AMD One-Ups Intel With Cheap Desktop Chips 362
CWmike writes "Advanced Micro Devices on Tuesday announced inexpensive desktop microprocessors with up to six cores to put pricing pressure on rival Intel. AMD's new chips include the fastest AMD Phenom II X6 1075T six-core processor, which is priced 'under $250' for 1,000 units, AMD said. AMD also introduced a range of dual-core and quad-core Athlon II and Phenom II desktop microprocessors priced between $76 and $185. By comparison, Intel's cheapest six-core processor is the Core i7-970 processor, which is priced at $885 per 1,000 units, according to a price list on Intel's website."
That would be all nice and dandy if only... (Score:-1, Informative)
If only for the fact that their 6 cores at 3.0Ghz+ are slower than 4 i7 cores at 2.4Ghz, and their 6core chip costs more than the i7 920. So no, they are not "One Uping" anyone. Not to mention, you can overclock that i7 920 to 3.5Ghz, and it leaves the 6 amd cores at 4Ghz in the dust.
For highly distributed, memory intensive tasks, like AI and Simulations, i7 performs better than AMD's 6 core variants. The fact that you get 3 memory channels on the 1366 board also helps.
AMD One-Ups Intel? Another misleading Slash story. (Score:5, Informative)
Pricing error in this article (Score:4, Informative)
Re:Technology (Score:3, Informative)
next year AMD will be launching its "bulldozer" architecture, which from what i've seen takes a rather novel approach to cores/hyperthreading (two 'cores' which share some execution units. I'm not saying bulldozer will suddenly revolutionize anything, but it is an interesting take on multi-core
as for your am2 cpu, yeah man, if your board takes a am3 cpu, go for it, you can pick up a quad core (which will trounce whatever you have in single threaded performance too), for under $100, if you still have a single core am2 chip, put in a $60-70 X3 and be amazed!
Initial review... (Score:5, Informative)
Overclockers.com has a review of the Phenom II x6 1075T processor. Looks like it's got pretty good overclock potential and performs well against similarly priced Intel chips.
http://www.overclockers.com/amd-phenom-ii-x6-1075t-review/ [overclockers.com]
Re:Legacy Implementation (Score:3, Informative)
4x PCI slots for legacy video capture equipment but fast processor for encoding
hmm, when I go on newegg the most PCI slots they sell on an AMD board or a current gen intel board is 3 while they have LGA775 boards up to 5 PCI slots.
Re:The thing I think you miss (Score:3, Informative)
"performance" by itself doesn't mean anything. You have to refer to "performance per dollar" or "performance per watt" or ... that's why you don't see everyone buying the top of the line CPU... There's a bunch of stuff that offers the highest performance, and that doesn't really sell much, if at all: fighter jets, formula 1 cars, thoroughbreds... We might feel all sexy at the idea of owning one of those, but the bare fact is, we can't afford it, and, when push comes to shove, we would be stupid to, anyway.
talking about perf/price, AMD is not that bad, especially for run-of-the-mill levels of performance, and once you take the MB cost into account (why are Intel's MB so much more expensive ?). On a fixed budget, I'd rather scrimp on the CPU/MB, and put more into the GPU and disks, because that's what limiting my PC, right now.
talking about perf/watt... I'm not running a server farm, I don't care. Intal is ahead, though.
Re:And 3 hours after reading this... (Score:1, Informative)
Re:And 3 hours after reading this... (Score:1, Informative)
Intel still charges insane money for their chips.. they only lower it to a fair price after AMD forces them ...
Re:? Do you really think Intels are 4x faster (Score:3, Informative)
But AMD provides the best performance for value. If you buy a 200euro amd you get the best bang for your buck. If you buy a 800 euro Intel you get more bang but pay more bucks per bang.
The Core i7-860 spanks everything AMD has at $280 @ newegg, there's only a few odd benchmarks AMDs $300 top six-core CPU wins. Then entire market from $250+ and up is Intel, Intel, Intel. The $100 market AMD wins, but their value gets worse the closer you come to the high end. You make it sounds like Intel only owns the Ferrari market, when in reality they own the whole $50,000+ car market.
Re:cache difference (Score:3, Informative)
AMD have no high end
Ever heard of the Opteron, particularly the 6100 series [wikipedia.org] released in March? 12 cores from 1.9GHz to 2.2GHz with 115W TDP or you can go up to 137W TDP and 2.3GHz or drop to 1.7GHz and closer to 65W TDP.
Re:Legacy Implementation (Score:4, Informative)
Re:? Do you really think Intels are 4x faster (Score:3, Informative)
You should also consider the price of the motherboard. Core i7 motherboards are very expensive.
Re:? Do you really think Intels are 4x faster (Score:2, Informative)
Will it require $50 activation like Intel? (Score:5, Informative)
Frankly if they don't do the cheesy-as-hell activation fee like Intel is proposing [thetechherald.com] I'm sold.
THAT'S BECAUSE INTEL CHEATS! (Score:4, Informative)
Some of the benchmark programs are compiled with Intel's C++ compiler, which generates CPUID checks for the manufacturer string 'GenuineIntel' and redirects all other manufacturer's CPUs to the slowest code path. So if you can't compile the benchmark yourself with a trusted compiler, its not worth the paper its printed on.
Intel also releases several libraries that other software vendors use in their products; these libraries contain the same manufacturer check [agner.org] which cripples their performance on chips by AMD, Via, etc. Commercial software products such as Matlab have unintentionally or intentionally shipped with these checks, with the result that they run slower than necessary on AMD CPUs. When the manufacturer test is patched out of the program, it is un-crippled and runs as fast or faster than a comparable Intel chip.
Intel settled out of court with AMD over this, and are in the process of also settling with the FTC, but have not actually stopped the practice. [agner.org]
Paying of OEMs is not their only trick.. (Score:5, Informative)
For many years, the Intel C++ compiler has discriminated against non-Intel chips by detecting their manufacturer using CPUID and redirecting all chips not manufactured by 'GenuineIntel' to a slower code path. (And that manufacturer ID is their trademark so other manufacturers may not use it). C++ libraries available from Intel (such as their math libraries) also contain the same discriminatory code checks. This artificially decreases their performance on AMD's chips.
It makes Intel chips look better, by slowing down the program on all of their competitor's chips. So the safest thing is NOT to use Intel's compiler for anything (most especially benchmarking). This is a problem because it has a reputation of producing larger, but faster, code. (Faster on Intel processors at least!).
The code it produces is actually quite decent on AMD chips too, as long as you patch out the generated version checks to un-cripple the performance on AMD chips. You can do it as a post-build step after compiling. It's a hassle that most software vendors don't bother with -- in most cases they aren't even aware that Intel's compiler generates the manufacturer-checks and redirects their program through slower code paths on AMD chips.
Re:And 3 hours after reading this... (Score:3, Informative)
EXCEPT:
The intel motherboard for those less expensive CPUs has zero forward-looking value. They are already obsoltete when you buy them now. Additionally, the memory on the board will be obsolete DDR2. Furthermore, there will be no USB3/SATA3 option.
The AMD board will be closer to current, can have DDR3 should you choose it, and will support AMDs newest quad core/six core processors currently out.
This is like me pointing out that a prius makes a hella good race car, because it can get over 15 laps out of a gallon of gas. A PC is a bigger picture than the CPU.
Not really an apples to apples comparison. (Score:3, Informative)
The i7-860 costs a lot more than a the X6 1075T
Sure the chip is only $20 more $250 vs $270 however you need a motherboard. i7 MB are notoriously expensive.
One can find a decent crossfire/SLI capable AM3 motherboard for $80 - $100. Not so with i7. Prices start at $160 and tier 1 brands are more like $180 - $200.
When you consider the additional cost of the motherboard your i7 solution is running 30% higher than the X6 platform. 30% higher for maybe 10% more performance.
Performance per $ is what matters. Sure you can get that i7 but it will cost you $100 more. Someone buying an X6 could spend that $100 on a better GPU making an overall better system.
Pure power Intel has always been a leader but many of their solutions fall apart in a price per value metric.
Re:Will it require $50 activation like Intel? (Score:2, Informative)
They still do this, and it's a nice feature :)
They ship completely loaded out machines, then we use Capacity Upgrade on Demand to enable the cores via software (though the HMC console). It sounds like idiocy to do this, but when this means that during our peak season we can enable/disable cores and save literally hundreds of thousands of dollars versus buying a fully loaded and enabled machine, it makes sense.
Keep in mind that this is not only for the hardware. Oracle licensing is HIDEOUS. We pay millions to license their database across our enterprise. We also pay $500K and more for Tivoli software (an IBM company). We pay for scheduler software. All these are based either on the processor capacity, number of processors, memory size, disk size, number of IO adapters, etc.. So adding a few cores here and there has an immediate impact on our licensing costs.
Re:? Do you really think Intels are 4x faster (Score:3, Informative)
I'd actually gone a less conventional route to start, then tried to fit it more in with the scheme. However, I find banging things to be a more entertaining image than bucking them, so I'm happy with my choice.
Re:? Do you really think Intels are 4x faster (Score:3, Informative)
Re:And 3 hours after reading this... (Score:3, Informative)
The features are small, but often quite worthwhile.
For example, in Word 2010, you can choose based on the program you're copying from how you want styles to be handled, whether to use the original style, local style, or paste as plain text. That saves me tons of time I would have spent messing with manually choosing paste type before.
2010 also has built in "dataleakage" detection, warning you about metadata you might be sending on accident.
The last version of Word I used was 2003, then switched to Open Office (which I still often use).
For collaboration reasons, I have to use Word for work and was quite pleased at many of the changed made in 2010 vs old versions. Yes, there's no groundbreaking features, but UI changes can a big deal and big time saver in programs also.
Re:And 3 hours after reading this... (Score:3, Informative)
That is just complete bullshit, Bulldozer will use AM3+ and probably be compatible with AM3 (just as the AM2+ CPUs were compatible with AM2 motherboards).
Re:Paying of OEMs is not their only trick.. (Score:3, Informative)
The fact that AMD's could run unoptimized code faster than the netburst optimized code is only testimonial to how crappy the netburst design was.
FACT: Intel's compiler is more than happy to selectively test for each of SSE, SSE2, and SSE3 at runtime and will use the best path it can, if and only if the vendor ID is "GenuineIntel." If it is not "GenuineIntel", then it runs a 80386 codepath intead. Thats right.. you heard me.. it runs a codepath for 25 year old CPU's.
Stop making excuses for the bullshit that is Intel's compiler shenanigans.
Re:Paying of OEMs is not their only trick.. (Score:4, Informative)
That's not the whole story - there was a deliberate attempt to not use optimized instructions.
See a long discussion at http://www.agner.org/optimize/blog/read.php?i=49 [agner.org]
Re:CPU manufacturers and I have a history (Score:3, Informative)
I said it before Intel did it, jackass. I'm not a fanboy, BTW. I'll buy Intel when it makes sense. I just prefer AMD after about two decades of experience buying products of both companies.
The 486 had a DRAM controller on its die? I'm going to have to ask for a citation. I think you're thinking of either the on-die L1 cache or the MMU (memory management unit), neither of which is a system main memory controller. Here's a citation to the counter: List of Intel Chipsets at Wikipedia [wikipedia.org]. See how the chipset determines the memory specs up until the Core i Series, including the 80486? Here's another: List of Intel Chipsets at World IQ [wordiq.com]. Here's another: Intel CPU and Chipset History at Overclock 3D courtesy of a forum post there by "PV5150" [overclock3d.net].
An MMU has nothing to do with controlling the actual SIPPs, SIMMs, or DIMMs. It's a multiprocessing ("multiprocessing" doesn't mean "multi-core") feature that allows the processor to enforce memory address range protection so that program A doesn't stomp on program B's memory range. That's a separate concern from getting data into and out of the processor from main memory.
Why don't you go get a copy of something like Upgrading and Repairing PCs [quepublishing.com] and inform yourself? Here's the ISBN.nu link for the 18th edition in case you're a bargain shopper: 18th edition [isbn.nu]. I have the fifth edition myself. I might get an updated version for the handy reference tables in the back featuring things like POST codes and error codes for SCSI controllers.
BTW, have you ever actually built a PC older than, say, a Pentium 4? Or owned one?