Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
CPU Intel P4 840 D P4 820 D P4 630 P4 640 P4 650 P4 660 P4 670 AMD Athlon64 3500+ 3700+ 3800+ 4000+ X2-3800+ X2-4200+ X2-4400+ X2-4600+ X2-4800+ 1-Way Opteron Opteron 144 Opteron 146 Opteron 148 Opteron 150 Opteron 152 2-Way Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Opteron 248 Opteron 250 Opteron 252 2-Way Dual Core Opteron Opteron 270 Opteron 275 nVidia GF 7800GT GF 6800GT GF 6600GT ATI R X850 XT PE R X850 XT R X800 XT PE R X800 XT R X800 XL Memory Corsair Crucial Kingston Mushkin OCZ What are you shopping for?

Summary

High System memory density has recently fallen from grace, at least according to a report of severe performance penalties caused by filling up all four DIMM slots on the P4 - Intel i875 platform. Even at 2 GB, digital image processing was penalized with approximately 61 % longer processing time according to the timing tool built into Photoshop. Or so the story went.

There are performance issues associated with higher system memory configurations; clock skew caused by higher load on the command and address bus necessitates relaxation of some chipset timing parameters, and potentially the reinsertion of the additional pipeline stages that were disabled as a function of enabling PAT.

The penalties in those cases are, however, in the order of 3-6% and not 40-60% that were shown to incur after adding two extra DIMMs. There is something wrong in the state of Denmark or was that the state of Benchmark here? We have some good ideas where the bugs are and have the data to back it up.

A few days ago, a rather interesting article on "Building the Ultimate High-End Gaming Workstation" was posted on Firingsquad. In the benchmarking section of the article, the authors showed that increasing the amount of system memory beyond 1 GB caused a tremendous performance hit in all applications used for benchmarking except for the purely theoretical SiSoft Sandra Memory Bandwidth Benchmark that was run with Buffering and Block Prefetch enabled.

Performance Killers or Enhancers? 4 x 512 MB Mushkin PC3200 DDR with a total of 8 ranks and no less than 32 internal banks. Click for larger picture.

The system used for testing was an off the shelf MicronPC Client Pro 545 using the standard Intel D875PBZ "Bonanza" board. The most interesting finding was that doubling the amount of system memory from 1GB to 2GB resulted in up to 61% longer run-times per benchmark, equivalent to a 39% performance hit encountered after increasing the system memory density.

Benchmarks used were mostly related to digital photography and Mathlab applications and in conclusion, the article was concerned with the fact that despite the flood of benchmarks posted on the Athlon 64 launch or else in the numerous reviews of the Intel P4, nobody picked up the performance hit associated with filling more than 4 physical banks of memory.

There is a very good reason for that, though. We have been running the P4 systems with up to 4 GB of system memory in the past and we have not been able to see the effects that were described in the Firingsquad article.. While we have no doubts about the accuracy with respect to reporting the results, there are a number of issues that need to be looked into and brought into perspective in order to paint a "true picture of the situation".

next page:    => The Three Monkey (wrenches) =>