Benchmark Buffet: Updated 9/1/10

Today we will take a high-level look into the world of benchmarking: CPU benchmarks, benchmark tools and much more. Specifically, we will be looking at 12 point scoring benchmarks plus a new addition listed at hwbot.org as well as the more common utilities used for overclocking and submitting scores to their system. We also touch on system information and monitoring programs.

For those that have not heard of HWBot, it is the premier online repository for submitting PC benchmarks and as a hardware specification resource. They also feature “the world’s only non-stop overclocking competition platform known as the HWBoints rankings. 20,239 overclockers from all over the world compete in more than ten of the most popular benchmarks in thousands of hardware categories, resulting in over 340,000 benchmark results, a figure that is increasing by over 300 every day”.

“The benchmark results generate points awarded to both the member’s profile, the team’s profile and the country of origin and based on those points members are ranking in the Overclockers League, Hardware Masters League, Team League, and many more sub-rankings…”

Now you have a brief background on the most popular host for competitive benchmarking and where the world’s 17th ranked team (of over 1000!), the Overclockers.com Benchmark Team, hang their hats. Lets take a look at the benchmarks and utilities you will use to run them!

Utilities

Below is a list of the major applications used and a brief description. Their functions are anything from overclocking CPU’s and GPU’s to system status/monitoring applications needed to confirm and enter scores at HWBot, to temperature monitoring tools and programs. Of course, you don’t need ALL of these, but can choose which you like the best to use for your tweaking needs. Please be sure to verify compatible programs with your specific hardware.

CPUz (Current version 1.55)

This utility shows information about your AMD or Intel CPU, Mainboard, Memory, and System. It will be used in all of the benchmarks listed so it’s a must-have when submitting any scores.

CPUz Mainboard Tab.
CPUz CPU Tab v1.54.

CPUz Mainboard Tab.

Memory Tab.
Memory Tab.

SPD Tab.
SPD Tab.

GPU-z (Current version 0.4.5)

GPU-z is a program that shows information about your ATI or Nvidia GPU including but not limited to shader count, memory, core speed (both stock and overclocked), as well as temperature and voltage monitoring of the GPU. It will be used in 3D based screenshots and is also a must-have for submitting those scores.

Gpuz Main Screen.
Gpu-z Main Screen v4.2.

Sensors Tab.
Sensors Tab.

RivaTuner (Current version 2.24c)

RivaTuner is tweak utility designed to overclock Nvidia and ATI video cards. With this tool, one is able to set Core/Shader(Nvidia)/Memory speeds, adjust fan settings, enable/disable AA/AF, raise/lower detail settings, and many other features.

RivaTuner Main Screen.
RivaTuner Main Screen.

Details Screen.
Details Screen.

GPU Clockspeed Control.
GPU Clockspeed Control.

ATI Catalyst Control Center (Current Version 10.8)

CCC is for overclocking ATI video cards and is usually downloaded from the ATI driver package. Like RivaTuner, this video card tweaking utility you are able to set Core/Memory speeds, fan settings, enable/disable AA/AF, raise and lower detail settings along with many other features.

CCC Overclocking screen.
CCC Overclocking screen.

CCC Details screen.
CCC Details screen.

Evga Precision (Current version 1.9.6)

Precision is a small footprint, little frills video card overclocking program for Nvidia cards. It will allow you to adjust the Core/Shader/Memory clocks and Fan speed. It has a nice graphing function on it to see temperature, core/shader/memory speed, GPU use as well. You are also able to save overclocking profiles. You do not have to have an EVGA-based card to use Precision.

eVGA Precision v1.91.
eVGA Precision v1.9.1

MSI Afterburner (Current version 1.6.1)

MSI afterburner is a nice utility for both Nvida and ATI video cards. Like the other video card tweak programs, Afterburner allows one to adjust the core/shader/memory clocks and fan speeds, monitors clock speed/temperatures (min/max), fan speeds and frame rates. You can also save 5 profiles.

MSI Afterburner.
MSI Afterburner v 1.5.1

SetFSB (Current version 2.3.150.116)

SetFSB is a CPU overclocking utility that is used to raise the Front Side Bus (FSB) of Intel S775 (and more) chips. It is primarily used to get those last 100+Mhz out of your system as this adjusting is completed from within windows. You will have to locate the clock generator for your specific motherboard to properly use this Tool.

Set FSB.
Set FSB.

AMD Overdrive (Current version 3.2.3)

AMD:O is for overclocking AMD CPU’s. This utility allows for HTref adjustments (FSB/Bclk for you Intel folk!), voltage adjustments, built-in stability testing, as well as monitoring volts and temps.

AMD Overdrive Status screen.
AMD Overdrive Status screen.

Overclocking and Voltage adjustments.
Overclocking and Voltage adjustments.

MemSet (Current version 4.1)

MemSet is a windows based memory tweaking utility that will work for Intel DDR2/DDR3 without an integrated memory controller (IMC).  It shows the various common memory timings (and others) and allows for on-the-fly adjustments for your memory timings.

Memset v4.1
Memset v4.1

CPU-Tweaker (Current version 1.5)

CPU-Tweaker is a memory tweaking utility for the Intel i3/i5/i7 series. Like Memset, it shows the various common memory timings and a lot of others, as well as allowing you to adjust them on the fly.

CPUTweaker v1.4
CPUTweaker v1.4

RealTemp (Current Version 3.4) / Core Temp (Current version 0.99.7)

RealTemp is simply a temperature monitoring program for Intel chips. Core Temp is for AMD and Intel chips. You wouldn’t want to have these running while benchmarking, but they are needed when trying to find your highest clock speed on CPU and GPU.

RealTemp v3.4
RealTemp v3.4

CoreTemp v0.99.5
CoreTemp v0.99.5

A couple of things to note about some of these programs…

1. It is usually recommended to begin your overclocking adventure from the BIOS. The most common reason is for stability. You would generally want to use an application like SetFSB or motherboard specific programs like EasyTune6 to get the very last out of your overclock. Here are some local overclocking guides to get you started: C2D/C2Q, i3/i5/i7, AMD.

2. YOU DO NOT NEED TO USE ALL OF THESE APPLICATIONS. Do not be intimidated! Try them out and see which application you prefer to use!

2D Benchmarks

2D Benchmarks are programs that do not test the video card but test the CPU and subsystems.

CPUz (See above for DL link)

This benchmark program is simply used to capture a validation and screenshot of the clock speeds. In order for this score to be submitted to the ‘bot, you must validate the score and submit the link you received from your validation to Hwbot. Any online CPU-Z submission must comply to the following rules: use most recent CPU-Z version, have a valid canardpc validation link as verification, screenshots are no longer allowed since 19th May 2008. You can save your overclock and submit later if your bench rig is not connected to the internet. use the same nickname in the CPU-Z submission as your hwbot name. Example CPU-Z verification link: http://valid.canardpc.com/show_oc.php?id=218928.

CPUz Validation screenshot (link in browser window)
CPUz Validation screenshot (link in browser window)

Tips/Tricks – The goal here is reach and validate the highest speed your CPU can go. There is no such thing as stable here. So long as you get the validation link saved and can submit it to the bot properly, thats all that matters.

Super Pi 1M/32M (HWBot accepted version, 1.5 XS mod)

SuperPi is a program that calculates a specific number of digits of Pi. In this case we use 1 million and 32 million digits. It is commonly used by overclockers to test the stability and speed of an overclocked system. You must use the XtremeSystems version 1.5, having accuracy in milliseconds and checksum validation choose 1M/32M calculation, and have a valid screenshot (see example below): clearly show SuperPi time and “calculation done” popup, processor in CPUz, memory settings in another CPUz instance, and checksum.

Valid Super Pi Screenshot.
Valid Super Pi Screenshot.

Tips/Tricks – This application (like most benchmarks) is best run with nothing else going on in your system. This includes eliminating any unneeded active services as well as active processes. 32M times are influenced by memory timings and FSB/uncore/HyperTransport speed more so than 1M. XP is the OS of choice for this bench.

WPrime 32M/1024M (HWbot accepted version 1.55)

WPrime is a multi-threaded benchmark application that can quickly test your processor performance. It does so by calculating square roots with a recursive call of Newton’s method for estimating functions. In contrast with most other simple benchmark applications, wPrime is written to take full benefit of processors with multiple cores. In order to submit a valid score, you must use the wPrime 1.55 version, choose 32M/1024M calculation, and have a valid screenshot (see example below): clearly show wPrime time, processor in CPUz and memory settings in another CPUz instance.

Valid WPrime Screenshot.
Valid WPrime Screenshot.

Tips/Tricks – This application of course likes nothing else running with it on your rig again eliminating any unneeded active services you can as well as active processes. Windows XP 32bit / Vista 32 bit are the better OSes to complete this benchmark.

Hexus PiFast (Hwbot accepted version 4.1)

PiFast is used to compute pi with a very large number of digits. Any online PiFast submission must comply to the following rules: use Hexus PiFast version (4.1) and have a valid screenshot (see example below): clearly show time, DOS box, processor and memory tabs in CPUz.

PiFast Screenshot (From Hwbot.org)
PiFast Screenshot (From Hwbot.org)

Tips/Tricks – Again like all other benches, trimming your uneeded services and active processes will help in getting the best score.

3D Benchmarks

3D benchmarks are programs that are supposed to test the GPU. However many of them are becoming a pure CPU race these days because graphics cards have become so powerful.

3DMark 2001 SE (Current version 3.3.0)

3DMark 2001 is an old 3D benchmark application that measures videocard performance in DX7 and DX8. It has little to no relevance for current game performance. However it is a lot of professional benchers’ ‘favorite game’ to play.  In order to submit a valid score, one must have a valid screenshot (see example below): clearly show 3Dmark score, 3Dmark subtest scores, 3Dmark settings, processor in CPUz, video card in GPU-z, unless you provide a futuremark orb link. GPU-z in screenshot, drop-down must be visible if you run in SLI so all video cards are identified. Rivatuner may be used if GPU-z fails to identify video card properly (as of 8 oct 2007). You must have a valid futuremark orb link as verification for scores in the top 20.

3dMark 2001 Screenshot (From Hwbot)
3dMark 2001 Screenshot (From Hwbot)

Tips/Tricks – This application, being so old and run on a low resolution, loves CPU speed along with a video card overclock. So, be sure to push that as far as you can for best results. As is typical with 3D benching, make sure your card’s settings are at performance and not high quality. AA and AF should be application controlled or off. You are also able to run the individual tests that make up the benchmark in any order and as many times as you want. This is beneficial as some tests are harder on the GPU than others so you can get away with higher overclocks for some of the tests yielding a better score. The best OS for this is XP. Driver choice will play a role in scores as well.

3DMark 2003 (Current version 3.6.0)

3Dmark 2003 is another older video card benchmarking application. It measure performance in DX8 and DX9. Any online 3DMark03 submission must comply to the following rules: use default 3Dmark settings, have a valid screenshot (see example below): clearly show 3Dmark score, 3Dmark subtest scores, 3Dmark settings, processor in CPUz, video card in GPU-z, unless if you provide a futuremark orb link. GPU-z in screenshot, drop-down must be visible if you run in SLI so all video cards are identified. Rivatuner may be used if GPU-z fails to identify video card properly (as of 8 oct 2007). You will need to have a valid futuremark orb link as verification for scores in the top 20.

Valid 3DMark 2003 Screenshot (From Hwbot).
Valid 3DMark 2003 Screenshot (From Hwbot).

Tips/Tricks – Another 3D bench that really gets a boost from faster clock speeds. As is typical with 3D benching, make sure your card’s settings are at performance and not high quality. AA and AF should be application controlled or off. The best OS for 3DMark 2003 is Windows 7 32bit. Driver choice will play a role in scores as well.

3DMark 2005 (Current version 1.3.0)

3DMark 2005 is a slightly older 3D benchmark that uses DX9. It is only a DX9 benchmark so older cards that do not support that API cannot be used in this bench. 3DMark05 also shows significant improvements with CPU clock speed as well as overclocking the video card. Any online 3DMark05 submission must comply to the following rules: use default 3Dmark settings, have a valid screenshot (see example below): clearly show 3Dmark score, 3Dmark subtest scores, 3Dmark settings, processor in CPUz, video card in GPU-z, unless if you provide a futuremark orb link. GPU-z in screenshot, drop-down must be visible if you run in SLI so all video cards are identified. Rivatuner may be used if GPU-z fails to identify video card properly (as of 8 oct 2007). You must have a valid futuremark orb link as verification for scores in the top 20.

Valid 3DMark05 Screenshot.
Valid 3DMark05 Screenshot.

Tips/Tricks – Another 3D bench that really gets a boost from faster clock speeds. As is typical with 3D benching, make sure your card’s settings are at performance and not high quality. AA and AF should be application controlled or off. The a good OS for 3DMark 2005 is Windows Vista 32. Driver choice will play a role in scores as well.

3DMark 2006 (Current version 1.2.0)

3DMark 2006 is the 2nd newest version of Futuremark’s 3D benching applications. This version uses DX9 and Shader Model 2.0 as well as 3.0. Any cards without DX9 and SM3 support will not work with this benchmark. It consists of two SM2.0 graphics tests, using DirectX 9 SM2.0 features extensively, two HDR/SM3.0 graphics tests, using DirectX 9 SM3.0 features and floating point extensively, and finally, two multiprocessor CPU tests focusing on CPU intensive tasks in 3D games. At the end of the benchmark it will give a score for each test and a total. Any online 3DMark06 submission must comply to the following rules: use default 3Dmark settings, have a valid screenshot (see example below): clearly show 3Dmark score, 3Dmark subtest scores, 3Dmark settings, processor in CPUz, video card in GPU-z, unless if you provide a futuremark orb link. GPU-z in screenshot, drop-down must be visible if you run in SLI so all video cards are identified. Rivatuner may be used if GPU-z fails to identify video card properly (as of 8 Oct 2007). You must have a valid futuremark orb link as verification for scores in the top 20.

Valid 3DMark 2006 Screenshot (From Hwbot).
Valid 3DMark 2006 Screenshot (From Hwbot).

Tips/Tricks – Another 3D bench that really gets a boost from faster clock speeds AND the number of cores used. The more cores used, the higher your CPU score. As is typical with 3D benching, make sure your card’s settings are at performance and not high quality. AA and AF should be application controlled or off. The a good OS for 3DMark 2006 is Windows Vista 32. Driver choice will play a role in scores as well.

3DMark Vantage (Current version 1.0.2)

3Dmark Vantage is the most recent release by futuremark of their 3D benchmarking series. This version uses DX10 and Shader Model 3.0. Since it uses DX10, it will require a video card that supports it as well as an OS (Vista or Windows 7) that supports it as well. There are a couple settings on here and the standard at Hwbot is the Performance bench. If you run an Nvidia GPU you must disable PhysX for a valid score. Any online 3DMark Vantage – Performance submission must comply to the following rules: use Performance settings, have a valid screenshot (or valid futuremark orb link): clearly show 3Dmark score, 3Dmark subtest scores, 3Dmark settings, processor in CPUz, video card in GPU-z, unless if you provide a futuremark orb link, GPU-z in screenshot, drop-down must be visible if you run in SLI so all video cards are identified. Rivatuner may be used if GPU-z fails to identify video card properly (as of 8 oct 2007). You must have a valid futuremark orb link as verification for scores in the top 20.

3DMark Vantage Screenshot (from HWBot)

Tips/Tricks – While pouring on the clock speed and cores has its advantages here as well, it’s not as pronounced as the others especially when you get into the higher presets. As is typical with 3D benching, make sure your card’s settings are at performance and not high quality. AA and AF should be application controlled or off. The a good OS for 3DMark Vantage is Windows Vista 32. Driver choice will play a role in scores as well.

Unigine Heaven (DX11/Extreme Preset / DX9/Basic Preset):

The newest benchmark to grace Hwbots list is an application called Unigine Heaven. This is a popular benchmark among enthusiasts as it uses multiple API’s (DX9 and DX11 and OGL4.0 in this case) multi monitor support and some of the latest eye candy features like tesselation. This is how Hwbot.org explains the benchmark: The HWBOT Unigine Heaven Benchmark application consists of the Unigine Heaven benchmark and a wrapper software tool developed by HWBOT. The benchmark comes in two shapes: a DirectX9 (Basic) preset, allowing people to test and compare low-end and mainstream setups, and a high-end DirectX11 (Xtreme) preset to test and compare the latest graphics cards. Next to the two presets, the wrapper software allows users to submit scores directly to the HWBOT database or, in case the test system has no internet connection, the possibility to save your result.” Here are the rules from Hwbot for this benchmark.

Any online HWBOT Unigine Heaven submission must comply to the following rules: Use standard benchmark settings. Verification file is mandatory for submitting results. Have a valid screenshot: clearly show benchmark score, processor and memory information in CPU-Z, videocard in GPU-Z, GPU-Z in screenshot, dropdown must be visible if you run in SLI so all videocards are identified. Rivatuner may be used if GPU-Z fails to identify videocard properly (as of 8 oct 2007).

You can download the version used at Hwbot.org directly from their site (linked above),  I do not believe the version on Unigine’s site will be accepted at Hwbot. At this point in time, this is a non boint scoring application.

Unigine Heaven v1.0.2 (Courtesy of Hwbot.org)
Valid Unigine Heaven Screenshot (From Hwbot.org)

Tips/Tricks – I personally dont know any at this time, but like is typical with a lot of 3D benchmarks, this thing loves powerful GPU’s and CPU clockspeeds. Crank them both for best results.

2D/3D

A combination bench such as PCMark 2005 and PCMark Vantage have an assortment of tests from the CPU and subsystem to video card performance.

PCMark 2005 (Current Version 1.2.0)

PCMark 2005 is a dated, but still relevant release of a complete system benchmark. This tests CPU, GPU, Memory, Web page rendering, and Hard Drives. Having all these items tweaked are going to yield the best results. Any online PCMark05 submission must comply to the following rules: use default PCMark settings, have a valid screenshot (see example below): clearly show PCmark score, PCmark settings, processor in CPUz, GPU-z, Subtest Scores (XP startup speed!), unless you have a valid futuremark ORB url. You must have a valid futuremark orb link as verification for scores in the top 20. XP Startup cannot exceed 220MB/s.

Valid PCMark 2005 Screenshot (From Hwbot).
Valid PCMark 2005 Screenshot (From Hwbot).

Tips/Tricks – Since this is a complete system benchmark and all relevant items (CPU/GPU/Memory) need to be overclocked and tweaked where applicable in order to achieve the highest score. Obviously drives  in RAID 0 will score higher than a single drive of the same make/model as well as SSD’s generally scoring higher than most 2/3 drive RAID 0 setups. The faster the memory runs and tighter the timings, the better score you will get also.

Now that you have seen all the benchmarks, and have all the tools you need to be successful at it,  it’s time to get down and chilly and get overclocking! Also note that there are a lot more tricks the benchmarking team have up their sleeves with years of combined experience that will help boost those scores up! Join the fray by heading into the forums to the Overclockers.com Benchmarking Team’s sign up thread and completing 10 of the 13 benchmarks above.

* Note all rules were copied directly from the Hwbot.org site and some screenshot examples were as well.

EarthDog

About Joe Shields 326 Articles
Joe started writing around 2010 for Overclockers.com covering the latest news and reviews that include video cards, motherboards, storage and processors. In 2018, he went ‘pro’ writing for Anandtech.com covering news and motherboards. Eventually, he landed at Tom’s Hardware where he wrote news, covered graphic card reviews, and currently writes motherboard reviews. If you can’t find him benchmarking and gathering data, Joe can be found working on his website (Overclockers.com), supporting his two kids in athletics, hanging out with his wife catching up on Game of Thrones, watching sports (Go Browns/Guardians/Cavs/Buckeyes!), or playing PUBG on PC.

Loading new replies...

b
beenthere

Member

515 messages 0 likes

I noticed in several utilities such as Core Temp, etc. that you mentioned they should not be running when benchmarking. Is this to insure the best scores or will some of these utilities cause system crashes when running intensive benchmarks like Prime 95, etc. ?

Reply Like

Avatar of hokiealumnus
hokiealumnus

1

16,560 messages 25 likes

It's to ensure the best score. You want all resources you can get dedicated to the benchmark that's running. Running those will take away a few CPU cycles.

You mention Prime95 - that's a stability tester and not a benchmark per se'. You want to run CoreTemp (or other program) while that's running so you can monitor your temperatures. Prime and the like are to figure out if your 24/7 overclock is stable (and is within your thermal envelope). For benching, we only care if it's stable enough to complete them. :)

Reply Like

b
beenthere

Member

515 messages 0 likes

OK thanks. Yeah I should have used Futuremark as a benching reference not Prime 95. The reason I specifically mentioned Core Temp is that their "readme" file implies it may cause system crashes because it's hardware based. I just didn't know if you were seeing this in operation or not?

So if my PC is OC to 10 GHz. and will finish a benchmark, I'm the new King, right? :confused:

Reply Like

Avatar of hokiealumnus
hokiealumnus

1

16,560 messages 25 likes

OK thanks. Yeah I should have used Futuremark as a benching reference not Prime 95. The reason I specifically mentioned Core Temp is that their "readme" file implies it may cause system crashes because it's hardware based. I just didn't know if you were seeing this in operation or not?

I haven't experienced any instability solely because CoreTemp is running on anything from the AMD or Intel side, ever. I've used two generations of Core2Duo and several Skt1156 chips on the Intel side and from the AMD side, most iterations of chips from an Athlon x2 on through to a Thuban.

So if my PC is OC to 10 GHz. and will finish a benchmark, I'm the new King, right? :confused:

As long as you get the proper validation screenshot, yep! Heck, with CPUz, all you need to do is save the validation file (current world record is 8199.5MHz).

Reply Like

click to expand...
b
beenthere

Member

515 messages 0 likes

10 GHz. should be a piece of cake. ;)

Reply Like

Avatar of EarthDog
EarthDog

1

76,189 messages 2,916 likes

Riiiiiiiiiight. :p :)

Reply Like

b
beenthere

Member

515 messages 0 likes

Oh come on now... be an optimist! :confused:

InHell stated back in the PIII days that we'd all be using 6 GHz. flame-throwing PIII InHell CPUs within a year. I bought two fire extinguishers just for use with PIIIs.

Certainly ten years later 10 GHz. should be no problem from a CPU??????? :thup:

Reply Like

S
SecrtAgentMan

Member

554 messages 0 likes

Very useful, thanks

Reply Like

Avatar of ChanceCoats123
ChanceCoats123

1

9,572 messages 0 likes

Very nice write up ED!

Reply Like

Avatar of jiccman1965
jiccman1965

Member

1,349 messages 0 likes

Thank you great help

Reply Like