Quote:Why does it matter? It won't affect gpu performance and/or scaling by more than 1-2%. Benchmark sites have already thrown the x8x8 makes a difference versus x16x16 and integrated pci-e makes a difference vs. external pci-e arguments both out the window.
NF200 chips have a volatile history, particularly in creating issues with Creative sound hardware.
Why would the interconnect matter? Ｉf anything, sandy bridge would be faster because pci-e is on die, instead of relying on qpi (on x58) to get pci-e. Both platforms use dmi to connect to the southbridge anyways. And as I said before, Ｙou CANNOT compare performance in a benchmarks like that. Openchess is a heavily multithreaded benchmark, so it's pretty obvious that the 980x will be faster, because it simply has two more cores. Find single threaded comparisons, and you'll see sb is ~10% faster clock for clock.
i don't care about ondie pci-e, you can choke on your limited bus, which has to pass through a nf200 bridge chip which is slower than QPI > x58 anyway and adds additional latency.
i'll wait for
Sandy Bridge E, with its true 40x lanes and PCI-E 3.0 compliancy.
(01-13-2011, 01:42 PM)boogerlad Wrote: [ -> ]Why would the interconnect matter? Ｉf anything, sandy bridge would be faster because pci-e is on die, instead of relying on qpi (on x58) to get pci-e. Both platforms use dmi to connect to the southbridge anyways. And as I said before, Ｙou CANNOT compare performance in a benchmarks like that. Openchess is a heavily multithreaded benchmark, so it's pretty obvious that the 980x will be faster, because it simply has two more cores. Find single threaded comparisons, and you'll see sb is ~10% faster clock for clock.
He's just trying to defend his 1st gen Core i7 system. The facts are in the benchmarks. One or two synthetic benches got nothing on Sandy Bridge prowess performance.
(01-13-2011, 02:16 PM)tuanming Wrote: [ -> ] (01-13-2011, 01:42 PM)boogerlad Wrote: [ -> ]Why would the interconnect matter? Ｉf anything, sandy bridge would be faster because pci-e is on die, instead of relying on qpi (on x58) to get pci-e. Both platforms use dmi to connect to the southbridge anyways. And as I said before, Ｙou CANNOT compare performance in a benchmarks like that. Openchess is a heavily multithreaded benchmark, so it's pretty obvious that the 980x will be faster, because it simply has two more cores. Find single threaded comparisons, and you'll see sb is ~10% faster clock for clock.
He's just trying to defend his 1st gen Core i7 system. The facts are in the benchmarks. One or two synthetic benches got nothing on Sandy Bridge prowess performance.
any review finding Sandy bridge to be 10% faster, is doing it wrong.
Then maybe you should go tell all of the authors who wrote the reviews from SB, tell them "Hey! You're doing it wrong!".
any application that only needs a single thread, can probably already run just fine on a single core Atlon64 3200.
any application that can't, is probably written by Rockstar, or Activision.
I was looking at the passmark charts and found something odd - The 2600k beat the 2600 by over 1000 PTA, but the i5 2500 k and non k versions got almost identical scores. Why's that ???
Quote:NF200 chips have a volatile history, particularly in creating issues with Creative sound hardware.
NF100 yes. NF200 no. I've been using nforce boards with NF200 for 4 years now along side my creative x-fi xtrememusic edition in linux mint, windows xp, and windows 7. Still no problems.... All of the people who complain about nforce problems are using old nforce boards. Besides X58 also uses NF200.....so if you hate it that bad I guess you lose either way.
Edit: We discussed this on irc and determined it was because I was using PCI instead of pci-e so you can ignore this.
Quote:any application that only needs a single thread, can probably already run just fine on a single core Atlon64 3200.
any application that can't, is probably written by Rockstar, or Activision.
Or 2-3 threads. As long as it's 4 or less sandy bridge will DESTROY gulftown. Dolphin for example, and all emulators for that matter. It's pretty hard to come up with a real world scenario where gulftown will win (even though it costs 4 times as much and has 6 cores). Sandy bridge still manages to win 99% of the time in all of the applications that people actually use. I'de dare say their are more applications out their with 2-3 threads that are in need of more performance than highly parallel visual applications.
It's just flat out better hardware. It destroys gulftown in most areas for a fraction of the price and even if it DIDN'T it would still be a much better buy than bloomfield or lyynfield for the same price.
@fb39ca4
Unlocked turbo + pass mark doesn't benefit from HT.
On a more on topic note:
http://www.anandtech.com/show/4113/lucid...ndy-bridge
(01-13-2011, 02:37 PM)fb39ca4 Wrote: [ -> ]I was looking at the passmark charts and found something odd - The 2600k beat the 2600 by over 1000 PTA, but the i5 2500 k and non k versions got almost identical scores. Why's that ???
Most likely because that program is designed for multi-core/threads which takes advantages of the extra cores and threads, therefore you'd see bigger difference in 4 core with 8 threads than 2 core with 4 threads.