Yup if you shelled out for a Socket 1366 (high end i7), you're going to be sticking with Nehalem until Socket R comes out down the line.
If you went with 1156, which I did (P55 Classified + i7 860 @ 4.0 Ghz), then you're screwed, just earlier, since it's now Socket 1155, which isn't compatible even though it's just a 1 pin difference.
I wasn't very happy with Intel when I found this out, since they've recently switched sockets after holding on to 775 for so long, but from my understanding AMD has also done something with the AM-2/3 socket where some motherboards are back/forwards compatible, but others aren't. I think there is a derivative socket, Am-2/3+, that is backward compatible but the Am2-3 standard version isn't forwards compatible. Don't take my word on it though, my builds have been Intel since the Q6600 came out. AMD has done a better job of backward compatibility but the sweet spot for price/performance + overclocking has been Intel chips whenever I've done my last few builds, and I only do builds every few years, usually after new architectures are released so my motehrboards are usually replaced as well.
Anandtech covered upcoming socket changes in more detail in their writeup
I agree on the myth of HSPICE vs Spectre -- my models were SPICE-syntax (22nm PTM BSIM4 models) so it's just a couple of syntax changes to get them into Spectre.
My results found that Spectre fit much closer to the predictions, while HPSICE's were a bit farther out on simple process characterization tests, as well as on simple designs (FO4 Inverters, etc.). Since Spectre matched the trends closer, I'd definitely have to give it the nod.
The increasing foundry support is definitely a major plus, though all the classes here are taught on Cadence's gpdk090 since it provides full models, does a lot of the calculations automatically so that nobody is forced to learn SKILL on their own to speed things up, and provides layouts for transistors and lots of extras/integration into Assura, etc.
I didn't mean to come off against Spectre -- my experiences with it have been great -- most of the negatives I heard came from some engineers on DeepChip. Personally, I was most impressed with Ultrasim -- it maintains about 98% accuracy on both power consumption and delay of full chips versus Spectre with about 10x speedup on small (~15k transistors) designs and even more significant gains on larger designs. I've tested up to ~1 million transistors with ease on Ultrasim even with large amounts of mismatch across devices.
I just spent the past summer doing research at the 22nm level (designing L1/L2 caches with DVFS and other low-power techniques) and I can't agree more on SPICE/HSPICE's inability to converge.
I shrunk my designs down to the criitical paths (~12k transistors), and even providing the proper nodesets/initial conditions HSPICE was unable to converge or segfaulted quikcly. Fortunately, my university has a deal with Cadence through their University Alliance program -- Spectre may not be quite as accurate at HSPICE for analog circuits, but both it and Ultrasim (a FASTSPICE simulator for large designs) can handle much larger digital designs without complaint.
To the original submitter: Is there a good reason behind the no network connection requirement? If the university has a proper setup, students should be fine either on or off campus -- then it may be worth checking if your university has any deals with either Cadence, Synopsys, or Magma -- their tools are primarily Unix-based (Solaris, AIX, and Linux support), so it's just a matter of having the students SSH in with X forwarding or use VNC. This would even allows users with underpowered machines to simulate large designs quickly since everything is done remotely. I primarily run Windows on my local box, but either VMs with Linux or using Putty with Xming work properly for all these tools.
Slashdot Mirror
Yup if you shelled out for a Socket 1366 (high end i7), you're going to be sticking with Nehalem until Socket R comes out down the line.
If you went with 1156, which I did (P55 Classified + i7 860 @ 4.0 Ghz), then you're screwed, just earlier, since it's now Socket 1155, which isn't compatible even though it's just a 1 pin difference.
I wasn't very happy with Intel when I found this out, since they've recently switched sockets after holding on to 775 for so long, but from my understanding AMD has also done something with the AM-2/3 socket where some motherboards are back/forwards compatible, but others aren't. I think there is a derivative socket, Am-2/3+, that is backward compatible but the Am2-3 standard version isn't forwards compatible. Don't take my word on it though, my builds have been Intel since the Q6600 came out. AMD has done a better job of backward compatibility but the sweet spot for price/performance + overclocking has been Intel chips whenever I've done my last few builds, and I only do builds every few years, usually after new architectures are released so my motehrboards are usually replaced as well.
Anandtech covered upcoming socket changes in more detail in their writeup
I agree on the myth of HSPICE vs Spectre -- my models were SPICE-syntax (22nm PTM BSIM4 models) so it's just a couple of syntax changes to get them into Spectre.
My results found that Spectre fit much closer to the predictions, while HPSICE's were a bit farther out on simple process characterization tests, as well as on simple designs (FO4 Inverters, etc.). Since Spectre matched the trends closer, I'd definitely have to give it the nod.
The increasing foundry support is definitely a major plus, though all the classes here are taught on Cadence's gpdk090 since it provides full models, does a lot of the calculations automatically so that nobody is forced to learn SKILL on their own to speed things up, and provides layouts for transistors and lots of extras/integration into Assura, etc.
I didn't mean to come off against Spectre -- my experiences with it have been great -- most of the negatives I heard came from some engineers on DeepChip. Personally, I was most impressed with Ultrasim -- it maintains about 98% accuracy on both power consumption and delay of full chips versus Spectre with about 10x speedup on small (~15k transistors) designs and even more significant gains on larger designs. I've tested up to ~1 million transistors with ease on Ultrasim even with large amounts of mismatch across devices.
I just spent the past summer doing research at the 22nm level (designing L1/L2 caches with DVFS and other low-power techniques) and I can't agree more on SPICE/HSPICE's inability to converge.
I shrunk my designs down to the criitical paths (~12k transistors), and even providing the proper nodesets/initial conditions HSPICE was unable to converge or segfaulted quikcly. Fortunately, my university has a deal with Cadence through their University Alliance program -- Spectre may not be quite as accurate at HSPICE for analog circuits, but both it and Ultrasim (a FASTSPICE simulator for large designs) can handle much larger digital designs without complaint.
To the original submitter: Is there a good reason behind the no network connection requirement? If the university has a proper setup, students should be fine either on or off campus -- then it may be worth checking if your university has any deals with either Cadence, Synopsys, or Magma -- their tools are primarily Unix-based (Solaris, AIX, and Linux support), so it's just a matter of having the students SSH in with X forwarding or use VNC. This would even allows users with underpowered machines to simulate large designs quickly since everything is done remotely. I primarily run Windows on my local box, but either VMs with Linux or using Putty with Xming work properly for all these tools.