NVIDIA redeems itself... almost

I am glad that I held my fire and did not post yesterday.

As promised, NVIDIA delivered a device driver for its 8800GTX and 8800GTS graphics cards by the end of January. It is still Beta, but Version 100.59 was delivered today as promised, right on the heels of Version 100.54, which was released just the day before when Vista was officially launched to the public.

What has angered a lot of NVIDIA's customers about this entire Vista driver episode, which continues today, is NVIDIA's arrogance towards its customers.

It would have been so easy for NVIDIA to explain to its customers why a Vista driver for its G80-based graphics cards was not made available earlier. Its weak excuse that NVIDIA needed the time to test the driver to assure a high quality, working driver is belied by its release of a beta Version 100.54 on the Vista launch date, a driver that according to most, performed worse and had more defects than the leaked Version 100.30. Then to follow it up the very next day with another beta Version 100.59 without any explanation adds insult to injury.

This arrogance appears to be symptomatic of today's large corporations. NVIDIA is not alone, and the only way to correct this type of behavior is for consumers to speak out with their purse strings. Rant off.

The first sign of good news is the size of the Version 100.59 driver versus Version 100.30 - it is 47,026 KB versus 20,747 KB in size.

I uninstalled Version 100.30 using the Control Panel "Uninstall or change a program" of Vista which required a system restart.

At the start of installing Version 100.50, Vista displayed the Windows Security message: "Windows can't verify the publisher of this driver software."

I guess NVIDIA still does not have the appropriate trusted certificate for its drivers. In fairness to NVIDIA, the Version 100.59 driver installation completed without a hitch.

The "swapped monitors" problem that was present in Version 100.30 has not been fixed in Version 100.59.

And I am not about to test the problem of the video driver unable to turn the monitor(s) back on after powering them off under a power plan. I'll leave that for others to test.

The first test I ran was 3DMark06. To be honest, I did not expect it to run. I was surprised that it actually ran and completed.

I was even more surprised with the resulting performance score of 8731 3DMarks which bested the score of 8409 3DMarks on Windows XP MCE 2005 under almost the same exact conditions. Actually, Vista was handicapped with the Windows "Sidebar" running during the test. That is a promising start for the NVIDIA driver. Here's the 3DMark comparison.

The next test I tried was to run the new Vista Ultimate Extra "Hold 'Em" poker game which did not run properly with Version 100.30. It still does not run properly with Version 100.59, resulting in the same problem: flashing text.

At least for now, I have a working video driver which I can use to conduct further testing.

Thanks NVIDIA, but you could have made it so much easier on yourself and your customers if you simply took the time to inform them. No news is worse than bad news.

Windows Ultimate Extras

The first 2 "Windows Ultimate Extras" for Vista Ultimate became available via Windows Update today: BitLocker and EFS Enhancements and Hold Em Poker Game. However, the much anticipated Windows DreamScene is not available yet.

I was unable to play the new "Hold Em Poker Game" on Nighthawk because a fully working device driver for my NVIDIA 8800GTS graphics card is still not available. And the natives are getting restless over at the NVIDIA Forum. A full scale revolt is not unlikely if January ends without a G80 driver available from NVIDIA. Heads will roll.

ATI has released their Vista drivers but there is still no support for my TV Wonder 650 which is "Certified for Windows Vista" amd "Ready for the Future with a Windows Vista on your PC." I guess the future isn't here yet.

There were also 7 "Updates for Windows Vista for x64-based Systems" including 1 Important (KB930857), 5 Recommended (KB929777, KB929427, KB931573, KB928089 and KB929685) and 1 Optional (KB929735). Installing these and the Windows Ultimate Extras required 2 restarts of Vista.

International users of Vista will be very happy to know that 15 Windows Vista Ultimate Language Packs are also now available via Windows Update: Chinese (Hong Kong S.A.R.), Chinese (Simplified), Danish, Dutch, Finnish, French, German, Italian, Japanese, Korean, Norwegian, Portuguese, Russian, Spanish, and Swedish).

The great hunt for Vista device drivers continue...

Vista hype nears crescendo

Vista is less than 2 days away from its release to the general public, and the hype is building to a climax. The various Vista news groups are seeing a lot of activity as people are already starting to get their hands on Vista and asking lots of questions. My local CompUSA store will open its doors at 10 PM tomorrow night to start selling Vista, Office 2007 and discounted merchandise at midnight.

I checked the various store ads but I didn't really see any bargains. Most discounts are tied to purchase of Vista or Office 2007, and I wonder whether some of the software which will be given away for "free" with purchase of Vista will actually work in Vista. We'll see how this launch compares with the launch of Windows 95, NT and XP.

I was one of the faithful waiting to get my hands on Vista. It is a good operating system based on my experience with it to date. I will withhold judgement on whether it is a great operating system just yet. I would be much more positive about Vista if more working device drivers were available, but that's the way it is with the early adoption of any operating system.

There are some bright spots in the driver scene. I was pleasantly surprised to see Nokia support Vista early on with their PC Suite which allowed me to synch my Nokia 6682 mobile phone with Vista Ultimate x64.

Another bright spot is Vista support for networked printers "out of the box." I was able to connect both Blackbird and Nighthawk to my HP 952C Deskjet printer attached to a Netgear Print Server PS110. I simply pointed Vista Ultimate x64 to its static TCP/IP address on my network and specified it as a "generic 2-port print server" and selected the printer as the HP 952C Deskjet printer on Port 1. It was very easy and painless.

Actually, networking is very simple with Vista, in general. Setting up file sharing with Vista is easy.

I will be waiting for the release of the NVIDIA driver for my BFG 8800GTS graphics card on Vista launch date (January 30, 2007) before proceeding with the rest of my discovery and testing on Nighthawk.

My wife was recently provided with a Core 2 Duo laptop for use at home by her work, so I might be able to keep Nighthawk permanently for my own use for more experimentation. We shall see.

Another BIOS update from ASUS

Almost right on the heels of the BIOS Version 0910 official release just a few days ago, ASUS released BIOS Version 1004 (dated 01/22/2007) for the P5B Deluxe WiFi-AP motherboard yesterday.

It makes one wonder why ASUS felt the need to release this new version so soon. The release notes mention "enhance memory compatibility" and "support CONROE E0 CPU (FSB 1333)" - both are good updates, but hardly worth rushing the release of this version.

I suspect this release has something more to do with the impending release of Vista to the general public in 3 days. There are reports in the ASUS forums of Vista KMS Activation problems with the prior BIOS Version 0910.

It is also curious that BIOS Version 1004 does not appear in the "BIOS" or "Latest" sections of the ASUS download site. One must go to the "Map" section to find and download this version.

Regardless, I downloaded and installed BIOS Version 1004 on Nighthawk last night from a USB drive using the EZ Flash 2 tool in BIOS itself. This version did not require reloading my BIOS settings as BIOS Version 0910 did.

I ran my battery of stability tests overnight and Vista Ultimate x64 (RTM Build 6000) has been running on Nighthawk (with its overclocked settings) since this morning without any problem.

I did not see any new settings in BIOS, nor any performance improvements. At least, BIOS Release 1004 appears to be stable so far.

More Vista driver woes

Vista is now only 4 short days away from being released to the general public On January 30, 2007.

Based on my experience running Vista Ultimate x64 (RTM Build 6000) on Blackhawk since November 17, 2006 and on Nighthawk since January 20, 2007, I have very little confidence that computer component manufacturers will have non-Beta device drivers ready for the Vista launch.

In this post, I will focus on add-on TV Tuner cards.

Shortly after I installed Vista on Blackhawk in November last year, I went out to purchase the ATI TV Wonder 650 since it was advertised as "Certified for Windows Vista." I don't know how ATI can claim that this hardware is "Certified for Windows Vista" since there was, and there is still no driver to make this device work in Vista x64. I should have read the fine print - fool, me.

So I went out and purchased the Hauppauge PVR-150 since it is also "Certified for Windows Vista" and a Beta driver was available for download. It did not and it still does not work in Vista x64. Fool, me - twice!

As previously noted in a prior post, the Hauppauge PVR-150 TV Tuner card was recognized when I installed Vista Ultimate x64 (RTM Build 6000) on Nighthawk and it loaded the Hauppauge driver. Windows Media Center allowed me to configure the TV tuner. Unfortunately, the Hauppauge PVR-150 TV Tuner card still does not work in Vista x64. The frame rate for the TV signal is about 1 frame a minute, at best. Maybe I should have bought a quad core processor so I can watch TV on my Vista computer.

Today, Vista detected an updated driver for the Hauppauge PVR-150. Hooray! I'm finally going to be able to watch TV on my Vista computer. I installed the driver and restarted Nighthawk with anticipation.

The Hauppauge PVR-150 TV Tuner card still does not work in Vista x64. The frame rate of the TV signal has improved to 2 frames per minute. I guess doubling the frame rate is good progress. At least Nighthawk is not crashing with the "Blue Screen of Death."

The latest non-working version of the Hauppauge driver is Version 6.0.6000.16488 dated January 8, 2007. Hey, at least Hauppauge has a device driver while ATI has yet to release one.

Do you still want to be on the bleeding edge? Do you believe that a deluge of non-Beta device drivers will be delivered on January 30th when Vista is launched?

The weakest links

The ASUS P5B Deluxe WiFi-AP is a solid, high-quality, versatile, and overclocking-friendly motherboard. Unfortunately, it has 2 very glaring defects: the JMicron Controller and the SoundMAX Audio device. You don't have to search the Internet far to find complaints about these two on-board devices.

I have already discussed the many problems of the JMicron Controller in an earlier post. I am now going to turn my attention to the on-board SoundMAX audio device.

When I installed Vista Ultimate x64 (RTM Build 6000) on Nighthawk, Vista detected the on-board SoundMAX audio device and loaded its Microsoft audio driver Version 6.0.6000.16386 dated 06/21/2006. And I had sound from the "Speakers" and "Digital Output Device (SPDIF)" but not from my headphones connected via the front panel.

However, Nighthawk is deaf! None of the "Recording" (input) devices work. Not the 2 microphone inputs. Not the 2 "line in" connections. Zero. Nada. Zilch.

I did not expect much from the default driver, so I was not entirely surprised. I knew I could always download the latest SoundMAX driver from ASUS. And ASUS has SoundMAX drivers on their Web site for Vista: "SoundMAX Audio Driver v6.10.01.5110 for 32bit Windows Vista (WHQL) & v6.10.01.6020 for 64bit Vista(beta drivers)"

I downloaded and installed the SoundMAX Vista drivers. The first hint of trouble was a warning message indicating that Vista could not "verify the publisher of this driver software." I've seen this before, so I allowed the driver to be installed.

The Vista Device Manager confirmed that the SoundMax driver Version 6.10.01.6020 dated 11/02/2006 was indeed installed. However, note the warning symbol (exclamation mark on a yellow triangular sign) superimposed on the SoundMAX Integrated Digital HD Audio" entry in the device list.

OK. Maybe I just needed to restart as indicated by a message after the driver installation. I restarted and the SoundMAX audio device was still not working. I checked its status in the Vista Device Manager which informed me: "Windows cannot load the device driver for this hardware. The driver may be corrupted or missing. (Code 39)"

So much for that exercise. I used the Vista Device Manager to "Roll back driver" to the previous version - the Microsoft default driver which at least works for audio output.

I was aware of the many SoundMAX-related problems facing owners of the ASUS P5B Deluxe WiFi-AP motherboard before I bought it. That's the reason why I did not even bother with it and installed a SoundBlaster XFi sound card in Blackbird. I will probably do the same for Nighthawk.

However, don't think that the grass is much greener on the other side. It is a little bit greener, but no that much greener. SoundBlaster users including myself have been facing similar challenges with device drivers.

Do you still want to be on the bleeding edge and install Vista when it becomes available to the general public in 5 days?

BIOS updated to Version 0910

ASUS just officially released BIOS Version 0910 for the ASUS P5B Deluxe WiFi-AP motherboard. Nighthawk is currently running on BIOS Version 0804.

I usually wait awhile to hear about other people's experiences before I perform BIOS updates. In this particular case, I was not overly concerned since Nighthawk is a test system and this motherboard has the ASUS "CrashFree BIOS" feature.

I downloaded BIOS Version 0910 and installed it on Nighthawk using the AsusTek EZ Flash 2 in the BIOS from a USB drive. One always holds his/her breath while performing BIOS updates. I do and I did again in this case.

So far, BIOS Version 0910 for the the ASUS P5B Deluxe WiFi-AP appears to be stable.

I ran 3DMark06 from Windows XP MCE 2005 to check for any performance improvement from the BIOS upgrade. I did not see any. In fact, performance appears to have decreased some but this may be due to Windows XP MCE 2005 now only accessing 2GB of memory instead of 2.9GB before I enabled "Memory Remapping" in BIOS.

I will be running my usual battery of stability testing which will delay my experiments with using the on-board SoundMAX audio from Vista Ultimate x64 (RTM Build 6000).

Speaking of Vista, I just noticed that ASUS now has the P5B Premium Vista Edition and P5B-Plus Vista Edition motherboards. Hmmm...

Essential Vista?

NVIDIA has a logo with the words "Essential Vista" associated with their GeForce 8800 products, and the 8800GTX and 8800GTS graphics cards are the first to support the new DirectX 10 introduced with Microsoft's Vista operating system.

One would assume that a product designed for Vista will work in Vista, right? A lot of people believed so when they ran out to buy these cards when they were introduced late last year. Unfortunately, they did not read the fine print.

I read the fine print - there were and still no device driver for these graphics cards from NVIDIA.

I knew this fact when I bought the BFG 8800GTS but I had no plans on using it with Vista immediately. Until this past weekend, I was only using the BFG 8800GTS card with Windows XP MCE 2005.

Last week, a driver (alpha or beta version?) was leaked on the Internet - Version 100.30 for both Vista x86 and x64. Reports from the brave souls who installed this driver indicated that although short on features, this driver actually worked and allowed the use of Vista's "Aero" interface, albeit with some problems here and there.

I took the plunge and installed the Version 100.30 driver on Nighthawk. I am usually not this adventurous, but Nighthawk is a test system.

The driver installed without any problem, but I encountered the first problem after re-booting. I have 2 monitors connected to Nighthawk, and the driver decided to swap the monitors, i.e., physical Monitor #2 (on the right) is now Monitor #1 and physical monitor #1 (on the left) is now Monitor #2. Confused yet? I was! Other individuals have confirmed this problem.

No sweat! I switched the monitors in Vista's "Display Settings" and specified Monitor #2 as my main monitor and extend the desktop to Monitor #1. This solved the problem.

I noticed another minor problem where the mouse pointer would be left in the "home" position (upper left-hand corner) of Monitor #1 (now the monitor on the right) as I moved the pointer from the right monitor to the left monitor. To get rid of this problem, I would simply move the pointer back to the right monitor to "retrieve" it.

Why would one put up with a driver with obvious defects and incomplete feature set (the Control Panel is severely lacking in features)? Because beggars can't be choosy...

I re-ran Vista's Performance Check to update my Windows Experience Index base score and Nighthawk scored 5.8! After the new base score was updated, the "Aero" interface was automatically enabled. Alright!

A more serious problem happened later when my monitors were automatically turned off after 20 minutes. Vista with the Version 100.30 driver is unable to turn the monitors back on! This problem has been confirmed by other users as well.

I was reluctant to simply turn off the power to Nighthawk using the power switch on the case. I did not want to introduce any data corruption in Nighthawk's RAID0 arrays. I would have taken the chance if Nighthawk had RAID1 or RAID5 arrays instead of RAID0. I knew that Nighthawk was still running since I could access a shared folder on Nighthawk from Blackbird.

How can I perform an orderly shut down of Nighthawk while blind, i.e., no video? I decided to take a break for lunch and think about it.

The first step was to determine what state Nighthawk was in so I could visualize what would be displayed on the screen. I remembered that I had set "On resume, display logon screen" on Nighthawk's "Screen Saver" setting. Since the monitors powered off, the "logon screen" would be the only screen Nighthawk could be on.

I used Blackbird to memorize the keystrokes required to properly shut down Nighthawk. Using the mouse without a video signal is impossible.

The keystroke sequence I used:

"Home" to position the cursor at the beginning of the password field.

"Shift End" to highlight and select whatever was in the password field since I was trying different key combinations prior to this.

"Delete" to delete whatever was in the password field.

My password. (I won't be sharing this with you.)

"Start" (or "Windows" key) to bring up the "Start" menu.

"Right arrow" three times to navigate to the "Shut down" options.

"u" to select "Shut down."

And Nighthawk gracefully shut down!

After restarting Nighthawk, I promptly set the "Power plan" to never turn off the monitor.

I immediately decided to run the 3DMark06 benchmark to compare performance in Vista to Windows XP MCE 2005 using the same exact hardware platform. Unfortunately, a problem with the Version 100.30 driver prevents 3DMark06 from running the benchmark. I know it's not a problem with 3DMark06 running on Vista because I am able to run it on Blackbird with my BFG 7950GT graphics card.

Vista displays an error message: "Display driver stopped responding and has recovered." This is one huge advantage of Vista over Windows XP. A similar problem on Windows XP would have caused a BSOD (Blue Screen of Death).

3DMark06 also displayed an error message (and did not crash): "IDirect3DDevice9::Present failed: Device lost (D3DERR_DEVICELOST)"

Hopefully, NVIDIA will have a "real" working driver for the GeForce 8800 graphics card as they have been promising when Vista is made available to the general public in 8 days.

At least Nighthawk's "sight" has been partially restored. Its sense of "hearing" will be the next focus of my attention. The sound output works with the default Microsoft driver, but no sound input works.

On the bleeding edge

I have been running Vista Ultimate x64 (RTM Build 6000) since November 17, 2006 when it was initially made available to MSDN subscribers. Prior to that, I also installed and tested both Vista Release Candidates 1 and 2. In fact, I built both Blackbird and Nighthawk specifically to run Vista.

Since installing Vista on Blackbird, I have spent a lot of time searching for device drivers, and scouring numerous forums and news groups for Vista tips and help so that I can use my various computer devices and programs with Vista. Device drivers are either not available and if they are available, the device drivers are in beta and incomplete in functionality. Some programs also do not work very well in Vista, if at all. I probably made it more difficult for myself in choosing to run the 64-bit version of Vista.

The retail version of Vista will soon be available to the general public in 9 days.

Unless you have a lot of time and patience looking and waiting for device drivers and updated programs which work well with Vista, I would recommend waiting until more device drivers become available and the many kinks are worked out. I still have my Windows XP Professional system available as my primary system although I use Vista as much as possible.

Preparing to install Vista on Nighthawk

My plan is to install Vista Ultimate x64 (RTM Build 6000) in a RAID0 array and from a Plextor PX-755SA SATA DVD-RW optical drive, both of which are connected to the Intel ICH8R Controller on my ASUS P5B Deluxe WiFi-AP motherboard.

The Intel ICH8R Controller settings in the ASUS 0804 BIOS of Nighthawk are: SATA Configuration is "Enhanced" and "Configure SATA as RAID."

I made 2 changes in BIOS: Vanderpool Technology was enabled since I plan to run VirtualPC which may take advantage of Intel's virtualization technology; and the "Memory Remap Feature" was also enabled so Vista can take full advantage of the 4GB of memory installed in Nighthawk.

I was specially concerned about enabling the "Memory Remap Feature" since it was not clear how it would impact the existing 32-bit Windows XP MCE 2005 operating system already installed on Nighthawk. I am happy to report that Windows XP MCE 2005 works properly and the only impact of enabling the "Memory Remap Feature" is that Windows XP MCE 2005 can now only access 2GB of physical memory instead of 2.9GB it previously could access.

The last step in my preparation for installing Vista was to gather the latest drivers for the Intel ICH8R and JMicron Controllers, just in case they are required by the Vista installation process.

No drivers needed

The Vista installation on Nighthawk was started at 4:42 PM.

The "upgrade" option is disabled in the RTM release, but this option will be available in the retail version of Vista.

Vista recognized the 2 RAID0 disk arrays in Nighthawk. Since the first RAID0 array contains Windows XP MCE 2005, I selected the second RAID0 array to create a new partition to install Vista Ultimate x64 (RTM Build 6000).

At 5:00 PM (eighteen minutes later), Vista restarted from the RAID0 array for the first time to complete the installation. The screen went blank twice and Vista restarted for the second time without any warning message.

At 5:05 PM with the 2nd restart, I was presented with the Boot Manager menu to select Vista or the "previous version of Windows." I selected Vista and was prompted to select a username and password. Vista then executed a "Performance Check" and Nighthawk was assigned a Windows Experience Index base score of 1.0 because of graphics performance! Keep in mind that Nighthawk has a BFG 8800GTS graphics card installed. One can't get any lower than this score.

At 5:10 PM, Vista restarted for the 3rd and final time for the installation process. Vista immediately checked for and installed updates, including the following: driver for the Hauppauge PVR-150 TV tuner, Definition Update for Windows Defender, an update for Windows Mail Junk Email Filter for x64-based Systems, and an "ATK - system - ATK0110 ACPI Utility."

The entire Vista installation took approximately 30 minutes (wall clock time) from start to finish, and it went without a hitch.

Because of the Windows Experience Index base score of 1.0, Windows Vista Basic was enabled instead of "Aero." I was able to increase the screen resolution to 1600x1200, but with only 16-bit color. I was also happy to note the audio worked with the on-board SoundMax audio device.

After activating Vista online, I updated the chipset driver by using the Intel Chipset Software Installation Utility Version 8.1.1.1010 and installed the Intel Matrix Storage Manager Version 62.1.1002 (dated 1/4/2007).

The next step is to find and install a driver for the BFG 8800GTS graphics card so I can enable the Vista "Aero" interface and use my second monitor.

3DMark06 Speed Run

I could not resist for bragging rights.

After seeing the incredible 3DMark06 scores posted by owners of the Nvidia G80 (8800GTX & 8800GTS) graphics cards, I had to see how mine compared with these scores.

All of my 3DMark06 benchmarks up to now have been run with both of my monitors enabled, image set to "High Quality," Antialiasing set to "4x," and Anisotropic filtering set to "8x" - my preferred operational settings.

With my own tests, I have determined that having more than one monitor enabled while running the 3DMark06 benchmark decreases the score by approximately 100 3DMarks.

The incredible 3DMark06 scores I referred to earlier are also run at the "High Performance" setting with both Antialiasing and Anisotropic filtering turned off. So I changed my settings accordingly and disabled my second monitor to see how my overclocked BFG 8800GTS (Core Clock of 660 MHz and Memory Clock of 1000 MHz) would perform against the "big boys."

I was totally unprepared to see Nighthawk score a high of 10292 3DMarks! Here's the 3DMark06 comparison for this score. I ran the benchmark 7 times and all the scores were above 10000 3DMarks with the lowest being 10143 3DMarks after upgrading the Nvidia driver from 97.44 to 97.92.

That was a fun exercise but my image settings are now back to my preferred operational settings.

In yesterday's post, I reported that it was not clear what the Core Clock is for my BFG 8800GTS since both ATITool 0.26 and Everest Ultimate were showing different values. I set the Core Clock to 660 MHz and both tools were ambivalent about what the Core Clock was actually set to. This morning, I decided to check with the NVIDIA Control Panel: it reports Core Clock of 660 MHz so that's what I'll use.

This morning, ATITool 0.26 is now consistent in reporting the Core Clock at 660 MHz and Memory Clock at 1000 MHz. However, Ultimate Everest is still not sure. It reports "GPU Clock (Geometric Domain)" at 648 MHz while in the "nVidia ForceWare Clocks" section, it reports GPU at 660 MHz. I think this may be an error in Everest Ultimate.

I am now preparing Nighthawk for the installation of Vista Ultimate x64 (RTM Build 6000) alongside Windows XP MCE 2005. I will be sharing my experiences in this endeavor with you here.

Mission accomplished!

I am impressed with the overclocking potential of the Nvidia 8800GTS graphics card. My BFG version of this card managed to reach a stable overclock of Core Clock at 650 MHz and Memory Clock at 1000 MHz.

The first step in achieving these speeds was to increase the Core Clock from 600 MHz to 620 MHz and keeping the Memory Clock at 900 MHz. My usual series of tests proved the BFG 8800GTS graphics card to be stable and artifact-free at this clock speed. Nighthawk managed to score a high of 9439 3DMarks using 3DMark06. This score is an increase of 223 3DMarks over the previous high score with Core Clock at 600 MHz.

The next step step was increasing the Core Clock to 640 MHz. Note that Everest Ultimate reports the Core Clock at 648 MHz while ATIT00l 0.26 reports it as 640 MHz. Again, Nighthawk proved stable at this speed using my usual battery of tests. The top 3DMark06 benchmark score at this speed was 9567 3DMarks, an increase of 128 3DMarks.

The final step was to increase Core Clock to 660 MHz from 640 MHz and Memory Clock to 1000 MHz from 900 MHz. I normally do not like making more than one change at a time, but I decided to take a chance since I was perfectly content to fall back to the previous stable overclock setting, i.e., I really had nothing to lose and much to gain.

Note that while ATITool 0.26 shows the Core Clock to be both 650 and 660 MHz, Everest Ultimate shows the Core Clock to be both 648 and 660 MHz. I'll split the difference and call the Core Clock at 650 MHz. Can anybody explain the reason for this disparity? I wonder if my BFG 8800GTS graphics card has its Core Clock capped at 648 MHz. Remember that I first saw this reading when I set the Core Clock to 640 MHz in ATITool 0.26.

Regardless, the Core Clock is faster than the previous setting because the new 3DMark06 top score for Nighthawk is now 9605 3DMarks, an increase of 38 3DMarks. However, this small increase in 3DMark06 score may be die to the higher Memory Clock of 1000 MHz.

Here's the 3DMark06 comparison of these benchmarks.

This completes my BFG 8800GTS overclocking exercise. I am very pleased to have achieved a 26% overclock of my graphics card.

This also completes my Nighthawk overclocking exercise. I am also pleased that I was able to increase performance by slightly over 1000 3DMarks.

My Blog will continue with sharing my experiences installing Microsoft's new operating system, Vista, on Nighthawk, including a performance comparison of Vista with Windows XP MCE 2005 on the same hardware platform.

Matched 8800GTX OC Speeds

My BFG 8800GTS graphics card is now running stable at the same clock speeds as the BFG 8800GTX OC graphics card: Core Clock of 600 MHz and Memory Clock of 900 MHz.

Running ATITool 0.26 for over 2 hours scanning for artifacts and the three Nvidia 8800 Tech Demos (Smoke, Adrianne, and Froggy) to visually scan for artifacts and monitor the GPU temperature clearly show that my BFG 8800GTS has absolutely no problems running at these speeds. This graphics card has ramped up to these speeds so easily and effortless that I believe it can achieve at least a Core Clock of 650 MHz - my next target.

How about the performance of my BFG 8800GTS at these speeds?

I ran the 3DMark06 benchmark six times and my BFG 8800GTS managed to score a high of 9216 3DMarks, an increase of 174 3DMarks scored at 8800GTX clock speeds.

Here's the 3DMark06 comparison.

It's apparent that an increase in Core Clock speed has a larger impact on video performance than an equivalent increase in Memory Clock speed.

If my BFG 8800GTS graphics card can achieve a stable Core Clock speed of 650 MHz, I will be ending my video card overclocking exercise at this point even if there's an indication that it can do more.

With the release of Vista drivers which I have been patiently waiting for, I am anxious to install Vista Ultimate x64 (RTM Build 6000) on Nighthawk this coming weekend. I want to perform and complete testing so I can move the BFG 8800GTS graphics card to Blackbird where I have been running Vista Ultimate x64 (RTM Build 6000) since it was first released to MSDN subscribers on November 17, 2006.

Will my BFG 8800GTS graphics card achieve a Core Clock speed of 650 MHz? We'll see tomorrow.

8800 GTX Clock Speeds Achieved

In my prior post, I was able to overclock my BFG 8800GTS to a Core Clock of 575 MHz (the same speed as the BFG 8800GTX) and a Memory Clock of 850 Mhz. The next milestone I wanted to achieve was to reach the 900 MHz Memory Clock speed of the BFG 8800 GTX.

The first step was to increase the Memory Clock of my BFG 8800GTS graphics card to 875 MHz. After 2 hours of checking for artifacts using ATITool 0.26 and countless more hours running the three Nvidia 8800 Tech Demos (Smoke, Adrianne, and Froggy) to visually check for artifacts, my BFG 8800GTS appeared to be stable with no signs of any artifact.

I ran the 3DMark06 benchmark three times and Nighthawk scored 9042 3DMarks on its best run, an improvement of only 11 3DMarks over its best score at a Memory Clock of 850 MHz. Here's the 3DMark comparison.

The next step was to increase the Memory Clock to 900 MHz. I repeated the same routine of running ATITool 0.26 for over 2 hours to check for artifacts, and running the Nvidia 8800 Tech Demos (Smoke, Adrianne, and Froggy) for extended period of time to visually check for artifacts and to monitor the GPU temperature.

No artifacts were detected. The BFG 8800GTS running at 8800GTX speeds passed the tests with flying colors.

However, the performance benchmarks results using 3DMark06 were disappointing. My BFG 8800GTS running at 8800GTX speeds (Core Clock of 575 MHz and Memory Clock of 900 MHz) was unable to reach, much less, top its previous highest score of 9042 3DMarks achieved at a lower 875 MHz. Its results at the 900 MHz Memory Clock speed were: 8998, 9021, 9012, 9003, 9015 and 9018 3DMarks.

It is obvious that incremental increase in memory speed yields minimal, if any, performance improvement. When it comes to memory, I believe that better latency instead of faster clock speed contributes more to improved performance.

BFG Technologies now also has a BFG 8800GTX OC (Overclocked) graphics card which runs at a Core Clock of 600 MHz and a Memory Clock of 900 MHz. That's my next target. Will my 8800GTS hold up at these speeds?

Scored 9031 3DMarks!

Nighthawk has broken the 9000 3DMarks barrier!

After overclocking my BFG 8800GTS graphics card to match the specs for the BFG 8800GTS OC (overclocked) model in my previous post, I increased the GPU Core Clock to 575 MHz in the Geometric Domain (resulting in 1350 MHz in the Shader Domain) and the Memory Clock to 850 MHz in my first step to match the specs for the BFG 8800GTX model (Core Clock of 575 MHz and Memory Clock of 900 MHz).

I figured out why Everest Ultimate did not reflect the same clock values as ATITool 0.26 - I forgot to refresh the Everest Ultimate GPU display, wrongly assuming that it automatically refreshed like the "Sensor" display.

As usual, I ran the ATITool 0.26 "Check for Artifacts" for over 2 hours with no errors detected. I also ran the three Nvidia 8800 Tech Demos (Smoke, Adrianne, and Froggy) to visually check for artifacts and to monitor the GPU temperature.

My wife happened to walk in the room while I was testing with the "Adrianne" demo. She innocently asked: "Who is she and what are you doing staring at the screen?"

In response, I nonchalantly said: "Just testing your new computer, my love."

Speaking of the "Adrianne" demo, I have heard a rumor that it contains an "Easter egg" that strips Adrianne down to her earrings, necklace and the "chopsticks" in her hair with the correct combination of "secret" keystrokes. Don't ask me for the secret combination because I do not know. I don't even know whether there is any truth to this rumor.

But, back to overclocking.

I ran the 3DMark06 benchmark three times at this new overclock setting for my BFG 8800GTS and I was pleasantly surprised to see Nighthawk score a high of 9031 3DMarks - an increase of 284 3DMarks over the previous overclock settings equivalent to a BFG 8800GTS OC model. Here is the 3DMark06 comparison.

My next objective is to raise the Memory Clock to 900 MHz with an intermediate stop at 875 MHz. Achieving this would match the specs of a BFG 8800GTX model. However, keep in mind that the 8800GTX has 128 Stream Processors versus 96 Stream Processors for the 8800GTS.

Overclocking the BFG 8800GTS

BFG Technologies' specifications for the BFG 8800GTS are: Core Clock of 500 MHz, Shader Clock of 1200 MHz, and Memory Clock of 1600 MHz. Everest Ultimate shows the actual specifications of my BFG 8800GTS as: Core Clock of 513 MHz (better), Shader Clock of 1188 MHz (worse), and Memory Clock of 1584 MHz (worse). I got a little more speed with the processor and a little less with the memory speed, but pretty close to factory specs.

BFG Technologies now also markets a BFG 8800GTS OC (Overclocked) version which increases the Core Clock to 550 MHz and the Shader Clock to 1300 MHz. These are my initial overclocking targets plus increasing the Memory Clock to its spec speed of 1600 MHz. I set the Core and Memory clocks to these values in ATITool 0.26 and started testing for stability, i.e., "Scan for Artifacts." Note that the Shader Clock is linked to the Core Clock. Also note that Everest Ultimate does not appear to show the same clock speeds as ATITool 0.26.

I allowed ATIT00l 0.26 to "scan for artifacts" for over 4 hours while closely monitoring the GPU temperature which is 54C @ idle. The image on the right is a screen capture at just over 2 hours of scanning for artifacts with no errors. Note that the GPU temperature is at 61C. The GPU temperature would spike up to 62C once in a while, but it never exceeded 62C. After over 4 hours of scanning for artifacts, ATITool 0.26 detected no errors.

Next, I ran the Nvidia 8800 demos (Smoke, Adrianne and Froggy) for an extended period to visually check for artifacts and to observe GPU temperatures. The Froggy demo exercised both the CPU (96-100% utilization) and the GPU. While running the Froggy demo, the GPU temperature peaked at 65C/149F, the CPU temperature peaked at 53C/127F, and the motherboard temperature peaked at 30C/86F. These temperature readings are all within acceptable range and I also did not observe any artifact problems.

How well does an overclocked Nvidia 8800GTS perform?

Quite well.

I fired off 3DMark06 and Nighthawk scored 8747 3Dmarks compared to the 8409 3Dmarks scored without GPU overclocking, an improvement of over 300 3Dmarks! Keep in mind that the 3DMark06 benchmark is primarily used for graphics performance so this result was to be expected. Here's the 3DMark06 comparison.

I might be able to hit the 9000 3Dmarks score yet.

I concluded my stability testing by flying between Caribbean islands in Flight Simulator X. I am happy to report that the scenery was perfect (no artifacts) and excellent. I also had one of my best landings ever.

My next overclocking target is to inch closer to the Nvidia 8800GTX specifications. I have already increased Core Clock to 575 MHz and the Memory Clock to 850 MHz. ATITool 0.26 has been scanning for artifacts since almost 2 hours ago (with no errors so far) as I write this post. I will report the full results tomorrow.

Your mileage may vary

I have 2 almost identical systems (Blackbird and Nighthawk), yet their overclocking potential have been different. Blackbird is stable at 3.4 GHz overclock, yet the maximum stable overclock for Nighthawk is 3.2 GHz.

Both systems have the same motherboard (ASUS P5B Deluxe WiFi-AP), same processor (Intel Core 2 Duo E6700 with stock HSF), same memory (4 GB of XMS2 TWIN2X2048-8500C5), same power supply (Antec NeoHE550), and the same case (Antec Nine Hundred) with the same number of fans (one 200mm and four 120mm).

The only differences are the graphics (Blackbird has the BFG 7950GT while Nighthawk has the BFG 8800GTS), the number and model of disk drives (Blackbird has six Seagate 750GB SATA-2 drives while Nighthawk has two Maxtor 500GB SATA-2 drives), the audio (Blackbird has the Soundblaster XFi Platinum while Nighthawk uses the onboard audio), and the number and models of optical drives (Blackbird has a Samsung SATA DVD-RW Lightscribe drive while Nighthawk has a Memorex IDE DVD-RW and a Plextor SATA DVD-RW.)

None of these differences should make a difference in their overclocking capabilities, but one never knows.

It is commonly accepted that 2 systems with the same exact components will overclock differently.

Do not feel bad or pressured to match (or exceed) what others with similarly configured systems are able to achieve with their overclocking. If you are one of these "others," don't gloat and just be grateful for your fortunes.


On to overclocking my BFG 8800GTS Graphics Card

I have initiated overclocking my BFG 8800GTS graphics card using ATITool 0.26.

You might be wondering why I am using ATITool 0.26 if I have an Nvidia 8800GTS graphics card. As it turns out, ATITool can be used for both ATI and Nvidia graphics cards. I selected ATITool because of its many capabilities, e.g., ability to find max core and memory speeds automatically, artifact checking, and the ability to save and load multiple profiles.

The Nvidia 8800 graphics cards have excellent overclocking potential similar to the Intel Core 2 Duo processors.

The first milestone I want to achieve with my BFG 8800GTS graphics card is to overclock it to the same GPU clock as the BFG 8800GTS OC graphics card: core clock from 500 MHz to 550 MHz and shader clock from 1200 MHz to 1300 MHz.

That's where the BFG 8800GTS graphics card is running at the moment while I use ATITool 0.26 to test for stability, i.e., no artifacts. ATITool 0.26 has been checking for artifacts at this new speed for almost one and a half hours as I write this post.

I will share the results of this overclock (including 3DMark06 benchmarks) with you in my next post tomorrow.

Memory errors at DDR2-960 MHz

I set the DRAM Frequency to DDR2-960 MHz in BIOS which equates to a 12:8 (DRAM:FSB) ratio with the CPU overclocked to 3.20 GHz (10x multiplier).

My Corsair XMS2 TWIN2X2048-8500C5 should be able to handle this easy. It is rated at PC2-8500 or DDR2-1066 MHz. According to its EPP (Enhanced Performance Profile) named "High Frequency" (Optimal Performance Profile), this memory can run at the memory speed of DDR2-1066 (memory clock speed of 533 MHz) at 2.2V with timings of 5-5-5-15 (CL-RCD-RP-RAS) with a Write Recovery Time (tWR) of 8T.

I ran Memtest86 v1.65 to test the new configuration and it failed with an error after 3 hours (or maybe sooner since I was not constantly monitoring the test.) I attempted to change the DRAM Write Recovery Time to 8 DRAM Clocks in BIOS.

Note that the EPP specified a Write Recovery Time (tWR) of 8T, but the ASUS 0804 BIOS allowed a maximum of 6T.

I had no other choice but to set the DRAM Write Recovery Time to 6 DRAM Clocks. I restarted Memtest86 v1.65 and it failed after 1 hour and 40 minutes with 2 errors.

I decided to loosen the memory timings to 5-5-5-18 (CL-RCD-RP-RAS) knowing that increasing the DRAM RAS# Activate to Precharge (RAS) setting to 18 DRAM Clocks would have very minimal impact on memory performance.

I ran Memtest86 v1.65 once more with the new memory timing and the test failed after 4 hours and 30 minutes with one error.

I believe that the only way I will be able to run this memory at DDR2-960 and up to DDR2-1066 is if the ASUS BIOS will allow DRAM Write Recovery Time (tWR) to be set to 8 DRAM Clocks as specified by the memory's Enhanced Performance Profile.

Given that the lastest version (0804) of the ASUS BIOS for the P5B Deluxe WiFi/AP motherboard does not allow this and maxes out at 6 DRAM clocks, I reset the DRAM Frequency to DDR2-800 MHz, memory timings to 5-5-5-15 (CL-RCD-RP-RAS), and DRAM Write Recovery Time (tWR) to 5 DRAM Clocks in BIOS. These were the last stable (error-free) settings for my memory.

I highly recommend that if you are running your memory at or close to its highest rated speed, check its stability by running Memtest86 overnight, specially if you have "Configure DRAM Timing by SPD" enabled to confirm that it is stable. If Memtest86 detects errors, disable SPD and set your memory timings manually according to your manufacturer's specifications.

I am perfectly content to run my Corsair memory at DDR2-800 MHz. This brings my attempt to optimize my memory settings to a close.

Next up will be overclocking my BFG 8800GTS video card.

Memories

Up to this point, I have focused and concentrated on getting the most out of my Intel Core 2 Duo E6700 processor using the stock HSF, except for a short detour into improving the performance of my disk drives by migrating from a non-RAID set-up to RAID 0 using the Intel ICH8R Controller and Intel Matrix Storage Console.

I am pleased with accomplishing a conservative 20% overclock of the C2D E6700 from 2.66 GHz to 3.20 GHz with a very stable system.

During this particular phase of my overclocking exercise, I intentionally kept the DRAM:FSB ratio at 1:1 to keep the memory component "out of the way" while I experimented with the CPU clock since both the CPU and memory clocks are tightly linked together. Changing the CPU clock has an immediate and direct effect on the memory clock. Keeping this ratio at 1:1 made overclocking to this point so much easier.

But is this 1:1 (DRAM:FSB) setting optimal?

Hiker, in a post in the ASUS Forums about his overclocking experience, makes a case that this may not be the optimal setting for high performance memory.

Let's find out in this next phase of my overclocking exercise. Now that I have completed getting the optimal performance from my CPU and stock HSF, I turn my focus on getting the optimal performance from my memory: 4 GB of Corsair XMS2 TWIN2X2048-8500C5 memory.

As usual, my first step was to set a baseline at my current 1:1 (DRAM:FSB) ratio using the Everest Ultimate Cache & Memory Benchmark. The baseline DDR2-640 measurements are: 6470 MB/s for Memory Read, 5832 MB/s for Memory Write, 5915 MB/s for Memory Copy and 71.3 ns memory latency.

I then changed the DRAM:FSB ratio in BIOS to 10:8 by setting the DRAM Frequency to DDR2-800 MHz. After restarting the system (with the usual automatic temporary power-off followed by a power-on which is unsettling to people who are not used to this normal Intel P965 chipset behavior), the Overclock view in Everest Ultimate confirmed the new 10:8 (DRAM:FSB) ratio with the memory clock now running at 400 MHz (or DDR2-800 MHz).

Everest Ultimate also has a "SPD" view which you can use to check the SPD (Serial Presence Detect) and EPP (Enhanced Performance Profile) settings for your memory (if it supports either or both) which are read directly from the memory sticks themselves. More importantly, you can use the information in this view to confirm that you have "matched" memory sticks, i.e., memory with the same specifications, which is important for DDR2 operation.

I used this information to verify that the memory timings in BIOS are already at the optimal settings for my Corsair memory.

So, were there any performance improvements with the memory at 10:8 (DRAM:FSB) ratio (DDR2-800 MHz)?

The Everest Ultimate Cache & Memory Benchmark showed that Memory Write stayed the same, but there was a dramatic increase in Memory Read performance and improved Memory Latency.

Memory Read performance went up from 6470 MB/s to 7491 MB/s and Memory Latency improved from 71.3 ns to 63.2 ns. The improvement in Memory Read performance directly contributed to the improvement in Memory Copy performance from 5915 MB/s to 6397 MB/s.

I also ran 3DMark06 and there was a slight improvement. You can view the comparison here.

Hiker from Sweden may be right after all.

I ran Memtest86 v1.65 overnight for over 8 hours and 38 minutes (wall time) executing 18 passes to confirm system stability with this new memory setting. No errors were detected.

The next step in my overclocking exercise is to set the Corsair memory to 12:8 (DRAM:FSB) ratio or DDR2-960 MHz. I will report on this later this coming week since I will be gone travelling on business.

Hitting the wall at 3.4 GHz

I ran Memtest86 v1.65 overnight for over 9 hours and 33 minutes (wall time) executing 19 passes without an error.

I restarted the system and allowed Nighthawk to settle in an idle state. The temperature readings at idle and 3.4 GHz overclock are: motherboard @ 30C/86F, CPU @ 49C/120F, Core #1 @ 55C/131F, Core #2 @ 56C/133F, and the GPU @ 53C/127F.

The first benchmark I ran was 3DMark06 and Nighthawk scored 8491 3DMarks, an improvement of 82 3DMarks over the 3.2 GHz benchmark. The graphics scores were similar but there was an improvement of 138 Marks in the CPU score.

The next benchmark I ran was the Everest Ultimate Cache & Memory Benchmark. There were improvements in all results, and the Memory Read result of 6856 MB/s now exceeds my highest score of 6781 MB/s achieved at no overclocking when memory was running at DDR2-1066.

I then ran 2 instances of SuperPi mod1.5 XS and took temperature readings at the end of the 20th loop of the 32M calculations. The temperatures were: motherboard @ 33C/91F, CPU @ 56C/133F, Cores #1 and #2 both @ 62C/144F, and GPU stayed at 53C/127F.

Both instances of SuperPi completed faster at 3.4 GHz versus 3.2 GHz overclock, completing 32M calculations 1 minute 14.031 seconds and 1 minute 14.312 seconds faster.



Everest Ultimate System Stability Test

My prior experience with the Everest Ultimate System Stability Test is that it really stresses the CPU more than other similar tools since I have observed it generate very high CPU temperature readings. This test keeps the CPU usage at 100% most of the time, with the usage dipping down to 50% when it switches tests.

The CPU temperature peaks at 72C/162F at 100% utilization, with both Cores #1 and #2 hitting 75C/167F.

During the periods when the Everest Ultimate System Stability Test is using the CPU at 50% while it is in transition between tests, the temperatures come down to more reasonable values.

The CPU temperature goes down to 57C/135F at 50% utilization, with both Cores #1 and #2 going down to 63C/145F.

At these very high temperature readings, I was not comfortable running this stability test longer than 10 minutes.

The good news is that all the temperatures are back at "idle" levels within 5 minutes of ending the stability test. The CPU temperature returns to 49C/120F, with Core #1 at 55C/131F and Core #2 at 56C/133F. The mother board and GPU temperatures stayed at 30C/86F and 53C/127F, respectively, throughout this stability test.

The Wall

I was feeling very good at this point so I fired up Johnny Lee's SP2004 ORTHOS as my final stability test with the intention of letting it run the rest of the day and overnight.

Nighthawk hit the wall just a little over 2 minutes into this stability test. Orthos stopped with the following messages: "FATAL ERROR: Rounding was 0.5, expected less than 0.4. Hardware failure detected, consult stress.txt file."

Windows XP MCE 2005 appeared to be running normally even after this error. I re-booted and tried a few more times before I gave up.

I lowered the overclock to 3.2 GHz and I have been running Orthos on Nighthawk since then, over 2 hours ago as I write this post. I can probably get over this hurdle by replacing the stock HSF and tweaking the voltages, but I am perfectly content at being able to overclock Nighthawk to 3.2 GHz, a respectable overclock.

Another one the rules I follow when overclocking is: Don't be greedy.