Intel Developer Forum 2007 - Day 1: More Nehalem, Penryn Announced, and Gelsinger Speaks
by Anand Shimpi & Larry Barber on September 18, 2007 8:03 PM EST- Posted in
- Trade Shows
The Transistor Issue: Nehalem Smaller than Penryn?
We published Nehalem transistor counts earlier today and wanted to clarify one point: the 731M transistor figure was for quad-core Nehalem. With quad-core Penryn weighing in at over 800 million transistors, there's a slight reduction in transistor count for Nehalem and the explanation is simple: cache size.
Quad-core Nehalem pictured above, supposedly 8-core versions are in the works
With Nehalem Intel could conceivably go to an architecture with smaller L2 caches per core and a large L3 cache to feed all four cores on the same die. High end Nehalem cores will also sport an on-die memory controller, reducing the need for a ridiculously large L2 cache, and thus we may see an overall cache size reduction going from Penryn to Nehalem while still yielding a performance increase.
Facebook
Twitter
RSS
8 Comments
View All Comments
GhandiInstinct - Wednesday, September 19, 2007 - link
Kurt cobain would be proud...Intel isn't getting lazy cause its on top, it's doing what a great company should do, trounce and eliminate all competition while it's customers get the cheapest and greatest technology.
GO INTEL!!!!
Now with USB3.0 even more reason to buy an ultimate rig from my AGP in 2H 2008.
DigitalFreak - Tuesday, September 18, 2007 - link
So, if it truely is 10x the performance, that would equate to 4.8Gb. Wow, who really needs ESATA then if CPU utilization is minimized? (Yes, I know ESATA will still have it's place).spinportal - Wednesday, September 19, 2007 - link
If USB 3.0 chipsets can forgo the saturation and polling deadlocking of the southbridge (data channels for SATA, IDE, USB, Ether(giga)net, PCIe bus? vs Northbridge on memory?) for competition of bandwidth, this kind of speed is worthwhile. What is the point of plugging in a USB device and lockstepping your OS's multithreading while it's negotiating? What's the point of throttling a USB channel, when your other channels tank, starve and choke?psychobriggsy - Wednesday, September 19, 2007 - link
I somehow suspect that CPU utilisation will not be minimised via "USB-offload-engines", so expect 4800Mbps USB3 to require quite a bit more CPU resources. If you want an interconnect with lower CPU utilisation, then there is always Firewire, although USB3 promises to put the freeze on that technology. Firewire needs to get its act together methinks.luxor - Tuesday, September 18, 2007 - link
Anand, these pics are blurry. You know this. We would love to be able to read some of these slides.Tristesse27 - Thursday, September 27, 2007 - link
BTW, your pictures aren't blurry because you're sitting in the back. You're pictures are blurry because you have the wrong camera for the job.rqle - Wednesday, September 19, 2007 - link
IDF is by far the most entertaining conference of them all.mattsaccount - Tuesday, September 18, 2007 - link
"Johan made us late to the keynote and we had to sit in the back."Bad Johan!