X-bit labs

Tuesday, July 12, 2005

AMD64 Vs. AMD64: Venice or San Diego? AMD Venice and San Diego Cores Performance Compared

Advanced Micro Devices has introduced several core revisions within its current microprocessor lineup, which are different in terms of several features, including SSE3 support, performance and power consumption. As a consequence of that it is not a trivial task to pick up the best possible AMD64 chip for particular needs.

Both Venice and San Diego cores are produced using 90nm process technology and deliver lower power consumption and heat dissipation as well as improved built-in memory controller and SSE3 instructions set support compared to previous-generation offerings. But there is also a noticeable difference. While Vince core boasts with 512KB of L2 cache, San Diego core provides 1MB L2 cache. Phoronix web-site has posted an article containing performance comparison of Venice and San Diego cores in real-life applications.

“Thanks to the 1MB cache in the San Diego, the extra cache proves to be beneficial in certain tasks, as shown in the Linux compiling process. The performance advantage is close to 10% in compiling LAME. In other benchmarks, such as Doom 3, the advantage of San Diego is roughly 5% at best. For Doom 3, we saw a 3-5% performance increase from the Venice to the San Diego, though as the quality gets higher, the performance increase diminished to less than 2%. However, this is partially due to the graphic card used to perform this benchmark,” the web-site reports.

“In conclusion, for average users, we found that the San Diego offers an average of 3% higher return than its counter part, Venice, when running at the same clock speed. With the pricing of the cheapest San Diego around $330, Venice 3200+ around $195, and Venice 3500+ around $275, it comes down to whether paying the premium price of the San Diego is justified. The AMD San Diego is one hell of a screamer for single-core setups, but with its high price tag, the small performance increase isn’t truly justified unless you’ll be coding and compiling all day or a similar CPU intensive process,” concludes Phoronix.

• Phoronix: AMD Venice vs. San Diego Core Performance Review.

Principles of FB-DIMM Memory Explained. The Future of Random Access Memory in Computes

PCStats.com has posted an article about Fully-Buffered Dual Inline Memory Module (FB-DIMM) technology. Basically, FB-DIMM memory combines DDR2 memory chips with a brand-new point-to-point serial memory interface which links each FB-DIMM module together in a chain.

“FB-DIMM memory uses a bi-directional serial memory bus which passes through each memory module, instead of having a single bus which each module dumps its data onto. Similar to PCI Express and other modern serial technologies, FB-DIMM transmits memory data in packets, precisely controlled by the AMB (Advanced Memory Buffer) chips built into each FB-DIMM module,” explains the web-site.

Since the FB-DIMM type of memory was designed for servers and workstations, those should be the first systems to benefit from the new memory technology.

“In the near future, expect FB-DIMM memory to appear in Intel-based high-end servers first, probably in the fourth quarter of 2005. Intel is already manufacturing AMB chips, while Micron, Kingston and other memory manufacturers like Infineon are churning out as yet unofficial FB-DIMM memory. The JEDEC standards commission is currently working on the official FB-DIMM specifications, but it’s unlikely that there will be any major changes. The technology's debut will probably be with Intel’s upcoming “Blackford” server chipset which also supports dual-core 64-bit capable Xeon processors,” concludes PCStats.com.

“While FB-DIMM technology looks to remain in the server world for the near future, it’s hard to imagine the most basic and attractive component of the technology - the high-speed serial interface - not becoming a standard soon. As DDR2 speeds ramp up and processor capabilities increase, it’s quite possible that the parallel memory bus will become a bottleneck for desktop system performance,” adds the author.

• PCStats.com: Introduction to FB-DIMM Memory Birth of Serial RAM.

AMD Athlon 64 Chips Support 500MHz DDR Memory, But Unofficially. Web-Site Investigates Memory Speed Impact on Athlon 64 Performance

Rumours about possible support of higher-speed, beyond officially certified levels, DDR memory by AMD Athlon 64 processors have been floating around for more than a year already, but without any proofs.

This time it seems that the latest revision E AMD Athlon 64 central processing units (CPUs) have actually gained ability to clock memory at up to 500MHz without exceeding its own stock clock-speed or HyperTransport bus speed. The capability is only available on certain mainboards though.

The question is, however, whether AMD Athlon 64 processors do need additional memory bandwidth at all.

“The performance improvements themselves aren’t tangible, but if you are trying to squeeze every last ounce of performance out of your system, then these new memory dividers offer you one more avenue to do so. If you have memory that can run at higher than DDR400 speeds without any reduction in latency, then by all means, explore the new dividers; just don’t expect them to change your life,” AnandTech web-site explains.

“Based on the tests that we’ve seen here today, AMD’s reluctance to move to higher bandwidth DDR2 offerings makes a lot more sense. The plain fact of the matter is that at the current clock speeds at which the Athlon 64 and X2 line are running, most desktop applications see virtually no benefit from higher bandwidth memory,” AnandTech web-site explains.

• AnandTech: Athlon 64 Revision E: Unofficial DDR500 Support.