Print 57 comment(s) - last by JonB.. on Nov 23 at 12:39 AM

AMD guidance details the backwards and forwards compatibility of the AM2, AM2+ and AM3 sockets.  (Source: AMD)

AMD details its platform progression identically to Intel's "tick-tock" architecture roadmap.  (Source: AMD)
AMD announces its Spider Platform in what could be its biggest launch of all time

Tomorrow marks the launch of AMD's platform launch, codenamed Spider. The launch consists of three components: AMD Phenom processors, the ATI Radeon HD 3800 graphics processor and AMD 7-Series desktop chipsets.

The first new AMD desktop architecture in four years will debut with the Phenom 9500 and Phenom 9600. Both chips feature a 95-Watt thermal envelope and a shared 2MB L3 cache.

The AMD Phenom 9600 ships with a 2.3GHz operating frequency, while the Phenom 9500 features a slightly lower 2.2GHz clock. Both processors run on HyperTransport 3.0 and feature a total 2MB L2 cache; 512KB per core. The chips also come with an integrated DDR2 memory controller and support speeds up to the DDR2-1066 specification, which is still pending JEDEC approval. carries the Phenom 9600 at a retail price of $322.00 and the Phenom 9500 at $286.00.

AMD guidance originally stated that a 2.4GHz Phenom, dubbed the Phenom 9700, would also launch on November 19, however last minute roadmap updates indicate this chip will come in December instead.

These two Phenom processors are part of AMD's first-generation, 65nm Stars family. In the second-half of 2008 AMD will announce its second-generation Stars family, which will be a migration of the current K10 core to the 45nm node.

The second-gen Stars lineup consists of five variants, including quad-core Deneb, dual-core Propus, dual-core Regor and single-core Sargas. The second-generation Stars CPUs will also include an integrated DDR3 memory controller, a first for AMD.

AMD's Spider Platform launch also includes the official announcement of its ATI Radeon HD 3800 Series. The launch of the Spider platform will add two new video cards to AMD's lineup, the ATI Radeon HD 3870 and Radeon HD 3850. 

AMD's new Radeon series received overwhelmingly positive feedback during AMD's November 15 media event.  Newegg representatives tell DailyTech the company sold out all of its stock on the first day, but is quickly replenishing inventory.

To round off the launch of the platform, AMD also announced its 7-Series desktop chipsets. The AMD 790FX targets the ultra-enthusiast market segment, and corporate guidance sets the retail price of 790FX boards between $150-250. Several partners already announced boards supporting this new chipset earlier this year.

AMD designates the AMD 790X as its performance chipset, slotted just below the 790FX.  AMD guidance states this chipset will run the consumer between $99-150.  The mainstream AMD 770 chipset will not see store shelves for several months, but AMD claims this chip will round off the mainstream segment for AMD chipsets. 

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

RE: AMD continues to disappoint
By excrucio on 11/19/2007 6:06:18 AM , Rating: 1
and the spider falls.

I was hoping for a huge success over intel and again im let down by it.

What ever happend to the huge floating point benchmarks that OPTERON proved over xeons?

Couldn't the x4 be the same? =[

RE: AMD continues to disappoint
By JumpingJack on 11/19/2007 6:10:07 AM , Rating: 2
The problem is that those huge floating point benchmarks were 'rates', if you take just the base, non-rate SPEC2006fp, C2 wins those FP benches.

The FP rate benches demonstrate AMD's superior bandwidth, this is effective whent he working sets and data sets are extremely large, and you must move large data sets to and from memory.

In client applications, there are very few that will stress the BW needs of the bus to enable the same result you saw in the FPU rate benchmarks.

The trend was the same with dual core, the dual K8 opty was actually pretty close to woodcrest in the rate benchmarks, but of course C2D was dominating K8 in the client desktop... samething here... effectively, simply being 'true-native' quad core provides no distinct advantage.

RE: AMD continues to disappoint
By Master Kenobi on 11/19/2007 8:16:04 AM , Rating: 2
simply being 'true-native' quad core provides no distinct advantage.

It never did. "Native" anything has been subjective at best. In theory it should eliminate cross talk bottle necks between the cores because they will use the interconnect rather than HT or Bus. While it will do this, the real world gain is so far not apparent. With Intel switching to QuikPath and Native Quads with Nehalem in the July/August '08 time frame this will be a moot point shortly anyways. By the way, Hyper-Threaded native quad cores for Nehalem.... 8 execution paths? Intel is a beast :)

I have a theory that sandwitching 2 together might cause problems for the memory controller (obviously you would get 2 and one has to be disabled). But I'm starting to wonder if they use native designs simply because it would cause some cross talk depending on which side the working memory controller was on.. Intel's MC is still on the north bridge so it could sling data to the processor without much trouble, just a logic update to which side to send it to. Maybe I'm just talking out my ass here but it was just a random thought that jumped into my head while reviewing the native/non-native quad designs.

RE: AMD continues to disappoint
By JumpingJack on 11/19/2007 3:24:06 PM , Rating: 2
Well done, well explained -- yes... AMD's share L3 cache eliminates the cache coherency problem at the L3 level, but they are discrete at L1/L2, the native design makes the coherency snooping much faster. The MCM that snoop goes across the FSB, no doubt ... however, the penalty is obviously small and in raw performance, well, we see the results.

I can dig up the link, but I have seen a few reference that the normal coherency pattern is that most snoops (80 or 90%) result in a NAK, which means the MCM is not paying the penalty 90% of the time...

RE: AMD continues to disappoint
By MandrakeQ on 11/19/2007 8:34:19 AM , Rating: 2
AMD has been targeting all of their designs for the server space since the K8, and these benchmarks show those decisions do not pay off in the desktop realm.

Core's larger caches are great for typical desktop usage where applications use small working sets and there is a small amount of processors to minimize coherency traffic. Luckily for AMD, these advantages degrade for servers when the addressable memory spans 128 GB and the processor bandwidth is chewed up by coherency requests of 16-32 cores.

If AMD could meet demand in the server space, they would be in a lot better shape since those processors carry higher margins. With that in mind, this desktop launch can't be all bad since it gives them a way to discard all of their broken parts.

I'm hoping they can find a way to increase their yields on the Barcelona parts and somehow make it out of this ordeal.

RE: AMD continues to disappoint
By JumpingJack on 11/19/2007 4:28:18 PM , Rating: 2
This is a good point, for HPC and scientific, FP related activity (aspecially in high cluster super computers), this processor will work great -- AMD focused alot of transistor budget on increasing the bandwidth, where this really helps in these situations.

Unfortunately, in many commercial space applications it still lags... Kanter was quick to point out the lack of any real meaningful commercial benchmark runs on Barcelona launch:

Makes sense, don't show something you cannot win :) ... but there will be a niche market for this CPU where the margins are good.

"DailyTech is the best kept secret on the Internet." -- Larry Barber
Related Articles

Most Popular ArticlesAre you ready for this ? HyperDrive Aircraft
September 24, 2016, 9:29 AM
Leaked – Samsung S8 is a Dream and a Dream 2
September 25, 2016, 8:00 AM
Inspiron Laptops & 2-in-1 PCs
September 25, 2016, 9:00 AM
Snapchat’s New Sunglasses are a Spectacle – No Pun Intended
September 24, 2016, 9:02 AM
Walmart may get "Robot Shopping Carts?"
September 17, 2016, 6:01 AM

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki