|
OK, to all the AMD fanboys out there. Games are made for dual cores. Intel have more powerful cores. Hence, Intel have superior gaming CPUs. The person is looking for a CPU and SC2 is the benchmark he's using. Recommending an AMD CPU is therefor the wrong thing to do. AMD could release an 18 core CPU for $100, and it would still run SC2 slower than a Core i5.
The X6 architecture is great for things that are heavily multithreaded such as video transcoding and some forms of 3D processing, but these do not include games, office applications and browsing. If you run a anti-virus scan, while using every single internet security software, archive data, browse, and run Office, you will still not need more than a quad-core, and the Phenom II architecture will perform worse.
Phenom II X6 are generally power-guzzlers, which make for a lot of heat in the case (and the need for more cooling and noise), and the need for a bigger PSU. I would recommend high core count CPUs specifically for people who know how to use them and need to do so. But to most people, a high core count is practically useless.
|
1090T user here.
Just like you, I wasn't sure what to buy, I just knew I needed a lot of cores for special applications ( 3d, compositing and rendering tasks )
In that particular case, the Phenom is a little better than 4 cores i5, i7 in performance and money. I'm quite happy with it.
If you"re not doing anything that uses a lot of cores ( like gaming ) then just go for the cheapest route.
|
Like the others said, wait and see.
But remember you are comparing an already released cpu to one that isnt even out yet, so the newer one will almost always be faster. Keep in mind that AMD also have unreleased CPU's which should be coming out soon I believe? Or at least hope. It's not really fare to compare old cpus vs new cpus.
|
@Kazius
Many games use three cores today (especially X360 ports because that has a tri-core CPU). The Athlon II X3 fares well against Intel duals for the price, even in game performance. I agree that hex-cores won't, but there are plenty of uses for a CPU outside of gaming that it is competitive in.
The trend in CPUs is more cores. Games will catch up over the next few years, they are just starting to use 4 cores.
Bulldozer will solve that, it is coming soon. AMD will again be competitive on single-core performance AND have more cores than Intel (eight).
|
Sandybridge-E will have 8 cores and 8 threads (16 total) so no AMD won't have more cores than Intel.
|
I really don't think Bulldozer will match Sandy Bridge in single-threaded performance, though we're all just speculating at this point. Also, we don't have a clear picture of what pricing is going to be like. But if you look at the architectural changes, Intel's focus is clearly more on single-threaded performance (more relevant for most applications including games), while AMD's focus is on more cores and aggregate throughput per cost.
If you run applications that are well-threaded, Bulldozer priced right looks like it should be great. The way Bulldozer modules are organized--essentially, two integer execution cores sharing a front end and a floating point unit--they should be very efficient at handling parallel integer workloads.
edit: The articles are old, but Kanter's articles on Bulldozer and Sandy Bridge may be worth reading:
Bulldozer
Sandy Bridge <- the conclusion here has a (speculative) comparison between the two
|
On December 21 2010 06:20 skyR wrote:
Sandybridge-E will have 8 cores and 8 threads (16 total) so no AMD won't have more cores than Intel.
I haven't seen anything official saying SB 8-cores will come to the desktop. Server, certainly. But I believe they will be confined to high-end, 4-socket platforms.
AMD will have a 16-core chip on the server, again 4S only.
@Myrmidon
Agreed that it's just speculation now. However it must be more competitive than the current situation. And AMD will always price it like for like whatever the matchup.
BD's insane clockspeeds may help it a lot on single-core. Dresdenboy (someone similar to Kanter in technical knowledge) was predicting 4GHz.
|
5930 Posts
Intel plans to phase out the dual core with the quad core and make octo cores (???) the quad core of today with Ivy Bridge. Of course that's a die shrink (22nm) and who knows what could go wrong in the process but Intel is definitely planning to do that.
Not worth that much since its pretty far away but something to take into account I guess.
|
If you can afford it, Intel over AMD any day of the week. My i7 920 hit 4.2ghz over its stock 2.66ghz without breaking a sweat with HT enabled no less. Triple channel memory over dual channel, any day of the week. AMD is always a generation behind.
|
On December 21 2010 10:01 Womwomwom wrote:
Intel plans to phase out the dual core with the quad core and make octo cores (???) the quad core of today with Ivy Bridge. Of course that's a die shrink (22nm) and who knows what could go wrong in the process but Intel is definitely planning to do that.
Not worth that much since its pretty far away but something to take into account I guess.
Yes, a 22 nm CPU would be an impressive process achievement. 0.5 nm is the physical oxide thickness limit, and 22 nm scaling would require it. They have to replace the CVD silicon oxide above metal1 with exotic materials to lower the capacitance, a difficult task to fabricate. If they can succeed it will lower a lot of barriers to even smaller feature sizes.
I am fairly shocked that the major companies haven't tried anything with GaAs logic yet, some university projects have successfully integrated RISC instruction sets on GaAs, and the potential clock speeds from GaAs are huge. It does have its problems-namely its huge electron mobility comes at a price of really slow hole mobility, making CMOS next to impossible to implement(though nmos is really damn fast on it), which really increases power usage. Researchers have achieved 110GHz amplifier operation with GaAs MMICs (in the microwave world we have mostly moved to GaAs), and experimental GaAs logic circuits have been published as having 3-5x the speed of equivalent silicon ICs.
|
|
|
|
|
|
EPT
