Best way to accelerate graphics display performance on T60p?

[Sandstorm83]

It's a 14.1" WXGA+ (1440x900) T61.

There's a utility called GPU-Z which will show you the GPU activity when viewing that page. On my Nvidia NVS 140m the GPU activity jumps up to 60%.

[Backslashnl1]

With my T60 config (see signature) at 1600x1200, I get between 6-12 fps, cpu running at 85~100% cpu usage.
Running the test again at 1024x768, performance increases to 12-20 fps.

[ScottWitte]

With my T60 config (see signature) at 1600x1200, I get between 6-12 fps, cpu running at 85~100% cpu usage.
Running the test again at 1024x768, performance increases to 12-20 fps.

The only relevant difference between your and my system seems to be the GPU, which obviously still has a big impact even if much of the work is being done by the CPU. 3DMark06 for the 950M is 132 vs 1203 for the V5200. That difference, while not directly scalable, is easily seen in your results.

And thanks for the brilliant idea of testing different screen resolution simply by changing it! The fewer pixels that need to be calculated each frame translates to faster FPS. In my case I get:

• 1600x1200 = 20FPS
• 1440x900 = 29/30FPS
• 1400x1050 = 28/30FPS (This is the closest line count to an HDTV's 1080 that the driver seems to be capable of.)
• 1024x768 = 40FPS average. (40FPS is the target so it would be the limit.)

A personal revelation: If I show my work to a group using a projector or conference room HDTV screen I should never try pushing more pixels than that display can handle and I will get better performance (which should be obvious).

[Backslashnl1]

Today I switched to Win7 x64 and ran the test again. Interesting I now have a different outcome:

1600x1200: 11-20 fps
1024x768: 20-46 fps, average around 35-40

Simply upgrading to Windows 7 increased my score, the hardware is exactly the same.
Also I tried to play some 1080p MKV movies, which would always stutter on winXP... and now they play smoothly! I did not expect this.

So it looks like the GPU has nothing to do with these results. I assume you run on Win7 also?

[ScottWitte]

Simply upgrading to Windows 7 increased my score, the hardware is exactly the same.

So it looks like the GPU has nothing to do with these results. I assume you run on Win7 also?

I'm running WinXP32 using 2GB (2x1GB) RAM. Your results are curious. Thinkpadophile's results suggested that Windows 7x64 would have no effect. That system lists 3GB RAM. If it is really 3GB, not 4GB using only 3GB (a T60 limitation), that may be the difference. You configuration could allow faster memory usage. I'm just blown away if that alone could make this much difference. I also don't know if combining a 2GB + 1GB RAM is even possible on a T60.

[RealBlackStuff]

I also don't know if combining a 2GB + 1GB RAM is even possible on a T60.

Yup, it's a good option.

[ScottWitte]

I also don't know if combining a 2GB + 1GB RAM is even possible on a T60.

Yup, it's a good option.

Hmmm. "Good" maybe in terms of valid but not "best" in terms of performance as I understand it. Isn't there something about interleaving or parallel access to memory that is possible with matched RAM modules vs a collection of RAM modules with only linear access?

That said, I wouldn't have thought it would make enough difference to account for the difference between Thinkpadophile's and Backslashnl1's results. But if not that, what?

[ThinkPadophile]

My T60p has 2 matched 2GB DDR2 modules for 4GB total (3GB usable).
Incidentally, I made a mistake in my earlier post. My external monitor is not UXGA (1600 X 1200) but WUXGA (1920 X 1200). That raises an interesting problem: why do we both report 20fps if your CPU is 16% faster than mine but my monitor is higher resolution than yours? Something seems amiss.

[ScottWitte]

My T60p has 2 matched 2GB DDR2 modules for 4GB total (3GB usable).
Incidentally, I made a mistake in my earlier post. My external monitor is not UXGA (1600 X 1200) but WUXGA (1920 X 1200). That raises an interesting problem: why do we both report 20fps if your CPU is 16% faster than mine but my monitor is higher resolution than yours? Something seems amiss.

Nothing amiss. You are pushing more pixels, 1900x1200 vs 1600x1200, which would slow you down. The slower CPU is an additional handicap. The fact that we got the same FPS despite that just supports the thesis that Win7x64 really does make a positive difference -- at least with 2x2GB memory.

Any chance you could undock the laptop and test with the UXGA screen?

[hoax32]

I get about 25-45FPS @ FULL 1280:800 in HD & FULL screen with my T7300 & NVS 140m

[ThinkPadophile]

Undocked and displaying on the UXGA LCD fps went up to 23.

[ScottWitte]

I get about 25-45FPS @ FULL 1280:800 in HD & FULL screen with my T7300 & NVS 140m

Thanks. It seems the reported numbers bounce around more when you are moving at the target speed of 40FPS. I would guess that is what is happening for you.

For comparison, at 1280x800 I get 30FPS. I'm surprised it isn't faster considering that I get 28/30 at 1440x1050.

Undocked and displaying on the UXGA LCD fps went up to 23.

Thanks. That sure supports Win7x64 adding a significant speed boost for CPU intensive operations. I can see yet another upgrade in my near future.

[Temetka]

I went to your T60 test page and I was able to get 32FPS on my T61 with Nvidia GPU and a T7500 on battery. I did not check under the "high performance" power option (i.e. cranking the system to max).

I'd second what has been mentioned before. Get a T61 motherboard for your machine. It is quite clear that the graphics alone are superior and of course the C2D is a more efficient CPU clock for clock as compared the CD in most T60 machines.

Other than that I also wanted to comment on your website. I really like your layout and it is very easy to navigate. Nice work.

EDIT: I plugged in my AC adapter and set Windows to High performance. I then viewed your test page full screen and received on average between 33 and 35 fps.

[ScottWitte]

I went to your T60 test page and I was able to get 32FPS on my T61 with Nvidia GPU and a T7500 on battery. I did not check under the "high performance" power option (i.e. cranking the system to max).

Thanks for checking. What resolution screen do you have?

When you say "high performance you are referring to the Thinkpad power management setting, correct?

[Temetka]

Yes, I am referring the power management setting.

My T61 runs at 1400x1050.

[ScottWitte]

My T61 runs at 1400x1050.

Thanks. I should have also asked which Nvidia GPU, which must also affect results.

As mentioned earlier, screen resolution has a major impact on this test. The more pixels computed for each frame the lower the FPS. I saw that when testing my system at different screen resolutions. So at 1400x1050 I got 28/30 FPS on AC compared to 20 FPS at 1600x1200. That was using WinXPx32, T7600 and V5200GPU.

Definitely the Frankenpad option will result in better performance, although I'm hoping it would be more than 10-20%. I'm thinking there are a few more tricks available to push this old system a little further before resorting to major surgery.

[ScottWitte]

Update: I did a clean install of Windows 7x64 then upgraded RAM from 2GB (2x1GB) to 4GB (2x2GB) and....Wow. Comparatives:

WinXPx32, 2GB, 1600x1200: 19/20 FPS
Win7x64, 2GB, 1600x1200: 23/24 FPS
Win7x64, 4GB, 1600x1200: 24/25 FPS

Curiously, my CPU was running around 80% under WinXPx32 while it runs in the low 90's under Win7x64. I assume this is what allows faster pixel crunching and hence higher FPS. I suspect it is faster/better memory access that is allowing the CPU to run faster.

So, simply upgrading from WinXPx32 to Win7x64 resulted in a 20% increase in FPS. Going 2GB to 4GB added another 5%. That is pretty slick, IMO.

Next step: Upgrade the fan and try some overclocking.

For anyone else wanting to run this test go to this: T60p Test Page. When it loads:

• Click the "Full Screen" button at lower right. (important!)
• Note the average FPS (Frames per Second) number at upper left while it is moving and after it has settled in.
• Just let it run automatically, don't zoom in as that will affect results. There will be brief pauses every 120°.
• Report FPS number back here with your hardware configuration (CPU, GPU, RAM, OS) plus anything special you may have done (overclocking???) and we will have apple to apple comparisons with little speculation.

[twistero]

Uh, using a Flash animation for benchmarking? I wouldn't call that authoritative. Flash isn't known for its programming efficiency. Also, different browsers or Flash plugin versions may have huge impacts on the result. You should probably consider a dedicated video benchmarking software for a consistent evaluation.

[ScottWitte]

Uh, using a Flash animation for benchmarking?I wouldn't call that authoritative.

Of course not. Except.... if you read the original post(s) you may see that my specific problem is this exact performance and the need to improve it. So it is authoritative for this sort of situation.

Nonetheless, this test should also be strongly indicative of the effect various hardware and software enhancements may have on other CPU intensive programs, including those that need to affect the display. I can tell you that I am seeing very noticeable improvements in browser performance, Photoshop and other real world programs I use. I, indeed most of us, very rarely need to run benchmarks in day to day work

[twistero]

my specific problem is this exact performance and the need to improve it. So it is authoritative for this sort of situation.

Whoops.
But I'm curious now, what browser do you test it on? Have you tried different browsers to see whether they perform differently? Or even the proprietary Flash plugin in Linux?

[ScottWitte]

But I'm curious now, what browser do you test it on? Have you tried different browsers to see whether they perform differently? Or even the proprietary Flash plugin in Linux?

Tested in IE8 and IE9 and FF11 (which uses same plugin as Safari and Chrome). Also tried a couple versions of Flash 11.x and later 10.x. ALL got essentially the same result, which is to be expected for this hardware. Linux really isn't an option for me.

A version of Flash 11 due later this year is scheduled to extend hardware acceleration back to the ATI X generation of GPUs such as in the T6X Thinkpads as I understand it. That would result in a significant performance improvement. Until then I want to tweak whatever I can out of this old boy.

[goofyGAguy]

My T42 with 1.8 GHz Dothan and ATi 9600 clocks in at a whopping 10 FPS.

[hoax32]

My T42 with 1.8 GHz Dothan and ATi 9600 clocks in at a whopping 10 FPS.

now $h*t is getting really serious!!!

[onix]

ScottWitte:
14/15 FPS
T60p, T2500, 2GB RAM, 15 inch Flexview screen 4:3, ATI Catalyst 10.2 drivers, WinXP Pro 32, FF11, Flash 11.1.102, CPU Usage 85%, RAM (Commit Charge) 989M/3939M.

dr_st:
I got the 16-19FPS with my T60, T7200, ATI X1400, Flash 11.1.102.63 with plugin-container (Firefox process responsible for running the flash plugin) taking up to 70% of CPU time, at full screen WSXGA+ resolution.

With another T60, T7200, Intel 950, Flash 11.1.102.55 I got 27-30FPS and SXGA+ resolution with plugin-container taking up to 85% of CPU time.

I wasn't aware the processor could be changed out on TP's. Really?? I have a 14.1" SXGA+ (1400x1050) T61p (a 128MB NVIDIA) and get between 31 ~ 44 FPS with a median around 36FPS running on Flash version 11,2,202,233 in Firefox 11.0. The CPU is a T9300 @ 2.50GHz with 4GB of RAM and on 64-bit Win7 Ultimate. I get a 4.2 Windows Experience Index.

[dr_st]

The CPU can be changed on most Thinkpads. On some of the slim models (X series and T***s series) it is soldered to the motherboard and can only be changed out with the whole board.

[ScottWitte]

An update: I tried ATI Tray Tools to accelerate the GPU. It made exactly no difference in results on this test. I still get 25/26 FPS in full screen mode at 1600x1200. CPU is running roughly 95%. whether or not GPU is accelerated.

For Comparison I also tested at 1400x1050. The FPS bounces around a lot more but gives me an average 30/32 FPS. CPU runs around 95%.

Conclusion: The current version of Flash doesn't use hardware acceleration for older GPUs so Flash rendering is highly CPU dependent. Since the CPU is running at capacity it is the bottleneck here. To get better performance I would either have to drop in a T7600G and overclock it or perform major surgery and create a Frankenpad. An overclocked T7600G should deliver a 20 to 25% performance increase so I should be able to achieve a solid 30/32 FPS at 1600x1200. I believe this would be comparable to a Frankenpad, interpolating from results that have been reported (at lower resolution). I don't think anyone with a 1600x1200 Frankenpad has reported, have they?

Come fall Flash should add hardware acceleration for the X era ATI GPUs. That should result in a significant performance increase.

[dr_st]

Come fall Flash should add hardware acceleration for the X era ATI GPUs. That should result in a significant performance increase.

Don't forget to come back and update this thread once it happens.

[hoax32]

OK I get about 9-15FPS on my Dell Latitude C610 which has a "Pentium III" 1GHz, a 16MB ATI Radeon and 512MB of SD RAM.

[ScottWitte]

OK I get about 9-15FPS on my Dell Latitude C610 which has a "Pentium III" 1GHz, a 16MB ATI Radeon and 512MB of SD RAM.

Always good to have a point of comparison Although, seriously, it's not that bad considering the hardware.

[Raceboy]

The CPU can be changed on most Thinkpads. On some of the slim models (X series and T***s series) it is soldered to the motherboard and can only be changed out with the whole board.

Umm, it is possible to change the CPU on X series without changing the board but it is not possible without few special tools (heat gun, BGA solder balls, stencil and BGA flux). I did it on my X31 (swapped in CPU from X41, Pentium M 758 1.5 GHz LV Dothan).
But if a person is not able to to this kind of mod, then yes, only with the board. Just sometimes you want better CPU than it was offered on specific board (like there was never a LV Dothan on X3x machines).