will have to keep looking. My comments were deliberately understated until I can find the links, because I cannot remember the exact descriptive adjectives used by the writers, but they were a bit more optimistic although neccesarily vague at the time. I just don't wish to misquote them.
I was very surprised by the macbidouille benchmarks. I expected a big performance increase, but nothing that great. I really hope it is true. This is just the beginning, and the road map looks good for the future, no matter what Intel brings out. I actually look forward to seeing how a dual 1.8 970 does stack up against a dual 3.06 Xeon.
It is no longer a scary proposition.
Originally posted by curiousuburb
IBM and other apparent leaks suggest initial ranges from 1.6-2.5
You won't see 2.5 GHz initially.
Originally posted by Shaktai
I actually look forward to seeing how a dual 1.8 970 does stack up against a dual 3.06 Xeon.
Want any tests run? I just got in a Dual Xeon 3.06 DP machine. I'm running 6Gb RAM, RAID 320 0+1
Originally posted by Mike
Want any tests run? I just got in a Dual Xeon 3.06 DP machine. I'm running 6Gb RAM, RAID 320 0+1
Could you please install Doom II or something? I like framerates!
Originally posted by Mike
Want any tests run? I just got in a Dual Xeon 3.06 DP machine. I'm running 6Gb RAM, RAID 320 0+1
I was going to purchase a Dual Xeon for 3D right before IBM started talking about the 970, and they (Cebit folk) all figured out that it would go into a Mac. From Much Older Speculations on the speed of the 970's. I came to a conclusion that a dual 1.8 would match a Dual 3.GHZ Xeon machine in raw speed based on early cebit crowd Speculations. I was not taking into account for Altivec because I am not sure how to rate this new Altivec unit. nor do I know what else the motherboard will include. Needless to say I put off that purchase until I saw what these new Machines would be truly able to do.
So one of two things could happen. We could put together some good showdown #'s at a later date, or we will be talking about how nice my render speeds are on that Dual Xeon I was talking about getting.
Truth is I can put off this purchase only for so long before I have to get on a faster boat.
But my boat of choice has always been a Mac, and I'd like it to stay that way.
Originally posted by Telomar
You won't see 2.5 GHz initially.
That is correct. All that has been announce is 1.2 to 1.8. Reliable word has slipped out that IBM is experiencing higher then expected yields of faster chips up to 2.5 ghz.
The key words are higher then expected. This does not mean that the yields are production level. If their expectation was to have a yield of one 2.5 ghz chip per 100 (random speculation), and they ended up with 2 per 100, that would be a higher then expected yield. It would still be far shy of producing enough chips in that range to meet demand.
Even MacBidouille indicates that the availability of the dual 2.0 is uncertain, because of chip yields. Expect 1.4 to 1.8 ghz machines in single and dual configurations. Speed bumps will come quickly when yields improve. I would expect at least one or 2 speed bumps before January.
What sort of prices are you seeing for dual?
If its in the $2500 range and a dual 1.6 970 system is in the $2200 range then things should get intresting soon.
Originally posted by keyboardf12
onlooker, i just did a quick check and on pricewatch i found a no name xeon system for 2k which i do not think was dual?!?
What sort of prices are you seeing for dual?
If its in the $2500 range and a dual 1.6 970 system is in the $2200 range then things should get intresting soon.
Not only with the Xeon either. The Opteron and Athlon64 will make the next 12 months a very interesting time for Apple fans. Why? Because once Apple get 'parity' with Wintelon they will want to see that stay that way or improve but they will be aiming at a moving target since Intel and AMD aren't standing still.
Very interesting times to be thinking of an upgrade.
MM