AMD: Opterons to hit 20 cores by 2012 Advanced Micro Devices can't beat Intel's wafer-baking process in the ongoing x64 server chip war, so it's going to fight with the weapons it has: higher core counts and better bang for the buck — or watt. That was the message delivered by Don Newell, the new chief technology officer for AMD's server-chip business when …
but can it run X264 well!
according to http://www.phoronix.com/scan.php?page=article&item=intel_corei7_970&num=4
even the current dual AMD Shanghai workstation (that's a dual socket CPU 8 core workstation board) AMD Does Not perform well on many REAL LIFE workload problems today and i dont think the new generation of AMD chips will match or out perform any sandy bridge in the same dual socket configuration, as even the Phenom II x6 cant match with i7 in the FPS encoding throughput x264 stakes.
its a shame, as i d like to actually see a lot more 2 socket AMD boards and CPUs that can at least keep up or better the real life x264 FPS and other real life application data throughput, no AMD for people shopping lists for the long time yet it seems.
ooh well we will have to made do with the 30% better than current i7 sandy bridge with generic internal video Encode/Decode engine and 10 gig "Ethernet over Light Peak" cards and kits coming early next year, and perhaps a few cheap dual socket boards for consumer use to put them in
what's with the bullshit 2 channel DDR3 2011 roadmap today, everyone knows you Need 4 channels (or even more) as standard today on all new CPU's/boards if you intend to gradually upgrade your PC over time.
there's nothing more irritating than having to rip out perfectly good ram and replace it with a higher
density stick because the motherboard and CPU vendors didn't see fit to include and allow more slots etc.
Both AMD & Intel can keep giving us more, but if the OS kernel isn't tuned for the scheduling of the resources, that over head limits you on the data you can push. Windoze loses. Linux in the middle. Solaris has been doing it for years. Sorry FreeBSD folks, don't know your kernel.
Soon, very soon, the x86 cpus will be the fastest on earth. First generation Bulldozer will be closely matching POWER7. But for a fraction of the price. You will be able to get 3-5 x86 servers for the same price as one POWER server.
In 2012 you will get 20-core Bulldozers - which will easily crush 8-core POWER7.
And let us not forget the Oracle/Sun server with 16.384 threads and 64TB RAM in 2015. :o)
Do you really think so? Actually, it would not surprise me if you thought so, as you have showed earlier, your reasoning is a bit weird.
For instance, you earlier claimed that even though you need four POWER6 at 5GHz to match two Intel Nehalem 2.93GHz at TPC-C benchmarks - the POWER6 is faster. Because the POWER6 core are faster. Great. Ask Jesper Frimann which CPU is fastest for TPC-C. And he will say: "Why, Power6 of course! You only need twice the amount of POWER6 cpus to reach a certain performance number, but ignore that fact. The POWER6 is faster than x86"
In the same vein, of course you need only one computer per continent, Jesper.
Great, Jesper! Continue with your thinking cap.
You need to learn several things my young friend.
1) Irony
2) Something about computer history.
3) Checking your facts.
Nope I would say that the fastest CPU is POWER7. I would even say that the fastest chip is POWER7. Although I do think that the T3 might be a little faster doing something like RSA en/decryption, due to it's special accelerators for that particular workload.
Now I wouldn't argue against that the Nehalem-EP chip at 2.93GHz with 4 cores is faster than a POWER6 Chip with 2 cores. The POWER6 core is still faster than Nehalem. And POWER7 which have shown up to 150K tpmc per core is of cause in a league of it own.
Not that there are any relevant TPC-C benchmarks to compare against at 5GHz as you try to insinuate. You really should check your facts.. but hey..
// jesper
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
