i4memory.com - different look at memory - Announcements in Forum : AMD motherboards / CPU

eva2000 · 11-06-2010 until 12-12-2010

If you use NOScript or other ad blocking scripts, you may experience issues viewing or using i4memory.com. To fix please read note here.

eva2000 · 23-03-2010 until 23-12-2010

i4memory.com ver 3.0 site and forum design has been rolled out. Full details here

eva2000 · 10-01-2010 until 10-12-2010

New feature added to forums - read here

eva2000 · 11-09-2008 until 31-12-2012

Just a reminder that new bbcodes have been added so make use of them for your screenshots when needed > [feedback] Additional custom IMG Tags

eva2000 · 08-09-2008 until 31-12-2012

I notice folks posting very large screenies including their whole desktop in background. It's alot easier to reduce and crop to only the area of a screen capture you want to show.

I thought i'd show you how much of a reduction you can get in size of the final screen shot.

Download www.irfanview.com and plugins...

prnt-scr -> open irfanview -> ctrl-v to paste image -> use mouse to highlight only area you want to show -> ctrl-Y to crop that area -> ctrl-R to resize it to 450-800pixels for forums -> save as -> jpeg -> options set at 65-75% quality

I used 65% jpeg quality on all below screen shots.

Full Desktop:
* prnt scr -> paste into irfanview -> jpeg 65% quality -> save as = 191kb

* print scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed INCLUDES taskbar/clock -> save as = 182kb

* print scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed crop away taskbar/clock -> save as = 160kb

* print scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed crop away taskbar/clock -> resized smaller to legible font size (CTRL+R) -> save as = 108kb

3x CPUZ screenies + A64Tweaker:

* prnt scr -> paste into irfanview -> jpeg 65% quality -> save as = 124kb

* prnt scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed showing -> save as = 111kb

* prnt scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed showing -> resized (CTRL+R) -> save as = 72kb

For resizing (CTRL+R) ensure preserve aspect ratio is checked:

Additional Tips:
1. You can also batch resize folders of pictures all at once. Read Irfanview help file to find out how

2. If you save a screenshot which is unedited, without resizing as PNG file, you will get a smaller size than any JPEG compressed quality picture. However, if you resize a screenshot then save as PNG file it will result in a size larger than JPEG compressed quality picture. For example below picture thumbnail (44.79KB) is resized, cropped JPEG 60% quality which links to an non-resized original screen capture saved as PNG (66.39KB)

Update: September 21, 2006
Discovered you can save a further 40-50% reduction in screenshot size if you use Irfanview to convert to greyscale BEFORE you save an non-resized screen capture to PNG format with PNGOUT enabled at compression level 9.

These screenshots with PNG format at compression level 9 were around 36-60KB in size but now check out the image sizes!

Updated: March 15, 2007
More optimisation tweaks for smaller screen capture file sizes from SLI_dog

By SLi_dog

This is a how to on creating small file size screen captures using irfanview which is a truely great app thats absolutely free

First up you'll want to grab your self a copy of Irfanview and the plug in pack too.

www.irfanview.com

I've found that the smallest size screen captures are almost always in .png format. Irfanview has a great plugin called "png out" which really helps reduce file size even further.

Here are some examples. The following pictures are of the full desktop captured using the "CTRL+PrtScn" keys.

The first is just a normal save with irfanview in png using compression value 9. Notice that the image quality is as good as you'd see it on the screen.

File size = 122kb

This is the exact same save in .jpg format with the default jpg compression value of 80%. Notice that the image quality has been lost from that of the normal desktop. The file size has gotten bigger too

File size = 147kb

Now I'll just do a simple image crop by selecting the area I want to crop to by holding the left mouse button and dragging a box around the area then pressing "CTRL+Y" to crop. You can always resize the area after by dragging the edges of the crop selection box. This drops the file size to only 106kb, that's already a 1/6th size saving which means faster loading web pages for easier browsing.

File size = 106kb

At this stage I'll introduce the pngout plugin. The following is using the pngout default settings. Notice that the image quality is still as good as you'd get on the screen.

File size = 85kb

Now the real trick to reducing file size is to reduce the number of colours in the picture. You do get some decrease in picture quality but it's still very readable and the file size saving are huge. Here's how you do it.

Make sure you UNTICK the Use Floyd-Steinberg dithering or you'll end up with larger file sizes

The following are advanced settings for the pngout plugin that I've found give the best reduction most of the time. You want to change the "Compression level" to 9, change the "Block split threshold" to 4096, and click on the "Keep chunks" button and hit "Select none"

And if you put together the Reduce Colour Depth and Advanced png out settings you can get a file size reduction like this without compromising image quality too much.

File size = 32kb

I should point out that this will only work for simple screen captures, if you have normal photos or have a photgraphic windows backdrop then you may want to use .jpg

I think that having a 3/4th reduction is size is worth not having such a pretty backdrop while benchmarking

Don't just take my word for it, play around and see if you can find any better settings. If you do I'd be keen to know.

Update: September 8, 2008

Just an update, if you find irfanview PNGout method too slow and don't mind sacrificing some compression for speed, check out PNGoptimizer as well as other tools listed at http://forums.legitreviews.com/about16702.html.

Hint:
1. only crop don't resize screen captures when you save for PNG format
2. reduce colours to 256bit or 64bit before optimizing PNG/screen captures

eva2000 · 22-08-2008 until 31-12-2010

Overclocking Methodology
by eva2000

Updated: Febuary 2nd, 2009

There's no one set of settings that will work for everyone. There's only a basic methodology to overclocking and it doesn't change regardless of platform you're on whether it be s939/AM2 or Intel 965P/975x/P35/X38/48/P45 or Nvidia 650i/680i/780i/790i. Even X58 follows the same method although the focus is much more on memory first as per brief general X58/i7 overclocking guide here.

Find your stable stock cpu/memory speed configuration settings. Yes that means to memtest86+ v2.11 and windows stability test (prime95/orthos/super pi/3dmark) you cpu and memory at stock speeds to confirm they're stable.
Find the max stable memory overclocking frequency and timings/subtimings of your memory. To do this try to keep cpu FSB at stock speeds or as close to stock speed/FSB and use memory dividers where possible to find your upper limits of your memory frequency. Regardless, of platform memory clocks and timings/subtimings respond to mainly 2 types of voltages, vdimm memory voltage and NB - northbridge voltage. On some platforms such as Asus Commando/P5B Deluxe/P5K series and Striker Extreme they also may respond to vFSB and SB - southbridge voltages. X58 notes: memory responds to memory and uncore voltages. Uncore voltages are sometimes termed CPU VTT or on Asus QPI/DRAM voltages. Intel max spec'd uncore = 1.35v
Next part is really a follow on from step 2, that is to find out your memory's max overclocking frequency and timings/subtimings with each memory divider to see if there are differences in internal latencies with each memory divider. Example, is on Asus P5B Deluxe 4:5 divider has a very very tight internal latencies compared to 1:2, 2:3, 3:5 on the Asus P5B Deluxe. This means while you could clock you memory to 600mhz on 2:3 divider with say 2.4-2.6v vdimm, you might need 2.4v just to get to 500-520mhz on 4:5 divider and that 500-520mhz on 4:5 divider might only be a fraction slower in performance than 600mhz on 2:3 divider. Each motherboard is different, so you need to test this for yourself. Again, the key is in 2 types of voltages, vdimm memory voltage and NB - northbridge voltage and basically how good your memory is. X58 notes: Core i7/X58 platform now uses memory multipliers x bclk to determine memory frequency speed. Most X58 boards will provide 6x, 8x, 10x, 12x, 14x, and 16x memory multipliers. So default Core i7 cpu has 133bclk. With default 6x memory multiplier = 133x 6 = DDR3-800Mhz speeds. Max frequency on X58 may vary depending on memory multiplier used as well. Higher the memory clock and sometimes bclk, means the higher the Uncore voltage needed. It's not uncommon to need 1.5-1.6v Uncore volts to reach >DDR3-1866 speeds even for manufacturer out of the box >DDR3-1866 rated memory.
Next, is to find your max FSB frequency. That means dropping your cpu multiplier if you can and trying your max FSB at 1:1. From step 2, you would of already found your memory's max overclock frequency so bear this in mind when you are testing for max FSB at 1:1. Max FSB on core2 duo platforms respond to several different voltages depending on the motherboard and it's chipset they'd include vFSB, CPU PLL voltage, GTL+ Reference voltages and the most important of them for Intel 965P/975x/P35 or Nvidia 650i/680i at least is NB - northbridge voltage. Better Northbridge and Southbridge cooling (for Asus P5K series seeing as CPU PLL volts is tied with SB voltage on this series), may result in slightly better max FSB. X58 notes: Max bclk is determined by individual cpu, amount of Uncore voltage used as well as in some instances IOH/ICH voltages used. I suspect the motherboard design also helps somewhat but too early to conclude since I've only tested one X58 board - DFI UT X58-T3EH8.
Regarding voltages and settings in bios in general, I usually start with AUTO values to initial tests, then retest stability and performance to compare with manually set values. Then for voltages I pick the lowest value I can select for manually setting the default voltages which I'd use for my base reference overclocks. I rarely will use AUTO voltages for the important voltages such as Vcore, Vdimm, NB and vFSB. Update: For P45 chipsets be careful of AUTO and even lowest selectable voltages for vFSB/CPU VTT, NB, CPU PLL etc and some P45 boards reported to auto jack up voltages as FSB rises and you could be using cpu killing voltages that you are unaware of!! X58 notes: For Core i7 /X58 key voltages are vcore, vdimm, uncore and to some extent IOH/ICH voltages. To keep cpu safe with greater than 1.65v memory voltage, the memory voltage must be within 0.5v of uncore voltage. Now some memory have XMP profiles which do all the voltage and memory settings for you automatically. Some manufacturer's XMP profiles set pretty higher uncore voltages like 1.5+ v especially for >DDR3-1866 rated memory. You may need more or less than that preset amount of uncore voltage. But I prefer to disable XMP profiles and manually test what voltages are needed. For 24/7, you'd want uncore voltage below 1.35v.

When you have tested for all 4 steps outlined above, you can then put it all together to find your optimal overclocks and test for overal system stability. This would require alot of reboots and testing with both memtest86+ v2.11 initially followed by windows stability tests - basically the same or similar procedure and applications I outlined for AMD64 Overlocking Guide AMD64 Overclocking Guide (07/11/06).

The above outlined steps you can see in action in nearly all my motherboard/cpu overclocking threads hence why I never really did an Core 2 Duo overclocking guide as basically every core2 duo cpu/board overclocking thread I started is a guide as such - all you need is to read the threads carefully

Some examples of the steps outlined include:

Note, not all threads follow the outlined steps fully since some of the steps are already done before hand i.e. you already know your cpus max FSB or max cpu mhz and/or memory's max overclocking frequency and timings/subtimings from prior tests with the same memory on the same motherboard. Hence, why I think memory overclocking is the most important part of the whole system overclocking process. Without good overclocking memory the rest of your system will suffer in terms of optimal overclocking potential.

If you have questions specific to your cpu/board/memory please DO NOT private message me but rather start a new thread in the relevant forum on i4memory.com outlining your system specs, bios settings, bios version etc.

11-06-2010 until 12-12-2010
eva2000 Administrator Join Date: Jul 22 2004 Location: Brisbane, Australia Posts: 22,181	[Note]: NOScript / adblock users - best i4memory.com viewing experience If you use NOScript or other ad blocking scripts, you may experience issues viewing or using i4memory.com. To fix please read note here.

23-03-2010 until 23-12-2010
eva2000 Administrator Join Date: Jul 22 2004 Location: Brisbane, Australia Posts: 22,181	i4memory.com Ver 3.0 design rolled out i4memory.com ver 3.0 site and forum design has been rolled out. Full details here

10-01-2010 until 10-12-2010
eva2000 Administrator Join Date: Jul 22 2004 Location: Brisbane, Australia Posts: 22,181	Unread Posts since your last visit - Added New feature added to forums - read here

11-09-2008 until 31-12-2012
eva2000 Administrator Join Date: Jul 22 2004 Location: Brisbane, Australia Posts: 22,181	New IMG/IMGF/IMGFS/URLF bbcode tags reminder Just a reminder that new bbcodes have been added so make use of them for your screenshots when needed > [feedback] Additional custom IMG Tags

08-09-2008 until 31-12-2012
eva2000 Administrator Join Date: Jul 22 2004 Location: Brisbane, Australia Posts: 22,181	Irfanview.com freeware - nice image editor [updated] I notice folks posting very large screenies including their whole desktop in background. It's alot easier to reduce and crop to only the area of a screen capture you want to show. I thought i'd show you how much of a reduction you can get in size of the final screen shot. Download www.irfanview.com and plugins... prnt-scr -> open irfanview -> ctrl-v to paste image -> use mouse to highlight only area you want to show -> ctrl-Y to crop that area -> ctrl-R to resize it to 450-800pixels for forums -> save as -> jpeg -> options set at 65-75% quality I used 65% jpeg quality on all below screen shots. Full Desktop: * prnt scr -> paste into irfanview -> jpeg 65% quality -> save as = 191kb * print scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed INCLUDES taskbar/clock -> save as = 182kb * print scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed crop away taskbar/clock -> save as = 160kb * print scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed crop away taskbar/clock -> resized smaller to legible font size (CTRL+R) -> save as = 108kb 3x CPUZ screenies + A64Tweaker: * prnt scr -> paste into irfanview -> jpeg 65% quality -> save as = 124kb * prnt scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed showing -> save as = 111kb * prnt scr -> paste into irfanview -> jpeg 65% quality -> cropped to only info needed showing -> resized (CTRL+R) -> save as = 72kb For resizing (CTRL+R) ensure preserve aspect ratio is checked: Additional Tips: 1. You can also batch resize folders of pictures all at once. Read Irfanview help file to find out how 2. If you save a screenshot which is unedited, without resizing as PNG file, you will get a smaller size than any JPEG compressed quality picture. However, if you resize a screenshot then save as PNG file it will result in a size larger than JPEG compressed quality picture. For example below picture thumbnail (44.79KB) is resized, cropped JPEG 60% quality which links to an non-resized original screen capture saved as PNG (66.39KB) Update: September 21, 2006 Discovered you can save a further 40-50% reduction in screenshot size if you use Irfanview to convert to greyscale BEFORE you save an non-resized screen capture to PNG format with PNGOUT enabled at compression level 9. These screenshots with PNG format at compression level 9 were around 36-60KB in size but now check out the image sizes! Updated: March 15, 2007 More optimisation tweaks for smaller screen capture file sizes from SLI_dog By SLi_dog This is a how to on creating small file size screen captures using irfanview which is a truely great app thats absolutely free First up you'll want to grab your self a copy of Irfanview and the plug in pack too. www.irfanview.com I've found that the smallest size screen captures are almost always in .png format. Irfanview has a great plugin called "png out" which really helps reduce file size even further. Here are some examples. The following pictures are of the full desktop captured using the "CTRL+PrtScn" keys. The first is just a normal save with irfanview in png using compression value 9. Notice that the image quality is as good as you'd see it on the screen. File size = 122kb This is the exact same save in .jpg format with the default jpg compression value of 80%. Notice that the image quality has been lost from that of the normal desktop. The file size has gotten bigger too File size = 147kb Now I'll just do a simple image crop by selecting the area I want to crop to by holding the left mouse button and dragging a box around the area then pressing "CTRL+Y" to crop. You can always resize the area after by dragging the edges of the crop selection box. This drops the file size to only 106kb, that's already a 1/6th size saving which means faster loading web pages for easier browsing. File size = 106kb At this stage I'll introduce the pngout plugin. The following is using the pngout default settings. Notice that the image quality is still as good as you'd get on the screen. File size = 85kb Now the real trick to reducing file size is to reduce the number of colours in the picture. You do get some decrease in picture quality but it's still very readable and the file size saving are huge. Here's how you do it. Make sure you UNTICK the Use Floyd-Steinberg dithering or you'll end up with larger file sizes The following are advanced settings for the pngout plugin that I've found give the best reduction most of the time. You want to change the "Compression level" to 9, change the "Block split threshold" to 4096, and click on the "Keep chunks" button and hit "Select none" And if you put together the Reduce Colour Depth and Advanced png out settings you can get a file size reduction like this without compromising image quality too much. File size = 32kb I should point out that this will only work for simple screen captures, if you have normal photos or have a photgraphic windows backdrop then you may want to use .jpg I think that having a 3/4th reduction is size is worth not having such a pretty backdrop while benchmarking Don't just take my word for it, play around and see if you can find any better settings. If you do I'd be keen to know. Update: September 8, 2008 Just an update, if you find irfanview PNGout method too slow and don't mind sacrificing some compression for speed, check out PNGoptimizer as well as other tools listed at http://forums.legitreviews.com/about16702.html. Hint: 1. only crop don't resize screen captures when you save for PNG format 2. reduce colours to 256bit or 64bit before optimizing PNG/screen captures

22-08-2008 until 31-12-2010
eva2000 Administrator Join Date: Jul 22 2004 Location: Brisbane, Australia Posts: 22,181	Overclocking Methodology by eva2000 Overclocking Methodology by eva2000 Updated: Febuary 2nd, 2009 There's no one set of settings that will work for everyone. There's only a basic methodology to overclocking and it doesn't change regardless of platform you're on whether it be s939/AM2 or Intel 965P/975x/P35/X38/48/P45 or Nvidia 650i/680i/780i/790i. Even X58 follows the same method although the focus is much more on memory first as per brief general X58/i7 overclocking guide here. Find your stable stock cpu/memory speed configuration settings. Yes that means to memtest86+ v2.11 and windows stability test (prime95/orthos/super pi/3dmark) you cpu and memory at stock speeds to confirm they're stable. Find the max stable memory overclocking frequency and timings/subtimings of your memory. To do this try to keep cpu FSB at stock speeds or as close to stock speed/FSB and use memory dividers where possible to find your upper limits of your memory frequency. Regardless, of platform memory clocks and timings/subtimings respond to mainly 2 types of voltages, vdimm memory voltage and NB - northbridge voltage. On some platforms such as Asus Commando/P5B Deluxe/P5K series and Striker Extreme they also may respond to vFSB and SB - southbridge voltages. X58 notes: memory responds to memory and uncore voltages. Uncore voltages are sometimes termed CPU VTT or on Asus QPI/DRAM voltages. Intel max spec'd uncore = 1.35v Next part is really a follow on from step 2, that is to find out your memory's max overclocking frequency and timings/subtimings with each memory divider to see if there are differences in internal latencies with each memory divider. Example, is on Asus P5B Deluxe 4:5 divider has a very very tight internal latencies compared to 1:2, 2:3, 3:5 on the Asus P5B Deluxe. This means while you could clock you memory to 600mhz on 2:3 divider with say 2.4-2.6v vdimm, you might need 2.4v just to get to 500-520mhz on 4:5 divider and that 500-520mhz on 4:5 divider might only be a fraction slower in performance than 600mhz on 2:3 divider. Each motherboard is different, so you need to test this for yourself. Again, the key is in 2 types of voltages, vdimm memory voltage and NB - northbridge voltage and basically how good your memory is. X58 notes: Core i7/X58 platform now uses memory multipliers x bclk to determine memory frequency speed. Most X58 boards will provide 6x, 8x, 10x, 12x, 14x, and 16x memory multipliers. So default Core i7 cpu has 133bclk. With default 6x memory multiplier = 133x 6 = DDR3-800Mhz speeds. Max frequency on X58 may vary depending on memory multiplier used as well. Higher the memory clock and sometimes bclk, means the higher the Uncore voltage needed. It's not uncommon to need 1.5-1.6v Uncore volts to reach >DDR3-1866 speeds even for manufacturer out of the box >DDR3-1866 rated memory. Next, is to find your max FSB frequency. That means dropping your cpu multiplier if you can and trying your max FSB at 1:1. From step 2, you would of already found your memory's max overclock frequency so bear this in mind when you are testing for max FSB at 1:1. Max FSB on core2 duo platforms respond to several different voltages depending on the motherboard and it's chipset they'd include vFSB, CPU PLL voltage, GTL+ Reference voltages and the most important of them for Intel 965P/975x/P35 or Nvidia 650i/680i at least is NB - northbridge voltage. Better Northbridge and Southbridge cooling (for Asus P5K series seeing as CPU PLL volts is tied with SB voltage on this series), may result in slightly better max FSB. X58 notes: Max bclk is determined by individual cpu, amount of Uncore voltage used as well as in some instances IOH/ICH voltages used. I suspect the motherboard design also helps somewhat but too early to conclude since I've only tested one X58 board - DFI UT X58-T3EH8. Regarding voltages and settings in bios in general, I usually start with AUTO values to initial tests, then retest stability and performance to compare with manually set values. Then for voltages I pick the lowest value I can select for manually setting the default voltages which I'd use for my base reference overclocks. I rarely will use AUTO voltages for the important voltages such as Vcore, Vdimm, NB and vFSB. Update: For P45 chipsets be careful of AUTO and even lowest selectable voltages for vFSB/CPU VTT, NB, CPU PLL etc and some P45 boards reported to auto jack up voltages as FSB rises and you could be using cpu killing voltages that you are unaware of!! X58 notes: For Core i7 /X58 key voltages are vcore, vdimm, uncore and to some extent IOH/ICH voltages. To keep cpu safe with greater than 1.65v memory voltage, the memory voltage must be within 0.5v of uncore voltage. Now some memory have XMP profiles which do all the voltage and memory settings for you automatically. Some manufacturer's XMP profiles set pretty higher uncore voltages like 1.5+ v especially for >DDR3-1866 rated memory. You may need more or less than that preset amount of uncore voltage. But I prefer to disable XMP profiles and manually test what voltages are needed. For 24/7, you'd want uncore voltage below 1.35v. When you have tested for all 4 steps outlined above, you can then put it all together to find your optimal overclocks and test for overal system stability. This would require alot of reboots and testing with both memtest86+ v2.11 initially followed by windows stability tests - basically the same or similar procedure and applications I outlined for AMD64 Overlocking Guide AMD64 Overclocking Guide (07/11/06). The above outlined steps you can see in action in nearly all my motherboard/cpu overclocking threads hence why I never really did an Core 2 Duo overclocking guide as basically every core2 duo cpu/board overclocking thread I started is a guide as such - all you need is to read the threads carefully Some examples of the steps outlined include: Asus P5K Deluxe - Intel P35 Bearlake + Intel E6600 B1 ES / Q6600 Kentsfield / X6800 Asus P5B Deluxe + E6600/E6700/Q6600 Intel Core 2 Extreme X6800 B2 + P5B Deluxe E6600 B1 ES - aiming for 4500+ Mhz + Asus Commando E6700 L629B079 >4200Mhz + Asus Commando E6750 'G0' 8x333FSB + Asus P5K Deluxe Q6600 B3 ES @4009Mhz - Asus P5K Deluxe + 14-18C ambients Intel Core 2 Duo E6600 + Asus 680i Striker Extreme SLI motherboard Results Intel Core 2 Duo E6700 L629B079 + P5B Deluxe 0804/1101 bios Intel Core 2 Extreme X6800 B2 + Asus 680i Striker Extreme Intel Core 2 Duo E6700 + Asus 680i Striker Extreme SLI Intel Core 2 Quad Q6600 + Asus 680i Striker Extreme SLI preview results Note, not all threads follow the outlined steps fully since some of the steps are already done before hand i.e. you already know your cpus max FSB or max cpu mhz and/or memory's max overclocking frequency and timings/subtimings from prior tests with the same memory on the same motherboard. Hence, why I think memory overclocking is the most important part of the whole system overclocking process. Without good overclocking memory the rest of your system will suffer in terms of optimal overclocking potential. If you have questions specific to your cpu/board/memory please DO NOT private message me but rather start a new thread in the relevant forum on i4memory.com outlining your system specs, bios settings, bios version etc.