X-bit labs

Monday, April 24, 2006

Hit-Parade of Major Reasons for Computer Failures. Invalid User Is Not the Primary One

When you surf around any hardware forums, you find a plenty of topics discussing technical issues. Very often the questions have to do with failure of the entire system or its individual components. Only experienced users can take into account all risks and make sure that their systems run smoothly for years. However, sometimes the systems die very unexpectedly.

Our colleagues from PC Stats web-site conducted a detailed survey among their readers about the main reasons that led them or their friends to system failures. As a result, they managed to put together the following “hit-parade” list that gives you a great idea of what to watch out for:

1. 26% - PSU and power issues
2. 23% - Bad gear and user negligence
3. 13% - Heatsink related
4. 15% - Assembly and moving
5. 10% - Lightning strike and static
6. 3% - Computer cruelty
7. 6% - USB related
8. 2% - Overclocking

• PC Stats: Most Common Ways to Kill a PC

Amazing Conroe Performance. First Benchmark Results at 3.1GHz Frequency

We have been waiting for the performance results for the overclocked Conroe processor for quite some time, however, bad luck eventually prevented us from seeing them. Although some hardware enthusiasts were quite lucky and managed to get their hands on an early Conroe sample, they still failed to really enjoy its potential, because of the lack of worthy mainboards. Even though they managed to get one Conroe processor to run in a G965 based mainboard, the BIOS was still too raw and the overclocking options were simply not there at all.

The theoretical frequency potential of Conroe processors has already been revealed: the rates should range from 1.6GHz to 3.33GHz. However, the reality was expected to be much more complicated. Such things as limited overclockability during bus frequency overclocking or a surprisingly pleasing result over 3.33GHz were just a few things awaiting the lucky pioneers. It was still unclear how the new architecture featuring shorter pipeline would behave during overclocking and how scalable the Conroe performance would be in relation to its growing frequency.

Well, today we can already start discussing the first real results that come from the Hong-Kong HKE PC web-site. Besides the already well-known Intel’s claims about 40% performance advantage of their 2.6GHz Conroe CPU over the Pentium D 950 (3.4GHz), our colleagues shared some valuable data about their successful overclocking experiment performed over the engineering sample of Core E6400 working at the nominal 2.16GHz.

So, with the default air cooler borrowed from the Prescott based system, the Conroe CPU reached 3.1GHz. This result was obtained by raising the system bus frequency from 266MHz to 389MHz. the maximum clock frequency multiplier supported by this processor model equals 8x. By the way, all Conroe models support EIST that is why the serial processors will also support multipliers below 8x.

According to some estimates, 400MHz bus frequency is going to be the peak bus frequency for Conroe processors. Note that far not every mainboard can ensure that the bus would overclock that far. I assume that very soon the mainboard makers’ claims about their products ability to support up to 400MHz bus frequency will no longer seem something unrealistic to us.

Another issue is low clock frequency multipliers of the slower Conroe processor models. They will be 7x and 8x respectively that is why when you speed the bus up to 400Mhz the maximum processor clock speed will only be 2.8-3.2GHz. Of course, this is a pretty good result for low-cost processor models, but some of the hardware enthusiasts out there will certainly be disappointed. Especially since not all the mainboards will be able to guarantee 400MHz system bus frequency. In particular, the budget i946PL chipset is only designed to support 200 (800) MHz bus. So, you can consider yourself lucky if the youngest Conroe processor runs at 2.4GHz in this platform.

Speaking of SuperPI 1M benchmark results obtained on Core E6400 processor overclocked to 3.1GHz, we invite all of you rise in applause.  The result of 16 sec is an absolute world record as of today! Even the dual-core Pentium XE 965 overclocked to 6.8GHz came only close to 19 sec with a liquid nitrogen cooling system. Conroe CPU manages to show a way better SuperPI score with regular air-cooling.

By the way, we were slightly puzzled by the fact that the Core E6400 engineering sample features 4MB L2 cache. The thing is that the mass production processors are expected to have only 2MB of L2 cache memory. Looks like only engineering samples can boast 4MB L2 cache, or maybe the company ha revised the specifications for some selected Conroe models.

In conclusion I would like to say that the overclocking potential of Conroe processors is very satisfying. At least, they ensure adequate performance improvement without any extreme cooling involved. Finally, very few overclockers hunt for relative performance boost rates, everyone else does it to get the practical performance improvement. Here, Conroe is the man!