All chips, no fish -- it's Wednesday's IT Blogwatch: in which AMD launches three new processors. Not to mention clowning about with custard pies...
Advanced Micro Devices Inc. (AMD) has pumped up the performance of its Athlon 64 X2 dual-core processor and cooled down two single-core Athlon 64 processors. The top-of-the-range AMD Athlon 64 X2 dual-core processor is now the 6000+, with a clock speed of 3 GHz and 2MB L2 cache ... There's no speed boost for the single-core Athlon 64 processor, but AMD is now manufacturing the 3500+ and 3800+ models using a 65-nanometer production process ... The latest versions consume 45 watts, compared to 62 watts for the previous [90nm] generation ... Two weeks ago, AMD introduced new dual-core Opteron processors for servers that are either faster or consume less power than previous chips.
...
While AMD had begun to pull ahead in the market for low-power server and desktop chips, Intel Corp. is fighting back. Last month, Intel persuaded Sun Microsystems Inc. to use its Xeon processors in future servers and workstations based on the x86 architecture. Sun had used only AMD chips in those machines in recent years, and will now use chips from both manufacturers. At the top of the scale, Intel took the lead in the market for quad-core server chips last November, with the launch of its Xeon 5300 series, previously known as Clovertown. AMD won't introduce its range of energy-efficient quad-core server processors, code-named Barcelona, until the middle of this year.
The Athlon 64 X2 6000+ has been available from some online retailers since January, but AMD has now given the chip a formal launch ... [it's] AMD's fastest desktop offering for Socket AM2 yet. As evidenced by AMD's updated price list, the Athlon 64 X2 6000+ has an official price tag of $464. That's a good $138 more than the next chip down the line, the Athlon 64 X2 5600+ ... The 45W Athlon 64 3500+ is launching at $88, while the 3800+ is $93.
But Justin Mann is disappointed:
[The 3500+/3800+'s] TDP of only 45W [is] still higher than, say, a mobile processor, but significantly less than their 90nm cousins require.
...
Unfortunately, [the 6000+] still relies on the same X2 core we've come to love, meaning that it will not be able to stomp Core 2. The pricing on the low-power units is quite nice ... Definitely a great match for building a low-cost machine you could easily passively cool or use extremely quiet cooling with, yet still manage to have the performance of an A64.
Both single-core processors are based on 65nm technology (like Intel's Core 2 Duos). The difference is that AMD's chips draw only 45 watts of power... whereas Intel's draw around 65. So if you're trying to cut back on costs or need a quiet living room PC, AMD's new chips (although late to the game) are the better bet.
Adrian Kingsley-Hughes doesn't understand:
Why [is] AMD still releasing updated Athlon 64 X2's that aren't faster than Intel's processors? Seems to me that it could be spending R&D money better on establishing a line that seriously beat Intel either on price or performance (or both). A shift to a 65nm process would help - the 90nm process has been squeezed dry.
Anand Lal Shimpi tries to explain:
AMD has been taking a beating lately; when Intel released its Core 2 microprocessor family it was priced and performed aggressively, too aggressively for AMD to adequately compete with at the time. AMD's marketing even went into remission as reviewers weren't even seeded for speed bumps to the Athlon 64 X2 line, the X2 5000+ and FX-62 were the fastest Socket-AM2 parts AMD sent out for review and they were the first ones we'd ever reviewed as well. Sometimes no PR is good PR, and an article around the launch of the Athlon 64 X2 5200+, 5400+ or 5600+ simply proclaiming that Core 2 is once again on top would only work in Intel's favor.
...
the Athlon 64 X2 6000+ isn't even the most attractive part of AMD's lineup. Shave 200MHz off the 6000+ and you get the Athlon 64 X2 5600+, priced at $326. Note that the X2 5600+ is simply a FX-62 in disguise, what used to be a $999 processor has been reduced to less than a third of its cost - ain't competition grand? Of course a price war isn't in AMD's best interest when it comes to making money, but it's the best AMD can do until its new micro-architecture makes its debut later this year.
MD has been skimping lately on its cache. I have a sneaking suspicion that the majority of AMD's current performance issues are related to cache and lack thereof. The Intel chips carry 4 to 8 Mb of cache. The thing about the Intel architecture is that the cache is shared across both or all 4 cores. In contrast the AMD chips have a dedicated tiny 1 MB cache for the consumer chips and 2mb per core on the high-end parts. With that said, the reality of dual core computing is that one core is used much more heavily than the other. In Intel's case this means that one core is basically given the entire cache for its use - a significant performance boost when running a few tasks. In AMD's case the idle cache is inaccessible to the heavily loaded core.
The reason that makes me think that the cache is the current bottleneck is that the memory controller on the AMD chip is significantly faster than Intel's. Given that fact one would conclude that in non disk-bound applications that require large amounts of memory (games) the AMD chips would pull ahead. This is not the case. Of course there is more than just cache at play here but the fact that the Intel chips has 4 to 8 times more cache available to it has to make a fairly significant difference.
AMD chips don't need a lot cache as they have a build in memory controller and a better chipset to cpu, cpu to cpu link. In a 2-4 cpu server the direct cpu to cpu links with out havening to use the chip set also reduce the need. I think that Intel may have to add cache to the ram controller / main chip set soon in their 4 cpu severs.
Here's Chris Burke, with a perspective:
Intel has finally decided to unleash the flood gates on their own manufacturing and produce huge caches. Before the most recent generation of chips, Intel's desktop parts weren't sporting very big caches either ... Anyway, Intel has the best fab tech in the industry, some of the best circuit designers, and the most fab capacity. Combine this, and it is economical for Intel to put big caches on all their parts, and they decided to start using that advantage. AMD can't afford to follow suit -- not only are their caches larger in die area for the same storage, they also don't have the capacity to produce huge chips. AMD is already fab limited.
This is why the recent IBM announcement about eDRAM is significant. AMD has a tech sharing agreement with IBM. If eDRAM is practical in AMD's 45nm process, then that could eliminate Intel's advantage in cache sizes.
Buffer overflow:
Around the NetAround Computerworld Previously in IT Blogwatch
- Richi Jennings: Why You Shouldn't Buy from Spammers
- John Murrell: Viacom and Joost: Less like Web 2.0, more like TV 1.0
- Linux-Watch: Mono brings Visual Basic programs to Linux
- Jonathan Bruce: Relational Database Redux -> Yahoo! Pipes
- Marshall Kirkpatrick: Viacom to Sign Deal with Joost
- James Governor: Woohoo: Women doing the startup thing (Java too)
- Dave Risney: IPv6 URIs in IE7
- Coding Horror: Everybody Loves BitTorrent
- Dave Caolo: Watch Apple Store Boston's construction live
- Rmorril: Month of PHP Bugs
- Opposable Thumbs: Shopping for HDMI cables: who do retailers think they're kidding?
And finally... How To Throw Custard Pies. Bonus link: Tony spams for Britain
Richi Jennings is an independent technology and marketing consultant, specializing in email, blogging, Linux, and computer security. A 20 year, cross-functional IT veteran, he is also an analyst at Ferris Research. Contact Richi at blogwatch@richi.co.uk.