Correct! The rest of it is just socket space for those giant pins (or lack therefore of on the processor for LGA775).
~desu
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Intel's Penryn Chip Performance Boosts
- Thread starter MacRumors
- Start date
- Sort by reaction score
~desu
And it has to be, because at 3 GHz speed of light is only 10 cm per cycle. And electrical signals don't move at speed of light. These five cm are about one processor cycle at the speed of an electrical signal.
shuddup...
shuddup...
some of us still have G4s....dabnabbit...
lol, all in fun.. Actually I am expecting my MacPro 8core to be supposedly delivered tomorrow--a few days late because evidentually FedEx pilots wont take off in the rain over in China--boo hoo..heh.
shuddup...
some of us still have G4s....dabnabbit...
Oh man !! 3.33GHz!! Thats amazing ! That makes me feel even worse about being on my G5, though i love it
lol, all in fun.. Actually I am expecting my MacPro 8core to be supposedly delivered tomorrow--a few days late because evidentually FedEx pilots wont take off in the rain over in China--boo hoo..heh.
There is some interesting stuff coming out the next year. Just a couple of examples (don't have the links handy but they are all over the net):
- Blu-Ray burners
- High capacity NAND flash HDs 64-128GB
- 802.11n "final" spec products
- high res LED screens
- 300GB HDs
- Harware encryption of data within the HD itself
- etc.
2008 is definitely going to be an interesting year hardware-wise not just because of Penryn.
- Blu-Ray burners
- High capacity NAND flash HDs 64-128GB
- 802.11n "final" spec products
- high res LED screens
- 300GB HDs
- Harware encryption of data within the HD itself
- etc.
2008 is definitely going to be an interesting year hardware-wise not just because of Penryn.
And some of us still run G4s in the MHz... Like my rockin'-awesome 800MHz G4 iMac. Man that thing is a beauty. She's slowin' down though...
-Clive
I Am Trapped In The Wait For SS & Penryn Sink Hole
I fully respect your superior knowledge on all things technical Aiden. Are you saying that the confusion that Bare Feats finds among those cores in the "new" 8 core MP won't be mollified with SS & Leopard to a significant degree of improvement? I'm thinking that the delay in shipping Leopard is reason enough to wait for SS with Penryn as it will be shipping before end of year about the same time as Leopard. Is that stinkin' thinkin'?
Exactly right. I want all the brains that the 8 cores of power need and I'd rather have that bigger 12MB cache in Penryn as well now that Leopard is delayed.
I fully respect your superior knowledge on all things technical Aiden. Are you saying that the confusion that Bare Feats finds among those cores in the "new" 8 core MP won't be mollified with SS & Leopard to a significant degree of improvement? I'm thinking that the delay in shipping Leopard is reason enough to wait for SS with Penryn as it will be shipping before end of year about the same time as Leopard. Is that stinkin' thinkin'?
Exactly right. I want all the brains that the 8 cores of power need and I'd rather have that bigger 12MB cache in Penryn as well now that Leopard is delayed.
I'm thinking that the delay in shipping Leopard is reason enough to wait for SS with Penryn as it will be shipping before end of year about the same time as Leopard. Is that stinkin' thinkin'?
What I'm saying is that you may be far better off getting the current system and crunching at high speed, rather than poke along on your quad for a year waiting for Penryn Xeons.
Cinebench scores from BF:
Quad G5: 1105
Quad Xeon: 1601
Octo Xeon: 2323
Quad Xeon: 1601
Octo Xeon: 2323
The current octo is 2.1 times faster than the quad. By January or so, there may be a Penryn that's 2.3 or 2.4 times faster. (I'd suspect on your multi-threaded workflow, Clovertown would do pretty well - so that Penryn would be less of a boost.)
You have to decide whether it's better to wait a year at 1X, or run 2.1X faster for that year.
If you don't push the buy button, I can only assume that you really don't need the added speed. Otherwise, you'd buy.
http://www.barefeats.com/mvdcpc.html
http://www.barefeats.com/octopro1.html
What are you talking about? I bought a 300 GB hard drive years ago. The biggest ones today are 750 GB (I think I even heard of a 1 TB drive being available).
Yes.... by Hitachi... going to order two of them when I get my MacPro with 8 cores.
Hopefully we'll be seeing product refreshes soon. Jobs went from Core Duos to Core 2 Duos pretty fast, but there hasnt bee anything since. Hoping with new iMac looks come fast processors..whenever that happens.
I think what Bare Feats refers to as a memory bandwidth bottleneck is not a FSB bottleneck as people tend to point to but a memory controller bottleneck. In truth, a 1333MHz FSB is perfectly sufficient for all but the most extreme cases for a quad core chip. If you compare a 1.8GHz E4300 on a 800MHz FSB to a 1.86GHz E6300 on a 1067MHz FSB, there is actually very little performance difference despite the smaller FSB. Similarly, even comparing a 667MHz Merom to a comparable clock speed 1067MHz FSB Conroe yields very similar results. The Merom architecture is simply a lot more memory bandwidth agnostic than Netburst was and the 4MB L2 cache certainly helps.Exactly right. I want all the brains that the 8 cores of power need and I'd rather have that bigger 12MB cache in Penryn as well now that Leopard is delayed.
I fully respect your superior knowledge on all things technical Aiden. Are you saying that the confusion that Bare Feats finds among those cores in the "new" 8 core MP won't be mollified with SS & Leopard to a significant degree of improvement? I'm thinking that the delay in shipping Leopard is reason enough to wait for SS with Penryn as it will be shipping before end of year about the same time as Leopard. Is that stinkin' thinkin'?
The problem though is with the FB-DIMMs. They have terrible scaling and tests have shown that a quad channel DDR2 533 FB-DIMM setup with 4-4-4 timings performs like a dual channel DDR2 667 setup with 5-5-5 timings. This is why Conroes tend to outperform Woodcrests in dual threaded applications despite Woodcrest having a bigger FSB.
Seaburg will introduce a 2nd gen MC that should improve things. Sadly, there won't be DDR2-800 FB-DIMMs so any performance increase will have to be from the architecture. Intel has stated that they've been focusing on reducing memory latency which is good. Their tests have shown that Seaburg provides a 5% performance increase using the exact chips compared to Blackford/Greencreek. Seaburg also has a larger snoop filter optimized for quad cores to reduce cache coherency traffic which will help.
Penryn will bring 1600MHz FSBs, but in light of the memory controller bottleneck that will still remain with Seaburg, I don't see that as being a big feature. Besides, with only quad channel DDR2 667 FB-DIMM support, there isn't enough theoretical bandwidth to fill dual 1600MHz FSBs anyways. I'm not sure how beneficial the bigger caches will be anyways. Intel's actually skimping since they're only going with a 50% increase instead of the usual doubling so that already limits improvements. The major benefit of Penryn though ironically seems to be SSE4. The Divx demo they used that was rewritten with SSE4 support shows Penryn more than doubling the performance of Kentsfield. I'm sure most video applications will add support for SSE4 whether on Mac or PC, so that will definitely provide a nice boost.
..because other then Quad-Core processors not much new stuff is available at this time... soon.
Not totally true, the Mac Pros went from quad core's to 8 cores. But I dunno if that's what you're talking about...
still new proccessor = bada$$.
There are other factors that enter in though-- cost and software being two of the big ones. You're quoting the Cinebench test results, but MM seemed most interested in video encoding performance. The BF tests show no improvement on Quicktime export performance between the 4 and 8 core machines (unless you're running 6 simultaneous jobs exporting the same file to the same format...).
I don't know what's holding up QT, it's probably the applications' treading performance, but it doesn't make sense to pull the trigger on an upgrade until the new system does better at what you want to use it for. Rather than dumping a chunk of cash and then waiting for the software to catch up, or finding that there's a critical bottleneck in the system preventing it from performing your task efficiently, sometimes it's better to wait. The hardware will keep advancing while the software people make the changes they need to.
Was mainly talking about the iMac's and Mini's. But lookin on how fast we got the Core 2 Duo's and Octo cores, we should be get these chips aswell pretty fast (lets hope). I wonder if the iMac & mini refreshes could be tied in when they processors come out
yeah the mini is looooooooooooong overdue for a proccessor upgrade. They should have put the core 2 duo in it! maybe they'll just put a quad core xenon in it!! haha.
I wonder if Apple has any Cell Processor or Power 6 based prototype Macs in their labs
B
You can count on that. Just as they always had secret intel macs, they will have prototypes running on any other significant processor, AMD included.
As soon as one of the technologies is proving to be better than the current intel roadmap, then we could well expect another 'transition'.
Hence, it will not happen overnight, Apple will look for the bigger picture, and need to know a roadmap that is worth investing in, not just for 2 years and we switch again, it will have to be solid info, worth at least 5 years if not longer.
I thought that was obvious when I said 300GB, when we're at 1TB for 3.5" HDs
but thanks for clarifying.
MM's typical task is multiple parallel rip/encode/decode jobs.... The 8 simultaneous jobs (or four that can do 2-way threading) is what he'll be doing.
This makes sense, as long as you don't need the performance.
If you *need* the performance, then you hit the buy button regardless.
For example, if it takes you 28 hours to encode a day's worth of video work - you hit the buy button. If it takes 20 hours, then you can contemplate different purchase choices.
_________
Anyone who is considering waiting for Stoakley doesn't need the performance. Anyone who considers waiting for Penryn clearly doesn't need a Penryn.
They might want the extra performance, or some extra longevity - but they don't *need* the few extra percent.
On the other hand, though, I'm going to hit a big "buy button" when Stoakley arrives, because I have some projects that *need* VT-d and PCIe.V2. These can't even start until the new systems show up.
Agree but I add one more month, Santa Rosa is available in May, so Apple may adopt it for Jun, July.
For the Penryn I would add 2 months, say April 2008.
Not an expert on this but I saw claims of 32 or 34 nm for the next round of chips.
To my knowledge that is the process size and not the size of the chip. Intel has the actual chip size on their web site under their technical specifications.
No need to correct me, I know I know little to nothing about this subject, just quoting stuff I read previously (and probably badly at that).
LOL. Nope its the pianist/organist of my favorite band.
Im at Vanier. Are u in university?
Errr.... my bad.
But you are right from a purely processor standpoint Penryn is worth more overall, but as a general update I think Santa Rosa will allow for faster speeds even if just on a processor bases Penyrn will be better.
I love Intel, they are soo fast when it comes to updating, but then it makes my MB CD look slow, but i love it anyway.