AMD has been riding a wave of success since the introduction of the Socket 939 platform and the resultant boom to both processor sales and performance. This has put Intel back on their heels and allowed the Athlon 64 line to assume a position of dominance, and put AMD at the forefront of processor development for the very first time. The Athlon 64 led the way to 64-bit computing, but Intel managed to beat AMD to the multi-core and 90nm punch. The subsequent release of the 90nm Winchester, San Diego and Venice cores have blunted that advantage, and brought AMD right back to the head of the table.
The new Venice core, or E3 revision, is a bit different compared to the previous 90nm Winchester release, and offers not only a higher clock speed ceiling, but also an updated feature set and higher performance. The integrated memory controller has been given an overhaul, and it now includes support for mismatched DDR DIMMs and the population of all four sockets at DDR400 speeds. This last factor is important, because in previous core revisions, the Athlon 64 and 64 FX processors would drop down to DDR333 memory speeds with all four DIMM sockets populated with double-sided DIMMs, Overall memory performance has been also enhanced through improved memory mapping and lower memory latencies.
The 90nm Venice core also has support for SSE3, an addition that brings it on par with the Pentium 4 Prescott models, and not that far behind the Intel cutting edge. AMD has planned their support of these Streaming SIMD Extensions, and as each new revision starts to gain speed in the developer community, AMD quickly adds support. SSE3 will provide some extra power for multimedia and media encoding duties, but one of the core benefits to SSE3 is upgraded multi-threaded performance, but this is unlikely to aid AMD's single-core/non-HT processors.
In terms of physical design, the Venice core does feature lower power requirements, and the Athlon 64 3800+ 90nm model runs at 1.4V (or lower), compared to 1.5-1.55V for the Newcastle version. This not only means a potentially cooler-running core, but also allows for higher clock speeds in the future. We could hypothetically see the 90nm Venice core at 2.6 GHz soon and potentially hitting 2.8 GHz or higher in the future, but this is doubtful since AMD is moving to a "dual-core Athlon 64 X2/single-core Athlon 64 FX" lineup. This is probably the final Athlon 64 90nm processor we'll see, so enjoy it while you can.
The other factors remain the same, and the Athlon 64 3800+ Venice still runs at 2.4 GHz, uses the Socket 939 platform, features a 128K L1 and 512K L2 cache design, and includes an integrated dual-channel DDR controller. Motherboards that adhere to the AMD specifications will need only a BIOS update to properly initialize these Venice-based processors, but in some cases, the actual motherboard is not physically compatible and a BIOS update may not be forthcoming. In our compatibility testing, we didn't find any serious problems with BIOS-certified boards, but certain features (Cool 'n Quiet) may not have been enabled, and some performance features had to be disabled or toned down because of the more stringent memory timings. As more time goes by, many of these niggling BIOS issues will be worked out.
When AMD introduced the 90nm Winchester core, it was only for models running at 1.8 to 2.2 GHz clock speeds, so core temperatures weren't really that big a deal, especially at the low-end of the scale. Now that we've hit 2.4 GHz, the benefits of a 90nm core start to appreciate, and become a base requirement for higher core speeds. But to ensure we know the real-world benefits of a smaller core, the actual CPU temperatures need to be gathered. Measuring core temperatures means keeping all components very consistent, as changing the motherboard, cooler, case or even PSU can yield different temperatures makes it impossible to run comparisons between two configurations.
For this test, we used the DFI nForce4 Ultra-D, because it sports a Venice-certified BIOS and produces consistent results for temperature evaluations. The heatsink-fan used was a ThermalTake Silent K8 cooper cooler with ShinEtsu compound. For this thermal test, we took the Athlon 64 3800+ 90nm Venice and 130nm Newcastle processors and measured them at idle, proceeded to run a series of CPU-reliant benchmarks and then took a load-type temperature reading. The readings are presented in the chart below.
At idle, the 90 nm Athlon 64 3800+ Venice shows a very nice advantage over the 130nm Newcastle, and checks in at more than 10 degrees Celsius lower. In terms of the load temperatures, we again see a significant advantage for the Venice core, but the 90nm part also doesn't give back any of the temperature gap and remains consistently lower than the 130nm Newcastle. This is a much better result than with the previous 90nm Winchester core at 2.2 GHz, and shows that not only does a higher core speed demand lower voltages, but also that AMD has made significant strides in their 90nm core design and development.
Unfortunately, we were not able to complete the overclocking side of the testing to our satisfaction, as there seemed to be some teething problems between our processor, the DFI nForce4 motherboard BIOS and potentially our reference memory as well. We could not get much out of the new Athlon 64 3800+ 90nm core without encountering stability issues, although we had standard overclocking performance out of other 90nm Winchester and 130nm Newcastle Athlon 64 models. We also tried the Athlon 64 3800+ Venice on a few nForce3 Ultra and K8T800 Pro motherboards (with certified BIOS) and it seemed to take to this a bit better and were able to achieve approximate 2.6-2.7 GHz clock speeds, but we're still not satisfied that this is the top-end of the clock speed range.
RSS feed
