AMD Launches First 45nm Shanghai CPUs 264
arcticstoat writes "The wait for AMD's next-gen CPUs is finally over, as the company has now officially launched its first 45nm 'Shanghai' Opteron chips for servers and workstations. 'AMD's move to a 45nm process relies on immersion lithography, where a refractive fluid fills the gap between the lens and the wafer, which AMD says will result in 'dramatic performance and performance-per-watt gains.' It's also enabled AMD to increase the maximum clock speed of the Opterons from 2.3GHz with the Barcelona core to 2.7GHz with the Shanghai core. Shanghai chips also feature more cache than their predecessors, with 6MB of Level 3 cache bumping the total up to 8MB, and the chips share the same cache architecture as Barcelona CPUs, with a shared pool of Level 3 cache and an individual allocation of Level 2 cache for each core.'"
Re:Which to buy now? (Score:4, Informative)
This news is about a server/workstation chip, and I don't do any purchasing of those. As far as desktop chips are concerned, AMD was ALWAYS competitive on a price-performance basis. The key word there being price.
Re:Which to buy now? (Score:5, Informative)
According to Anandtech's review, it's highly competitive for database servers. http://it.anandtech.com/IT/showdoc.aspx?i=3456 [anandtech.com]
Re:great hallmark (Score:3, Informative)
Thriving on life. I've just lived there for four months and wish I could go back.
Re:Oh please. (Score:2, Informative)
Frequently?
Do you have any actual backup for that? The last three desktops I've owned have all had AMD processors, and the only thing that's gone bad on any of them was the AGP slot went bad on one of the mobos after about 5 years.
Re:Oh please. (Score:3, Informative)
No he doesn't have any data to back that up. Well, not unless he is going back to the K6-2 and K6-3 days. I have used and supported many AMD system and only had 3 of them with problems with chipsets. Actually, this problems weren't even the chipsets, they were with cheap board manufacturers who used counterfeit parts and got a load of bad resisters. The ECS K7S5A had a short run on that.
Of course he might be referring to the older VIA chips that allowed the user to select optional components during the driver installs for windows 98 and early 2000 versions of the drivers but the problems there was with the USER not picking the correct options. I don't consider that as a bug in the chip either, that was back when people were expected to know what they were doing when mucking around with system drivers.
If you want to talk about buggy, I just put a Quad core pentuim system together. The store didn't have the mainboard I wanted so I settled for a paired down version of it. It said it supported my chip right on the box. After 3 days of blue screens, swapping memory out, and other problems, I found some bad reviews for the board online that said I needed a certain Bios Update before it would work with the quad core intels. Sure enough, there was a bios update at intel's site (the board shipped with a bios about 5 revisions out). I downloaded it and wouldn't you know it, the bios itself wouldn't fit onto a floppy (1.5 meg). SO I burnt a CD but the floppy boot disk couldn't find the SATA cdrom. So I spent the next day or so attempting to make a bootable CD with this bios update on it. No go with the freeware utilities and I ended up having to format the hardrive in Fat32 and copying the bios file there just to flash the bios. The techs and the shop I got the board from asked why I didn't just use the windows version of the bios updat and couldn't understand that if windows wouldn't stay running long enough to download it or even open the CDROM to run it, it probably wouldn't be wise to attempt to flash the bios from the OS in this state if I somehow did manage to get it on the machine. Anyways, 5 days to get something running that should have taken 45 minutes to 2 hours altogether. And yes, this was my fault because I didn't scope the board out before I decided to use it and I didn't know how to make a bootable CD that I could add files to. It wasn't the chipset's fault but I'm sure someone could somehow blame it if they wanted to. That's what some of the reviews were saying. Human error is often redirected to things or objects which is probably what the gp was doing.
Re:Which to buy now? (Score:5, Informative)
No, they weren't. For the past year Intel has boxed AMD in with chips at the same performance and lower price, or the same price and higher performance, or both.
And Intel has had performance segments (QX*) stretching well above AMD's, and pricing segments (Atom) well below AMD's.
AMD's short-lived price/performance superiority in the desktop sweet-spot in 2004 and 2005 has left many people thinking they're still in that position. That hasn't been true since Core 2 came out. HyperTransport gave them a slight edge in very-high-end servers for certain applications, but Intel stayed near them with reliably higher clock speeds, and is coming out with QuickPath in four days, wiping out those few use cases where AMD can make easy sales today.
What I'm saying is, right now you are likely to choose Intel in almost all situations, if you are objective.
Re:Oh please. (Score:3, Informative)
I think you pretty much nailed it. A few years ago I bought a amd 3000+ for 595 bucks because I wanted bleeding edge. A few days later a friend bought a amd 2800+ for 250 bucks. Less than half what I paid.
My extra almost 400 bucks got me nothing that mattered. Sure I was faster then he was but he could play all the same games I could just as well. Sure the web, read email, view porn. Just as good as I could.
The only place you could tell the difference was when encoding video or audio. I was just a few seconds faster. Well worth the 400 bucks....not.
Re:Oh please. (Score:4, Informative)
What happens is that during normal operation of any RAM there is a small chance that a particular bit will get flipped. Cosmic rays are often blamed as the culprit; of course if you overclock and overvolt your memory you increase the chance of errors, but even good quality RAM running within spec will get an incorrect bit every so often. If you use non-ECC memory there is no chance to spot this error; it just returns the wrong data. The old parity memory added one extra check bit for every eight bits, so most of the time it could detect (but not correct) a one-bit error. ECC stands for error correcting code (look it up on Wikipedia) meaning that if one individual bit is corrupted it can recover the correct data. If you are really unlucky and two bits in the same code word get corrupted at the same time then you still have problems, but that is unlikely.
If you are using non-ECC RAM then may be getting corrupted from time to time, but you don't notice.
Re: ECC RAM (Score:3, Informative)
Accorcing to Wikipedia (http://en.wikipedia.org/wiki/Hamming_code#Hamming_codes_with_additional_parity_.28SECDED.29 [wikipedia.org]), a scheme that can correct 1 bit error and detect 2 is typically used. So it can correct single errors. The most common reason is some form of radiation (see http://en.wikipedia.org/wiki/Soft_error#Causes_of_soft_errors [wikipedia.org]).
Against at least one of those (cosmic rays), even quality RAM is not immune. This said, only the vendors of quality RAM seem to be in the business of making the ECC version anyway. Which makes it more expensive, because you're buying more of the good stuff (the parity bits for the SECDEC algorithm require an extra chip on the module). But considering the overall price of RAM, that should be not a problem.
Re:Which to buy now? (Score:5, Informative)
cycles aren't everything in all cases... AMD still has more system bandwidth, which speeds up everything when talking about IO bound applications. FSB speeds up every aspect of the computer.
The applications where AMD is superior are IO bound applications like database servers, and music production.
Intel is better for video because you are dealing with a limited number of streams and it's computationally expensive, so is CPU bound.
With audio you can have hundreds of streams (often 4-6 per fader on the mixer), and at 24/96, will quickly overwhelm any intel based system. Since a lot of us use DSP cards ( think of it as GPU for sound) the data path capacity, especially to the DSP processors (PCI/PCI-e) is very important, and Intel simply can't touch AMD in this respect.
AMD architecture simply has untouchable plumbing. If you will notice, Apple is looking for a new chip vendor. This probably has a lot to do with it since most audio professionals use Apple gear.
If Jobs and Co. were smart, they'd offer both intel and amd architectures, depending on the job being done. Intel is fantastic for video and a lot of pro video peeps use Apple gear too. Those are two market segments that couldn't be more different in their requirements. To be the best of the best for multimedia, Apple needs to either build a new architecture or offer both AMD and Intel.
-Viz
Re:Chipset?... (Score:3, Informative)
Despite any advances AMD makes in CPU's, they still have such a sub par selection of chipset vendors.
What's wrong with their PCI-E bridge chip? It converts PCI-E to HT and back pretty well afaik. Or maybe you meant the southbridge? Yeah, that USB and SATA logic is really cramping my gaming rig.
The performance-interesting parts of the northbridge are on the CPU in AMD architectures (and now intel ones too), and they're great. I'm not sure what you're complaining about.
Re:Oh please. (Score:3, Informative)
I have to agree with Lonewolf666. I have been agonizing over going with AMD over Intel but the ECC issue is a deal breaker. It is only supported on Intel's more expensive and older motherboards. DDR3 in combination with ECC is not supported at all ruling out anything recent and an FB-DIMM solution would be more expensive yet.
All of AMD's recent processors have really nice support (note 1) for non-registered ECC DDR2 with the caveat that not all systems have support in BIOS for it. Gigibyte motherboards supporting embedded graphics (780G and 790G) do not while similar Asus ones do.
My current system is an old P4 2.4C and 875P chipset with 2 GBytes of ECC DDR and my system before that is an HX chipset with 1 GByte of ECC SDRAM that now does home server duty.
As for how necessary ECC is . . . there are a couple of papers available online that discuss measured soft error rates versus density and memory size. Below 1 GByte it is not much of a concern but with large arrays it can become a limiting factor in data and system reliability. My fuzzy recollection is that errors could be daily for an 8 GByte memory array. For a majority of systems running games and media applications there is of course no issue.
Note 1: AMD Barcelonas support a variety of ECC modes including scrubbing and chipkill while the Intel systems I considered only support correction on read.