This tower computer is built to support the Merced Highway 99 experiment - obsolete as of March 2013
This is nice information to have about your machine because it will be needed
0) to answer the base question in every help forum,
1) if your hard drive fails,
2) you catch a virus, that is format the drives and a fresh reinstall
3) you want to change hardware or the operating system,
4) you want to change drivers,
5) the operating system is corrupted
Model: CIFE-RGW MERCED 99
On 4/1/2008 I [a former CEE241 TA now CIFE research assistant] purchased a 1-year old dual-core XP-OS laptop on ebay, avoiding Vista due to issues found during the CEE241 course. After four years and numerous changes to hardware, software, and settings the laptop successfully constructed a 50MB Autocad Civil 3D model using the 2011 x64 release. 1) Through modification an existing and infamous heating issue with this laptop model was aggravated; an auxiliary fan was necessary for skype chats and 3D modeling. 2) The duration to complete the civil 3D model was not consistent with the intuitive expectation of the Merced 99 project management staff. Professor Fischer decided that this laptop has fulfilled its useful life as a VDC research machine and authorized a replacement machine. While the dv2120 laptop integrated graphics supports a respectable 512MB shared DDR2 800Mhz memory, the lack of a dedicated graphics card is the core limitation; therefore the next machine will have a dedicated graphics cards. Therefore, research for guideline specifications for a new machine were initiated.
A new machine was ordered
The new machine arrived. The following steps were followed to baseline the machine:
Today at startup a warning indicated that the harddrive files were corrupted, a disk scan corrected the issue and everything now works fine. GPU testing continues. Trying out the WHIPTHREAD command in C3D set to 3 http://www.theswamp.org/index.php?topic=16803.0;all
Posted in AutoCAD forum: my 2006 AMD laptop with a GeForce GO 6150 (similar to your graphics card) and maximum shared memory (some conflicting information on what this actually is, so say 512MB) compared side-by-side to my tower an AMD BE with a Quadro FX 5600 monster card: in 2D wireframe orbit the 6150 is incredibly faster, in 3D wireframe orbit they are comparable, in anything shaded or realistic the Quadro rules (similar results with an ATI v7800). So, maybe the graphics card means nothing, and I have read posts in forums that have found similar intuition. I am not going back to the 6150 but there was no 'event horizon' passed by going to the Quadro. Note, for the GeForce Go 6150 I used a driver and modified file from laptopvideo2go (allowed the driver to install) and not the driver posted to the HP site, this driver change (after several driver updates over time I found that some worked better than others and recently used 259.47) this made an incredible performance difference.
The comparison of the new ATI and used Nvidia graphics cards indicates that the buswidth may be the key characteristic. To test, a GeForce GTX 280 with a buswidth of 512-bit (650Mhz, 1GB, 240-core) was purchased today and will be tested to confirm an improvement. This card has a lower capacity than the tested Quadro FX 5600 and ATI v7800 cards, except the nearly double buswidth. After 3D orbit testing minimal improvements but in the right direction. The result is a 23-second start to finish time, a 50% improvement over the baseline 42-seconds and major improvement over the up to 1.5-minute observed. The buswidth does not appear a pragmatically significant factor.
5/25/2011 What next?
Hardware and software configuration testing is essentially completed - daily routine is back toC3D modeling. Corsair PC2000 memory is being used but with the CPU tuned to 4.3Ghz the memory is unstable. Even at baseline tuning the memory does not feel solid as the initial Patriot memory.
The new memory has stabilized after trying various configurations and NtfsMemoryUsage set to 2 that according to Win7 documentation should have no effect the Civil 3D orbit in 2D wireframe and 3D wireframe. and regen times have all improved a lot. The hardware configuration is complete.
8/29/2011 Opened the case and pointed the 14” floor fan (from laptop tests) into the case - temp at idle dropped from 45°C to 33°C
8/30/2011 Left the fan off today, just removing the case side panel lowers temperature from 45°C~50°C to 40°C
9/7/2011 Crashed OCZ REVO main drive OS: looking for solution but most threads end with reformatting and reinstall
|Yahoo Answers 1||restore - push F8 a lot, select option “Repair Computer” then logon before doing a system restore||x|
|Yahoo Answers 2||reinstall|
|Microsoft Answers 1||delete file - diskBoot Win7, click on repair, at some point it should ask about selecting your OS, but instead click the button that says load drivers, this will bring up a window that will let you explore your computer and delete the file to the recycle bin. The location should be C:\Windows\system32\nvvsvc.exe||see Nvidia|
|Make Use Of||log-in then goto services.msc start» search» type services.msc» highlight nvvsvc» right click» properties» startup tab» pick manual» apply» ok and set Nvidia drivr line to manual, and Nvidia stereoscopic 3D line to disabled. Then reboot||cannot|
|team liquid forum||reinstall - keep booting from disc and finally got the disc to display the setup|
|Microsoft Answers 2||reinstall - start from scratch and spend the next week or two reinstalling, updating, and reconfiguring all my highly technical work software.|
|Microsoft Technet||clear corrupt Event Viewer service logs manually||x|
|Kioshea forum||Unplug all USB items and restart||x|
|Techspot forum||boot in “Safe Mode with Command Prompt” and then switch to the CD Drive D: and launch the OS setup. I did an XP “Upgrade” and now all is well.|
|Nvidia forum||delete nvvsvc.exe (renamed with *.old extension)||x|
|personal blog||copy and paste contents of windows/system32/config/RegBack folder from OS that works to the OS that does not (nice I have two nearly identical installs of OS on separate drives next to each other)|
Nice, it has a name - Black Screen of Death
9/9/2011 I give up - moving to backup SATA drive - installing a few applications to bring it up to date, in the next couple weeks the PCIe drive will be formated and a fresh reinstall, then will be the new backup.
Now that the C3D modeling is completed the capacity provided by the PCIe drive is probably excessive and the SATA SSD is likley sufficient No - the PCIe drive is much better and the SATA SSD is annoyingly slow. As soon as the PCIe drive is ready I will switch back.
10/04/2011 Contacted OCZ to drop-off the revo drive and pickup a replacement but they have none in inventory and must test this drive before deciding to replace or just give it back. The turnaround on the RMA is a couple weeks. So I am using the Muskin SATA, it works, and the RMA process is sufficiently annoying that I have procrastinated to 11/18/2011 and probably will continue to do so for another week.
11/18/2011 I pulled the Revo OCZ drive and had to lower the CPU/NB speed to boot, else it went to blue screen when logging into windows. assuming this is because the OS was using the drive for virtual memory. Bumped the HTlink up a bit at the same time from 2100Mhz
11/28/2011 Dropped Revodrive off at OCZ san jose - Danny RMA rep; in testing now.
8/9/2012 changed back to CMX8GX3M2A2000C9 due to daily blue screen crash; memory felt warm, might be overheating - when time will test modules and if pass then try a memory fan
15 minutes later put the KHX2400C11D3K4/8GX back in - crashed with blue screen again, try dialing back the cpu
10/12/2012 still using the CMX8GX3M2A2000C9 memory due to stability issues. Also, compared to a new W110er laptop with Intel i7-3820QM and 8GB 1866Mhz memory - the laptop works as good and sometimes 8x better. One exception is importing C3D dwg file to Navisworks, the laptop locks up and the AMD machine has no problem.
01/23/2013 just read that ASUS mb stability improves if two memory slots left open - ordering 16Gb (2×8) modules to replace 16GB (4×4) modules. This thread references a different motherboard but we will see if it behaves similarly. http://www.tomshardware.com/forum/316710-30-will-1600-work-phenom
3/21/2013 never did order the memory - this machine feels obsolete, only the FX5600 and the PhenomII for some specific applications are useful for Navisworks import and visualization.
|Low Power State Link Speed|
|Wake on Magic packet|
|Wake on pattern|
|WakeOnLAN from PowerOff|
|Attribute||V7800||Quadro FX 5600||GeForce GTX 280|
|Form||PCIe 2.1 x16||PCIe 2.0 x16||PCIe 2.0 x16|
|Series||FirePro 3D V7800||Quadro FX 5600 –> GeForce 8800 ultra||GeForce GTX 280|
|Memory Bus Width||256-bit||384-bit||512-bit|
|Max. Shared Memory|
|power consumption||138w 1×6-pin||171w 2×6-pin||236w|
|Announced||April 26, 2010||2006||June 17, 2008|
|Device PCI ID|
|3rd party website for card|
|wikipedia page for card||1||1|
|vendor page for card|
|forum posting for card (guru3D)|
|Civil 3D||visual OK||visual good||visual good|
|test 1/0||degree||test app.||driver||website||driver code||date||installed||notes|
|ACAD C3D||2011 ACAD Perf Win7x64||Nvidia||2/4/2011||02/16/2012||Muskin|
|ACAD C3D||265.70 Win7x64 2012Cert||Nvidia|
|ACAD C3D||275.65 Server2008R2x64||Nvidia||184.108.40.20665||7/11/2011||8/24/2011||Revodrive|
|ACAD C3D||275.65 Win7x64||Nvidia||4/4/2011||8/24/2011||Revodrive|
|ACAD C3D||276.14 Perf Server2008R2x64||Nvidia||10/4/2011||02/16/2012||Muskin, ACAD Perf is not integrated in driver|
website CASE MID TOWER ANTEC P193 BLACK
POWER SUPPLY CORSAIR 750W SLI READY PSCS750
CF-LIQ-H50 LIQUID COOLER CORSAIR H50
Solid State Drive Alignment Resource, these drives were installed new with Win7 - except for the Intel X-25 drive which is a clone from a 7200rpm drive - so alignment should not be an issue. http://www.ocztechnologyforum.com/forum/showthread.php?93422-Very-Slow-OCZ-Revodrive-X2-220GB
OCZ REVO DRIVE 240GB 1RVDX0240
(removed 11/1/2011) 2nd boot Harddrive with x32 OS
[Backup, nearly duplicate (and nearly full; 90%) of Revo drive]
Mushkin Callisto Deluxe 120GB 34nm SandForce SF-1200 MKNSSDCL120GB-DX
Alignment check: cmd → fsutil fsinfo ntfsinfo C: [not real clear how this is done, search the web for some confusing documentation]
ATTO Disk Benchmark, ordered by transfer size KB; total length 256MB; Overlapped I/0; Queue depth 4
|State||Full OS||Partitioned||Full OS||Full OS||Full OS|
Summary: for future use the exact results of this summary are not as important as a benchmark for how to obtain these results and compare them. As you can see the performance numbers and the real world feel of each drive differs. From this comparison, I'd predict that an Intel X25-M G2 drive with x64 may perform the best of all these drives. I am not the first person to notice that the REVODrive is fast but not as fast as advertised and is comparable to other SATA SSD.
Write summary: The OCZ REVO and Intel X25-M G2 drives have comparable 70MB/s write speed and the Mushkin Callisto Deluxe has nearly double the speed of 230MB/s.
Read summary: All three drives top at 230MB/s, the OCZ REVO and Intel X25-M G2 drives have nearly equal ramp-up and lead over the Mushkin Callisto Deluxe by twice the speed in the smaller blocks and by 30% near where they converge at the 128KB block size. At the top end the Mushkin Callisto Deluxe has the stablest speeds of the three drives.
Pragmatic summary: Using these three drives, the Intel X25-M G2 with the x32 OS (have not tried x64) has the fastest startup time by a wide feel, in use the OCZ REVO performs the best with Civil 3D (C3D is balky with the other two drives or possibly the OS install differs, the Intel drive has an x32 OS so C3D is nearly unusable) comparable with both the Intel X25-M G2 and Mushkin Callisto Deluxe for any other application. The OCZ REVO in this test is running at half the advertised speed.
(added 12/5/2011 for virtual memory) fujitsu MHZ2320BJ G2, 7200 RPM 320GB
Motherboard supports 4 x DIMM, 16 GB, DDR3 2000(O.C.)/1600/1333/1066 ECC,Non-ECC,Un-buffered Memory
Timings (nanoseconds, lower is better): BIOS to AUTO
Forum question: What is my memory speed
Approach: For the MER99 experiment, first we developed an approach using an HP dv2000 laptop. During the process I customized it with a 2.3GHz processor, Intel solid state drive, a 64x OS, and 8GB Muskin memory. After these modifications the laptop was adequate for integrated models of building projects - the 3D models were provided and not modeled with this laptop. The approach to modify the laptop was driven by resolving performance issues as they arose but tuning was not an option and I had not discovered benchmarking applications. The MER99 experiment uses AutoCAD Civil 3D and the scope of the 3D model is greater then a laptop form can support - so following the approach of resolving issues with increased hardware we changed to a tower form and added the ability to tune. The baseline tower using the top AMD processor, harddrive, and RAM memory was inadequate. A tuning procedure became necessary. Our approach to tuning is 1) research, 2) trial and error, 3) benchmark, 4) document. For research we rely on web searches of gaming and overclock web forums. For trial an error the research assistant plans a series of test parameters and then makes changes incrementally. For benchmark two software applications are used, these are CPU-Z to record the settings, and GeekBench to measure a performance index. The results are then documented in this wiki and periodically posted to a web forum for peer review.
Benchmarking tools: For the Civil 3D application we need a benchmark to test single core processing and are not concerned with multicore metrics
Tuning tools [none of these were tested for this experiment]: Check if only CMOS clearable boards
|Try||CPU||CPUbus||Ghz||PCIe||DRAM||CPU/NB||HT link||CPU v||CPU/NB v||DRAM v||HT v||NB v||timing||Div:DRAM||approach||result*||CPU-Z||Geekbench|
|7||16.5||229||3.8||100||1527||2061||Auto||1.392||1.1||1.65||1.2||1.1||9-9-9-24-41-2T||3/10 722.4||BIOS tune||OK|
|9||18||217||3.9||100||Auto||Auto||2384||1.3625||1.2||1.65||1.2||1.1||9-9-9-24-35-2T||3/10 722.4||BIOS tune||43°C||9/28/2011||8216|
|11||17||217||3.7||100||Auto||2167||1734||1.65||1.2||9-9-9-24-34-2T||3/10 722.4||BIOS tune||40°C||9/30/2011||7704|
|14||18||217||3.9||100||Auto||Auto:2167||Auto:2167||1.452||Auto||1.65||Auto||Auto||9-9-9-24-34-2T||3/10 722.4||BIOS tune||43°C||10/11/2011||8184|
|Try||CPU||CPUbus||Ghz||PCIe||DRAM||CPU/NB||HT link||CPU v||CPU/NB v||DRAM v||HT v||NB v||timing||Div:DRAM||approach||result*||CPU-Z||Geekbench|
|30||18.5||210||3.9||100||Auto||2950||2739||Auto:1.40||Auto||1.50||Auto||Auto||7-8-7-24-27-1T||3/8 561.9||BIOS||bit unstable||20111011||8284|
|32||18.5||210||3.9||100||Auto||2949||2106||Auto:1.40||Auto||1.50||Auto||Auto||7-8-7-24-27-1T||3/8 561.9||BIOS||20111117||8286 v2.1.12|
|41D||18.5||220||4.1||100||1766||A2207||A1987||1.500||1.35||1.70||Auto:||Auto:||8-A9-A9-A24-A40-1T||1/4 883||BIOS||48°C - daily crash||20120702||10743v2.2.7|
*max loaded temp for AMD Phenom II 55°C - result temp measured at end of geekbench run
*change in v2.1.12 to v2.2.3 resulted in 25% improvement in score and so are not comparable
1crashed when skype screenshare started by conference 3rd party - otherwise worked
* indicates installed
forcing driver instal for Server versions
Give the date, drive, and version
careful, some applications install to the C: by default, this drive is F
Clear these folders (Win7) if you need drive space; I did for a virtual machine to run a 32-bit OS.
Next build - attempt #2 at field PC;
investigate: change to intel processors for CAD applications due to single core speed http://www.informationweek.com/news/231600921 TDP laptop 35w TDP tower 130w
|AMD 10h||65nm||9/2007||Black Edition||Socket AM3|
|Nehalem (microarchitecture)||45nm||9/2008||x2||Bloomfield (microprocessor)||LGA 1366|
|Sandy Bridge||32nm||1/2011||x4||K-edition 2500K, 2600K & 3930K||LGA 1155 LGA 2011|
|Bulldozer (processor)||32nm||9/2011||Black Edition||Socket AM3+|
The bottle-neck in the Merced99 experiment system is the harddrive and memory. So if we focus on that in the next machines:
Try: 13.3” ASUS $400 ebay computer with 540M gpu - replace x2 2410M cpu with $1200 x4 XM unlocked processor and install $150 1600Mkz memory = $1750 for nonexistant 13” laptop.
Issue is socket compatibility - research of blogs finds they may be compatible.
These optional and add-on changes will be required over the next five years either as entirely new machines or incremental changes to the existing cases:
These changes should give the machines another 8 years but is building new machines and reusing the existing cases and power supplies. It will also take longer to reconfigure and require some research and knowledge-building. These machines will not have service contracts so we have to keep them going ourselves but I have not seen any Dell technicians around anyways. The existing parts that are removed can be sold on ebay.
Some of these specifications sound extreme and that building machines rather than buying as a complete box is a hassle but on the Merced project these are the type of specifications and approach to computing that is working. For the longterm the core supporting factor is a motherboard with an unlocked BIOS, slots and room for later expansion, cooling, and a tunable processor.
http://www.gskill.us/forum/showthread.php?s=5937aa96edec4d46b475ac0e8092be6e&t=9913&page=2 G.SKILL Ripjaws Z Series 16GB (4 x 4GB) 240-Pin DDR3 SDRAM DDR3 2400 (PC3 19200) Desktop Memory Model F3-19200CL9Q-16GBZMD
Lower to approach 1333 and CAS2
Test: you can give this a try. open the autodesk sample file:
C:\Program Files\Autodesk\AutoCAD Civil 3D 2012\Help\Civil Tutorials\Drawings\Corridor-5c.dwg. set interval to 1m dont' touch anything else. do that in corridor property. then hit ok then start timing til become responsive. took me .34 seconds. i think that's not bad.
but what i realized from this is again emphasize how much time we waste on waiting for the program to responde. only if it can use more cores even 4 cores we would see a marked improvement. imaging cutting down all these time by 75%. http://forums.autodesk.com/t5/AutoCAD-Civil-3D-Customization/graphics-card-comparison-test/m-p/3350335#M6203
try: http://www.youtube.com/watch?v=bzHiAgE1LKU&feature=player_embedded#! turn off “Enable Orbit Auto Target.”