Intel's Nocona Xeon / Supermicro SuperServer 7044A-82R
Posted on: 08/21/2004 05:00 AM

Since I've taken on the task of combining a CPU review with a Server/Workstation review, I'd best chop my conclusion into two separate and distinct sections. First, I'll give you my concluding thoughts on Intel's latest processor, Nocona.

Nocona has brought certain technological advancements to the Xeon line that were certainly necessary. Xeon's 533MHz FSB had been showing its age, especially when you factor in the resource sharing requirements of two logical processors on one physical processor core. As such, the FSB boost to 800MHz is quite welcome and as you can tell from our benchmarks, it has a direct relation to performance. Especially when dealing with bandwidth intensive applications like media encoding.

That said, I don't believe Nocona out-paces its predecessor by as large a margin as I may have expected, especially considering its 400MHz advantage in clockspeed. Of course, this isn't exactly uncommon with the most recent releases of Intel processors. When Intel released Prescott and it was compared against Northwood, the results were not overly positive. However, the changes Intel made (deeper pipeline, for example) were made with the intention of ramping clockspeeds beyond what Northwood was capable of. This may also be the case here. In theory, Nocona may be able to scale beyond the 4GHz barrier and at such a clockspeed the performance delta's would doubtless increase to a more acceptable range.

Let's quickly shift our spotlight to the L2 cache sported on Nocona. With the last of the 533MHz FSB Xeons, Intel tried to compensate for its slower bus speed by adding additional L3 cache to the die of the processor to reduce the amount of fetches the CPU would have to make to main memory. Since Nocona finally brought the faster bus speed to the table, the additional cache was deemed unnecessary and Intel decided to go with 1MB of L2 cache instead of 512KB of L2 and 1MB or 2MB of L3 cache. It's easy to question this decision, but as of this moment, I'm not 100% sure that I disagree with it. Cost becomes a major factor when you're dealing with this much on-die cache, which is why processors like the Pentium 4 Extreme Edition are astromicially priced. To ensure this processor appealed to the Xeon-loving masses, I'm sure keeping the cost within a semi-acceptable range was a factor in this decision. Will Intel release another version of this processor with even more L cache? I hope so, because right now, if I was building a mid-range database server based on Intel processors, I may opt to go in another direction.

PCI Express and EM64T? PCI Express is definitely a step in the right direction for dealing with the increased bandwidth demands of current and future expansion cards and peripherals. As I said previously, my time spent with PCI Express to this point has been nothing but positive. EM64T on the other hand is still somewhat of a gray area. After talking about the technology with a couple of "the guys" on IRC, it became clear that Intel processors have been able to do 64-bit integer math for quite some time. Furthermore, the memory support issue is somewhat ambiguous. While the jump to 64-bit would seemingly take the amount of physical memory that Xeon can support through the roof (exabytes?), it almost seems as if Nocona can't address any additional memory than Xeon was already capable of, via Physical Address Extension (PAE). For the curious, that number has been 64GB for quite some time. Is EM64T more marketing than substance? I'll reserve judgement until I see the benchmarks with my own eyes. Of course, when I know, you'll know.

Moving on to Supermicro's SuperServer 7044A-82R, I must say that my experiences with this machine have been flawless to this point. The hardware included is of a high quality, which we've come to expect from Supermicro's products. While the name indicates that this machine is targeted towards the server market, it's clear that with the addition of a PCI Express video card and the BIOS fan adjustments switched to "SuperWorkstation", the 7044A-82R will be able to handle double-duty as a workstation without so much as blinking an eye. If you're in need of a high-end workstation or mid-range server, I can easily recommend this machine. Whether or not Nocona is the right processor for your needs will probably depend on the application you intend on running. If you're looking to get involved in high-end encoding or rendering, I think it will give you a definite performance advantage over its predecessor.

I'd like to express my thanks to Supermicro for providing the server for evaluation and affording me enough time to test it the way we're accustomed to. I've had the unit in my possession for several months and have been using it for a variety of tasks during that time. The only disappointment is that I have to send it back.

Printed from (,11.html)