CPU reviews Windows 7 transition - opinions

[Omega]

Hi guys!

I'm gonna need a little help from you on this one. Many of you are probably already aware of CPU reviews being published recently on TPU. I really hope, and we are working hard to continue doing reviews and even expand and upgrade them to bring you the best readers experience possible.

So far, two reviews were published and third is on the way. All three tested CPU's are AMD's lower end products and they have a secondary function… to help me go trough TPU reviewer learning curve. After the third CPU is reviewed and published, the test setup will be transferred on Windows 7 OS, and here's where I need your help.

When all things considered, reviews are made and published for our readers (you) and our goal is to deliver to you all the information's and specifications you need, but keep everything easy and interesting to read. First of all, I need opinions whether to go with 32bit or 64bit version of Windows 7. The only difference here is that 32bit version enables higher stable overclock frequency's for CPU. In both cases amount of memory will be software limited to 3.2 GB for 32bit or 4 GB for 64bit version. This is a must when it comes to testing triple channel memory. Three sticks of ram (6 GB) will be used for LGA1336 Core i7 CPU's but amount of available memory limited to 4 GB, to make results comparable to LGA1156 and AM3 platforms.

So, long story short… would you prefer higher overclocked CPU's or ~800 MB larger system memory?

Second thing is the current benchmark suite used to test the CPU's. While doing reviews I noticed some flaws that could be improved with transferring to Windows 7. Bellow is the current benchmark suite:

Benchmark Suite​

• Everest Ultimate - Read/Write/Copy/Latency
• PC Mark Vantage
• 3D Mark Vantage v1.0.1 Performance preset (CPU and Total score)
• wPrime 32M v2.0
• SuperPI

• Handbrake DVD rip to .mkv
• Xilisoft Video Converter Ultimate v5.1.20 - .mov 1080p encode to .mpeg-4
• DivX Converter 7.0 - AVI encode with Home Theater preset

• BonkEncoder v1.0.13 .flac encode with LAME MP3 320 kbs
• I'm Too Audio Encoder v2.1.77 - WMA encode with LAME MP3 320 kbs

• Photoshop CS4
• Cinebench R10
• Blender 2.49
• POV Ray 3.7b

• 7zip v4.65 32 MB mutithreding
• WinRAR 3.9 Benchmark
• WinRAR 3.9 compression - 688 MB file

• Tom Clancy's HAWX v1.1 - high details
• Resident Evil 5 benchmark - medium details
• Section 8 v1.0 - high details
• GTR Evolution v1.0 - high details, Level 1 AA
• Far Cry 2 v1.3 - DirectX 10, High Details

• Prime95 for maximum heat and consumption
• CPU-Z 1.52
• AMD Overdrive

To be honest I'm not happy with audio encoding tests as they use only one core and are very frequency dependent . Also I'd like to change list of games for testing CPU performance in gaming. The maximum number is 5 games, and they should cover FPS, RTS, Simulations, Adventure/RPG platforms.

Again, to make the long story short. Do you have any proposition what other test to incorporate into benchmark suite? I would appreciate suggestions for audio encoding programs that use more than one core, video encoding programs/tests and game titles for gaming performance testing.

Thank you in advance for any given advice, cheers!

 

[newtekie1]

I'd prefer the 64-bit.

 

[FilipM]

I'd go with 64bit Windows 7, but i didnt understand the bit "32 bit os allows for higher oc" - is there that much of a difference?

In terms of benchmarks, Arma 2 is a really good CPU benchmark with a decent graphics card in the system, you could also try Brothers in Arms, hell's highway and Men of War/COH for RTS titles.

 

[Omega]

I'd go with 64bit Windows 7, but i didnt understand the bit "32 bit os allows for higher oc" - is there that much of a difference?

64bit Windows loads more instructions on CPU. Max overclocking potential downgrades anywhere from 100 MHz to 600 MHz, depending on the cpu.

As for games, only titles that have "in-game" benchmark or "replay mode" can be taken in consideration.

 

[Kursah]

I would rather see more real-world stuff done on x64. Especially since that's the route things appear to be going, personally I've not lost on my OC's going from x86 to x64, but then I'm not pushing for every mhz I can get either. But, I'd rather have 800 more MB of ram, than a small handful of extra mhz's from the CPU any day. Really I don't think losing 100-600mhz when your OC is already healthy (at least on the intel side) is that bad of a deal...though I can't say I know anyone personally that has had an OC loss going to x64..so maybe say 0MHz-600MHz potential loss instead.

 

[MK4512]

Definitely 64 bit.

It creates a much realistic overclock, as most people will be using the 64 bit edition of Windows 7.

 

[Omega]

..so maybe say 0MHz-600MHz potential loss instead.

You've got a point there. I also made a change from 32bit XP to 64bit W7 and my OC stayed the same. But it all depends on the processor and levels of overclock. These things will become clear fast enough when actual processor testing on 64bit begins. So far, most of the sites and overclockers use 32bit OS.

For now, let's keep it @ 0-600 MHz. 600 because I've witnessed that number on my friends CPU when he made the transition to W7 64bit from Vista 32bit.

 

[Velvet Wafer]

i experienced no problems so far, besides the usual problem,
that i cant cross the 4 ghz in 64bit,with water

 

[Omega]

Ok.... let's try this way.

1. Which applications do you use for video encoding/converting, and which format do you like the best (mkv, avi, mpeg4...)?

2. Which application do yo use for audio encoding/converting, and which format do you like the best (mp3, flac, wma...)?

3. Name 5 newer games you liked the most

4. Any other CPU demanding application you use not covered in benchsuite?

 

[W1zzard]

my vote goes for 32 bit so you can also compare processors without 64-bit support.

for the benchmarks you have there i see no point going 64-bit. the memory working set of all benchmarks is also small enough to fit into less than 2 gb of memory. 3.2 or 4 gb ram - no difference

 

[Omega]

my vote goes for 32 bit so you can also compare processors without 64-bit support.

I can't think of any new CPU with no 64bit support, and i see no point testing processors before first AMD's K8 Athlon or Pentium D series.
Also, I think 64bit native processors should be tested in their native environment - 64bit OS.

Going back to 32bit would also deny me the opportunity to test some 64bit applications (like WinRAR), and see how the processors handle 64bit apps.

 

[vagxtr]

But, I'd rather have 800 more MB of ram, than a small handful of extra mhz's from the CPU any day. Really I don't think losing 100-600mhz when your OC is already healthy (at least on the intel side) is that bad of a deal

I'd agree with you that more ram matter much more though we won't see this in above mentioned test cause Omega will limit it to 4GB only. btw. Omega how do you mean limit triple channel systems to 4GB only?
But as for OC potential goes, you cant compare totally different architectures in terms how they scale in 64b and 32b mode. Especially not K8/K8L vs. Core2 cause AMD shows it's full x86-64 potential while C2 hasn't that in itself just 64b compatibility Maybe some of test should compare that 32b vs. 64b scaling of that two-three competing old chips more directly in future reviews, and Nehalem on top of them. Even direct comparing 32b vs. 64b on C2D vs. Nehalem ... per core basis if it will be possible to disable affinity to some core on particular test

 

[mlee49]

First off thank you for the opportunity to provide feedback and even help offer suggestions on the review process.

I suggest using a 64 bit OS primarily of the fact that most people who actually read the reviews use a 64 bit OS. I personally think it gives more accurate usage of the product. That and with 64x on the rise with preproduction machines its gaining ground more rapidly.

 

[Omega]

Omega how do you mean limit triple channel systems to 4GB only?

Start>Run>msconfig>boot>advanced>max. memory = 4 GB

I need some help choosing the games for future CPU tests. Maximum number of titles is 5 and should cover most of the game genres.

1. Tom Clancy's HAWX

2. DiRT 2

3. GTR Race ON

4. Warhammer: Dawn of War II

5. BattleForge

6. FarCry 2

7. Crysis Warhead

8. Modern Warfare 2

9. Resident Evil 5

CPU gaming performance will be checked with Radeon HD 5850 in three resoulutions: 1024x768, 1680x1050 and 1920x1200.
Opinions?

 

[FilipM]

Dirt 2, FarCry 2, GTR Evo/Race, Crysis Warhead. Cannot choose from any of the others, never played them

 

[Deleted member 3]

my vote goes for 32 bit so you can also compare processors without 64-bit support.

for the benchmarks you have there i see no point going 64-bit. the memory working set of all benchmarks is also small enough to fit into less than 2 gb of memory. 3.2 or 4 gb ram - no difference

Ignoring new technologies (64 bit isn't even new anymore) may give false results though. Isn't it relevant how new technologies affect performance?
Will you disable any SSE4 support and the likes just so you can compare to an Atom and whatever other CPU's don't support such extensions? Will you run everything single threaded just so it's fair for some single core Atom/Celeron?
I think it's fine to run some tests on a level field, but I think you should at least run 1 or 2 benchmarks to focus on the added values of new features. Do new extensions increase performance? Do extra cores make a difference?

Also, I think running too many games has no added value. Run 2-3 and you're done, run encoding tests and other CPU intensive tasks.

 

[Omega]

I would beg to differ... RTS and FPS games have different approach on how to use CPU power, so while AMD can chunch some impressive numbers in FPS (example not a fact), it gets beaten by Intel in FPS. I think gaming community is much larger than "A/V enconding" one, and they deserve to know what CPU is best for their favorite game genres.

And since there are many genres, I have scaled the number of game tests to five: FPS, RTS, Simulation, Adventure, Arcade/other

 

[Omega]

How's this:

1. Crysis Warhead
Crysis and FarCry are based on same engine, so no use having both games as the performance ratios would be pretty much the same between cpu's.

2. Modern Warfare 2
Most recent and very popular FPS title

3. Warhammer II
Representing RTS genre

4. DiRT 2
DX11 game representing arcade/sim driving genre

5. GTR Race ON
Older based graphic engine, but possibly the only game that scales almost 100% with more CPU cores

5th place is still open for suggestions.

Here's the final benchmark setlist for CPU reviews:

Benchmark Suite​

• Everest Ultimate - Read/Write/Copy/Latency
• PC Mark Vantage
• 3D Mark Vantage v1.0.1 Performance preset (CPU and Total score)
• wPrime 1024M v2.0
• SuperPI 1M

• x264 benchmark V3.0
• Handbrake DVD rip to .mkv (Standard High preset)
• Xilisoft Video Converter Ultimate - .mov 1080p encode to .mpeg-4
• DivX Converter 7.2 - AVI encode with Home Theater preset

• dbPowerAmp .flac to 320 kbs MP3 Lame
• BonkEncoder v1.0.13 .flac encode with LAME MP3 320 kbs
• I'm Too Audio Encoder v2.1.77 - WMA encode with LAME MP3 320 kbs

• Photoshop CS4
• Cinebench R10
• Blender 2.49
• POV Ray 3.7b

• 7zip 32 MB mutithreding
• WinRAR Benchmark
• WinRAR real life compression - 688 MB file

• Crysis Warhead
• Modern Warfare 2
• Warhammer II
• DiRT2
• OPEN FOR SUGGESTIONS

• Prime95/Orhos for maximum heat and consumption

 

[Chad Boga]

Benchmark Suite​

• Everest Ultimate - Read/Write/Copy/Latency
• PC Mark Vantage
• 3D Mark Vantage v1.0.1 Performance preset (CPU and Total score)
• wPrime 1024M v2.0
• SuperPI 1M

• x264 benchmark V3.0
• Handbrake DVD rip to .mkv (Standard High preset)
• Xilisoft Video Converter Ultimate - .mov 1080p encode to .mpeg-4
• DivX Converter 7.2 - AVI encode with Home Theater preset

• dbPowerAmp .flac to 320 kbs MP3 Lame
• BonkEncoder v1.0.13 .flac encode with LAME MP3 320 kbs
• I'm Too Audio Encoder v2.1.77 - WMA encode with LAME MP3 320 kbs

• Photoshop CS4
• Cinebench R10
• Blender 2.49
• POV Ray 3.7b

• 7zip 32 MB mutithreding
• WinRAR Benchmark
• WinRAR real life compression - 688 MB file

• Prime95/Orhos for maximum heat and consumption

How many of the above are free downloads?

I am putting together a system for myself and after I have done that, I would like to see how well it is running.

 

[Omega]

Benchmark Suite​

• wPrime 1024M v2.0
• SuperPI 1M

• x264 benchmark V3.0
• Handbrake DVD rip to .mkv (Standard High preset)

• dbPowerAmp .flac to 320 kbs MP3 Lame
• BonkEncoder v1.0.13 .flac encode with LAME MP3 320 kbs

• Cinebench R10
• Blender 2.49
• POV Ray 3.7b

• 7zip 32 MB mutithreding
• WinRAR Benchmark
• WinRAR real life compression - 688 MB file

• Prime95/Orhos for maximum heat and consumption