915 Motherboard Roundup: Socket 775 for the Rest of Us
by Wesley Fink on December 7, 2004 12:25 AM EST- Posted in
- Motherboards
AOpen i915Pa-EFRII: Overclocking and Stress Testing
FSB Overclocking Results
| Front Side Bus Overclocking Testbed | |
| Processor: | Pentium 4 Prescott LGA 775 560 ES (2.8GHz-3.6GHz) |
| CPU Voltage: | 1.425V (1.3675V default) |
| Cooling: | Thermaltake Jungle 502 |
| Power Supply: | OCZ Power Stream 520 |
| Maximum CPU OverClock: | 217x18 (3996MHz) +9% |
| Maximum FSB OC: | 217FSB x 14 (+9%) |
When the 915/925X boards first appeared, there was a lot of talk about a 10% overclock built into the 915/925X chipset. Manufacturers of boards that cater to enthusiasts, like Abit, Asus and DFI, quickly found ways to get around this design issue, but others did not seem to see this as a big problem. AOpen falls into the latter category here. Whether at stock multiplier or the Speedstep reduced 14X, the AOpen will reach 217 and nothing more.
Most users never overclock, and the AOpen is a great board for them, since it contains all the important performance features that are missing on many other 915 boards in the roundup. However, overclockers should look elsewhere for a 915 board as they will be very frustrated to be limited to a 9% overclock.
Memory Stress Test Results:
The memory stress test measures the ability of the AOpen i915Pa to operate at its officially supported memory frequency (533MHz DDR2), at the best performing memory timings that Crucial/Micron PC2-4300U will support. Memory stress testing was conducted by running DDR2 at 533MHz (stock 3:4 ratio) with 2 DIMM slots operating in Dual-Channel mode.| Stable DDR533 Timings - 2 DIMMs (2/4 DIMMs - 1 Dual-Channel Bank) |
|
| Clock Speed: | 266MHz |
| Timing Mode: | 3:4 (200:266 - Default) |
| CAS Latency: | 3.0 |
| Bank Interleave: | Auto |
| RAS to CAS Delay: | 3 |
| RAS Precharge: | 3 |
| Cycle Time (tRAS): | 10* |
| Command Rate: | N/A |
The AOpen i915Pa was completely stable with 2 DIMMs in Dual-Channel at the DDR2 settings of 3-3-3-10 at 1.8V default voltage. These are much faster settings than the DDR2 SPD and rated timings of 4-4-4-12.
Filling all four available memory slots is more strenuous on the memory subsystem than testing 2 DDR2 modules on a motherboard.
| Stable DDR533 Timings - 4 DIMMs (4/4 DIMMs - 2 Dual-Channel Banks) |
|
| Clock Speed: | 266MHz |
| Timing Mode: | 3:4 (200:266 - Default) |
| CAS Latency: | 4.0 |
| Bank Interleave: | Auto |
| RAS to CAS Delay: | 3 |
| RAS Precharge: | 3 |
| Cycle Time (tRAS): | 10 |
| Command Rate: | N/A |
DDR has a long benchmarking history that makes it easier to figure out where to go with memory timings. We are still in the learning curve with DDR2, and the old DDR patterns just do not work in DDR2. We found that no matter what the rest of the settings, 4 DIMMs with DDR2 generally required CAS 4 for stability. Timings like 3-4-3 or 3-4-4 appeared to work fine, and then we were suddenly facing hard drive corruption. Our advice for now is to use CAS 4 with 4 DIMMs for stability, although 4-3-3 timings did work fine on most of the test boards. If you plan on trying to raise voltage to attempt lower CAS with 4 DIMMs, then please backup your system first.
Facebook
Twitter
RSS
26 Comments
View All Comments
Live - Tuesday, December 7, 2004 - link
Sorry Didn't see your reply before I posted Wesley.Sure there is some value to be had but not "outstanding". I still don't agree with you but I guess my mind is made up. Intel needs to come out with something new before I go back.
As a roundup it was very good reading tough. I can't wait for the next AMD roundup to hit AnandTech.
Live - Tuesday, December 7, 2004 - link
#12"The P5GD2 is an expensive motherboard, at about $240 on the web, but you can get almost all the same features in the P5GD2 Deluxe for about $50 less."
Thats expensive to me. Compare that to the 134.99 for the 939 Gold Editors Choice winner "MSI K8N Neo2 Platinum"
But thats not the point. If the 915P was substantially cheaper then a 939 system you might call it value for money but is it not. Mind you a 939 board is generally not cheap either but at least it delivers in comparison.
The CPU used in the review that hardly beat the much cheaper 3500+ had a max overclock of 14% and I bet you would find it hard to reach that high without the CPU overheating and start throttling http://www.anandtech.com/news/shownews.aspx?i=2345...
LGA775 CPUs does not offer great overclocking headroom compared to the much cheaper earlier Intel platforms or AMD for that matter. Sure they still overclock but nothing that we haven't seen before at higher cost and temperature. Again not what I would call outstanding.
Wesley Fink - Tuesday, December 7, 2004 - link
#11 - We just ran the 3500+ benchmarks in the same configuration this morning, and we do agree that the 3500+ is a particularly good value in performance for the dollar. However, the larger picture of prices of AMD CPUs compared to Intel show the Intel processors are a good, if not outstanding, value.Our conclusion was based on Anand's value analysis in the 3.46EE/1066 launch review at http://www.anandtech.com/cpuchipsets/showdoc.aspx?... There he compared the 3800+ at over $600 to a Intel 560 3.6GHz at about $450 and found the 3800+ the winner but probably not a big enough winner to justify the price premium for the 3800+. At that time, there was no 570 (3.8GHz) and the 3.6 was the fastest Intel CPU unless you considered the $1000+ Xeon-based EE processors. Price changes continue, and with them the value relationships do change.
A quick check of prices today shows
Intel 520(2.8GHz)- $160 AMD 2800+(754) - $128
Intel 530(3.0GHz)- $180 AMD A64 3000+ - $152
Intel 540(3.2GHz)- $220 AMD A64 3200+ - $194
Intel 550(3.4GHz)- $282 AMD A64 3400+ - $269
Intel 560(3.6GHz)- $455 AMD A64 3500+ - $270
Intel 570(3.8GHz)- $795 AMD A64 3800+ - $630
AMD A64 4000+ - $716
AMD A64 FX55 - $812
With current prices we would have to agree that there is really no great value advantage to Intel any more. But there is good value in the Intel processors from 2.8GHz (520) to 3.6GHz (560). Certainly the 4000+, at $80 less than the 570 and faster performance, and the FX55 at about the same price as 570 and significantly faster in most benchmarks, are better value at the top. But we still stand by Intel being good value in the middle.
deathwalker - Tuesday, December 7, 2004 - link
"outstanding value and performance for your buying dollar" ?????????? at $240 for a Mobo?..I guess I need to retake Economics 101...Bah...Intel just continues to shot themselves in the foot. A side note not related to this review..Dell must be deep inside of Intel's pocketbook with there contiued refusal to market AMD based products.danidentity - Tuesday, December 7, 2004 - link
#11 - 915P motherboards are not expensive. They are equal or cheaper in price than socket 939 A64 motherboards.LGA775 CPUs offer great overclocking headroom if paired with the right board. Intel CPUs have traditionally have had more OCing headroom than AMD chips. That still holds true, for the most part, today. Especially when talking about the low-end chips, like the 2.8GHz.
Live - Tuesday, December 7, 2004 - link
I'm sorry but I don't see the "outstanding value and performance for your buying dollar"Expensive Motherboards and CPUs with little overclocking headroom compared to the Athlon 64 competition. How does that translate to excellent value and performance? even the much cheaper 3500 comes out on top on most benchmarks.
Sure there are niche markets where the Intel platform excels but for the big majority of us AMD is where its at right now.
I don't think this review is in sync with your conclusion so either list some valid arguments for your point (Since its not there in the benchmarks) or edit the conclusion.
danidentity - Tuesday, December 7, 2004 - link
Wesley, thanks for including tests from a more comparable AMD CPU. Listening to your readers is always appreciated.Wesley Fink - Tuesday, December 7, 2004 - link
#4,#5,#6 - The Athlon 64 results with the FX55 were included for Reference, and not direct comparison. However you do make a good point.The closest A64 we had in the lab to a 3.6GHz 560 was the 3500+ based on the 90nm process. This should provide an advantage to the Intel 560. Since there are complaints here the FX55 is too high end, these new tests tilted toward Intel should balance the playing field. The 3500+ costs about $265 and the Intel 560 (3.6GHz) is about $455, so the 560 is about 70% more expensive than the 3500+.
The added 3500+ benchmarks were also an opportunity to test with the SAME ATI X800XT PCIe we used in benchmarking the 915 boards. Enjoy!
Color codes have been updated and there are now 3500+ results on the Gigabyte nForce4 with the ATI X800XT PCIe in all benchmarks.
Original plans were to include the Intel 570 in this roundup, but much of the testing was already done when the Intel 3.8GHz CPU was launched. This Intel 3.8 is priced at around $800, which is very close to the FX55. You can see how it compares to the FX55 in performance in Anand's launch article at http://www.anandtech.com/cpuchipsets/showdoc.aspx?...
CrystalBay - Tuesday, December 7, 2004 - link
It's a pity that all these 4 dimm slot, dual channelMB's have such a rough time doing it. Imagine trying
to run 4 1GB dimms in DC, this goes for ddr1 as well 2.
Glassmaster - Tuesday, December 7, 2004 - link
Now that Northwood and 865/875 are on the way out, only a fool would buy Intel.Glassmaster.