Slashdot Mirror
1.3GHz Duron Arrives
zebadee writes: "Tom's Hardware has the news that AMD have released a 1.3GHz Duron to the "mainstream PC market" that has been optimised for use with WindowsXP. The article also asks 'why haven't AMD gone with the MHz doesn't equal performance as they have done with the new XP/MP chips, as it would be assumed the market for these will be consumers who don't generally look at benchmark figures?' More information can be found at the AMD website."
Does anyone have info on this?
"Flyin' in just a sweet place,
Never been known to fail..."
Why haven't AMD gone with the MHz doesn't equal performance as they have done with the new XP/MP chips, as it would be assumed the market for these will be consumers who don't generally look at benchmark figures?
Because the Duron competes against the P3-based Celeron, not the P4 that runs far slower clock-to-clock than its predecessor. If Intel hadn't deliberately designed the P4 for clock speed at the expense of performence, AMD would not have needed their True Performence Initiative.
In this case, two wrongs DO make a right. At least AMD's "wrong" is just marketing fluff rather than deliberately misdesigned engineering.
It might have something to do with the Durons 1.0Ghz and above having the newer core, while the old 1.3Ghz T-Birds used the older core. There are a lot of applications where the 1.3Ghz Duron will smoke a 1.3Ghz T-Bird.
Athlon's are being marketed against the PIV's however, and the PIV's have changed their architecture significantly. This has the effect of the PIV actually being slower at equivalent clock speeds. A PIII 2GHz would be faster, for most apps, than a PIV 2GHz. To counter this unfair MHz advantage, AMD came up with their PR numbers to show that Athlon XP's perfrom equivently to a higher rated PIV. Of course, once software is programmed to take advantage of the PIV's new architecture (rememebr when Pentium Pro's hit the scene?), I wonder if AMD will push those XP ratings down. =)
And as an enthusiast, I like knowing the actual MHz
Why?
If a 1.6 GHz (AMD) chip is faster than a 2.0 GHz (Intel) chip, then this seems to be a singularly useless number... if only had any meaning when the two companies had more similar architectures where the MHz figures were at least roughly comparable.
Note, incidently that the original speed measure was MIPS (Millions of Instructions Per Second), but this was not MIPS of the CPU/computer in question, but rather MIPS normalized to a VAX 780 having 1 MIPS.
How on earth we got to the point that people started to measure speed by MHz is beyond me. For the previous generation x86 CPUS it was admittedly semi-reasonable, but across architectures it was always useless.... Have you ever checked the clock speeds of the top SPECINT scores...
The difference bewteen athlons and durons is the cache (im not sure how much) and the Front Side Bus, the athlons have 133x2 and the durons have 100x2. this makes a BIG difference. A duron will run ok but a athlon will always be faster. Answer your question?
-
The Athlon, of course, is competing with the Pentioum 4 and is able to keep performance pace, but not clock pace. In that market clock speed causes confusion about actual performance, so the performance rating makes sense there.
That's my best guess as to why they don't use the performance rating system on Durons.
It could also be that AMD has no problem saying that the Duron achieves performance and clock parity with the Pentium 4. I haven't seen any benchmark comparisons between the Pentium 4 and the Duron, nor have I looked for them, but I have no problem believing that a 1.3GHz Duron qualifies for a 1300+ rating, or even a 1500+ rating. Giving it that rating, however, would place it in direct competition with hte Athlon. That maybe seems a little underhanded, but ask yourself what's more underhanded: limiting competition between your high- and low-end products through naming convention, or limiting it by intentionally crippling the low-end product?
Under capitalism man exploits man. Under communism it's the other way around.
The 1.1 and higher Durons are the exact same architecture as the Althon XP/MP, with less cache.
Wrong, on the Tyan Tiger MP the Duron, Tbird and XPs can be run in dual configuration, as reported by several hardware review sites, and the members of various forums such as the ones at amdmb.com. However, the newer mp athlon boards(ones using the MPX chipset) are locking down on the ability to use processors other than the Athlon MPs as required by AMD's contracts with the mobo manurfacturers.
I've got a Best Buy ad here for three Compaq computers. In the "Processor" line of the table it says: "Intel Celeron 1.3GHz", "Athlon XP 1700+ QuantiSpeed Architecture operates at 1.47GHz", and "Pentium 4 1.8GHz".
It clearly and openly states, in print as big as the Celeron GHz number, the Athlon QuantiSpeed number, and the Pentium 4 GHz, the actual GHz rating of the Athlon chip.
How is that not telling as it is?
I have a dual Athlon (Tyan's mobo) with two Athlon XP 1600+ CPUs.
They work quite fine: my current uptime is 47 days.
There are two differences between XPs and MPs: the price and the guarantee that MP will do SMP. As far as the latter one goes, I've never heard of an XP that won't run in an SMP configuration
Especially now, since the price of the XP processors are ever increasing. For less than $150 I can get a Duron 1200 and a very capable companion motherboard, the ECS K7S5A. For $65 after shipping from Newegg.com (No, I have no vested interests in Newegg, I just like their customer support and shipping), you get a solid performing, stable, compatable motherboard with built in Sound and Lan. Nothing to scoff at. It even has overclockability using such programs as CPUFSB. The Durons are a great family, I think I may go ahead and get one now :)
Tom's Hardware has a nice review and comparison of the athlon 1300 vs. the Celeron 1300.
These Durons are based on the Morgan core, which for all intents and purposes is the same as the Athlon XP processors, but /w less cache. This means you get SSE & instruction pre-fetch, like the XP, as well as lower power dissipation.
The T-Birds are based on an older core that didn't have those enhancements, but did have more L2.
So basically, given an equal clock speed, the rankings of the processors would be:
Old Duron -> Athlon T-Bird -> New Duron (this article) -> Athlon XP
The caveat is that in certain apps where L2 is the deciding factor, the T-Bird might be faster, but as a general rule, the core enhancements of the newer Durons (and XPs) outweigh the larger L2.
"Optimized for Windows XP" has little to do with the performance and/or features of the chip. It has more to do with the "standard" PC that Microsoft and Intel require you to have to get the "Designed for..." logos on your computer system.
The PC Design Checklist has a set of requirements that your hardware must meet in order to be certified for use with Windows XP. Basically, if you're a system manufacturer and want to pre-install Windows, you must follow these guidelines and use only parts that also meet the guidelines. AMD certified their products with Microsoft so that OEMs would be able to include them in PCs that have Windows pre-installed.
Microsoft and Intel come out with new requirements every year for PCs to get the latest "Designed for..." label and to preinstall Windows. Some of the latest requirements are that PCs are not allowed to have ISA devices and that the PC must be at the desktop/login stage no more than 35 seconds after the power button is pressed. (See the link I posted above for full details.)
It's not a conspiracy by Microsoft and Intel, or anything of that sort. It's the same thing that a lot of manufacturers go through to say that their products are "certified" for such-and-such uses. Note that you, as a PC manufacturer, are free to not certify your computer with Microsoft, but you lose the OEM discounts on Windows preinstalls if you do, and you lose the free advertising provided by Microsoft.
This, overall, is a Good Thing -- otherwise, cheap manufacturers might still be trying to force-feed us ISA devices and no USB ports.
5% fluctuation sounds more like the effects of different chipsets and other mainboard internals. Linux speeds up dramatically when I recompile for Athlon on my T-Bird 1100, much more than just 5%. WinNT and friends only have 2 kernels on board as far as I know (and I went looking for them). One is the plain vanilla single processor affair, the other has SMP. Both of these must be careful to cater for the lowest common denominator: a P1 Classic. Including instructions from anything higher than a P1 would cause problems on these CPU's. Not many people will be running XP on a P1, but still... Of course there's some dynamic stuff in there based on availability of MMX/SSE etc. I just don't think that's going to make the same difference as -march=athlon does on my Linux box. Maybe a few benchmark runs would be in order. I don't have the time or the box for such a test right now, maybe sometime next month, but this thread did start me thinking..
Learn from the mistakes of others. There isn't enough time to make them all yourself.
It may not be where you heard the story, but it is on page 177 of "The Practice of Programming" by Kernighan and Pike.
Because it's deceptive marketing. I'm an AMD user myself (TBird 1.2 GHz), but it really annoys me that they would be willing to mark a processor as something that it is not . Please don't try to tell me that 'performance rating...blah blah...equivalent speed in old architecture.' They are marking the new processors as speeds that they do not run at, period
That is horse shit. For the Athlon XP 1900+, where do they ever say it runs at 1900Mhz? No where. All they are doing is letting people understand that their architecture allows their XP 1900 to run just as well as the Intel chips running at 1.9Ghz. And if you look at the benchmarks that many sites are coming out with, you'll see the 1900 actually beats the 2+ Ghz chips from Intel.
If you want to talk about 'deceptive' practices, how about Intel designing the P4 so that the clock speeds are MUCH higher than their P3 chips, but the actual performance is not improved by that much?
AMD is just trying to make it easier for shoppers to compare. As long as they don't start naming their products "XP 2500" when they don't compete with the 2.5Ghz Intel chips (for example), then I think their naming system is good.
"And like that