Intel's Clarkdale CPU + iGFX

[M_S]

Hell, you should be cruising just fine with that rig. The CPU doesn't draw that much power and as long as the graphics are in 2D mode, there is no stealing of cycles by the GPU so you get the whole benefit of the CPU power for what you need.

[M_S]

That looks like a pretty impressive DAW.

With respect to processors, Intel is very good to me, as are most other companies, so I don't really worry about those too much.

Let us know how it works out.

[DavidC1]

*snip snip*

Ah, there's the man. Where did you go all this time? That busy?

Anyways, please check UT3 out again. Set the screen percentage options in video from 50% to 100% and it won't look so pixellated anymore, and also perform normal.

BTW, I have done all tests with 3 driver revisions. There's total of 6 driver versions now, I stopped checking after the 3 as for the apps I have tested, the changes were very minor. I know the latest 2086(15.7.1) supports proper 24p playback, but I don't use this as HTPC.

A side effect of testing was confirmation of belief that hard drive performance DOES impact frames directly. When my freespace on my SSD dropped from 25GB to 15GB, performance dropped 10% in average.

Hell, you should be cruising just fine with that rig. The CPU doesn't draw that much power and as long as the graphics are in 2D mode, there is no stealing of cycles by the GPU so you get the whole benefit of the CPU power for what you need.

In reponse to Jimmyv:

The resource allocation between the CPU and the GPU core in Clarkdale doesn't work in a way so the CPU aids the GPU when necessary. No, when there's a situation where software VS might be faster than hardware VS(not true for Clarkdale), the entire VS pipeline is re-routed to run on the CPU instead.


A truly awesome way is if it can work completely with synergy to each other, so the CPU can "top-off" the performance required for 3D. But that might be too complex to do. It's not happening on here though.

[M_S]

*snip snip*

Ah, there's the man. Where did you go all this time? That busy?

Anyways, please check UT3 out again. Set the screen percentage options in video from 50% to 100% and it won't look so pixellated anymore, and also perform normal.

BTW, I have done all tests with 3 driver revisions. There's total of 6 driver versions now, I stopped checking after the 3 as for the apps I have tested, the changes were very minor. I know the latest 2086(15.7.1) supports proper 24p playback, but I don't use this as HTPC.

A side effect of testing was confirmation of belief that hard drive performance DOES impact frames directly. When my freespace on my SSD dropped from 25GB to 15GB, performance dropped 10% in average.

Hell, you should be cruising just fine with that rig. The CPU doesn't draw that much power and as long as the graphics are in 2D mode, there is no stealing of cycles by the GPU so you get the whole benefit of the CPU power for what you need.

Sorry, I just got back from Europe after two weeks of Cebit and other things in Munich and my backlog just ate me up, plus I am getting ready for another trip. For one or the other reason, the i5 rig is not cooperating with me right now, I guess I have to rebuild it and see what the story is.

[M_S]

Actually, I noticed myself some anomalies with low resolutions

I did not know UT3 wasn't supposed to look like that actually. I know it looked just like that on the G965. Maybe I'll download the demo and try out again.

Update: I have figured out what's wrong with Unreal Tournament 3. You have to put the "Screen Percentage" in settings to 100%.

I am a bit lost with that setting. I am always using "Fullscreen" mode, which would be 100%. Again, I can only guess what you mean here, but it makes sense because in "windowed" mode, you create a resource conflict between the 2D and the 3D engine, which you can eliminate by setting fullscreen mode, which completely turns off the 2D mode.

The same "issue" applies to a variety of different scenarios. When at the G4-TV (Attack of the show) we had to play games in "windowed" mode to disable SLI / Crossfire because otherwise the video feed wouldn't work. In other words, in that case, going to 100% screen percentage turns on Crossfire / SLI and disabled the 2D engine. The interesting thing about this is that aside from technical reasons, some of this simply relates to DRM, that is, if you are in Full-Screen mode, DRM demands disabling of "high-quality" video output to any secondary port as a means of copy protection.

These things are not really "technically related" but current implementations follow this scheme and therefore, as a side effect, you get more performance out of Fullscreen mode in applications where video resources are marginal

[DavidC1]

I am a bit lost with that setting. I am always using "Fullscreen" mode, which would be 100%. Again, I can only guess what you mean here, but it makes sense because in "windowed" mode, you create a resource conflict between the 2D and the 3D engine, which you can eliminate by setting fullscreen mode, which completely turns off the 2D mode.


There's a difference between "Screen Percentage" and "Fullscreen Mode". Here's a good article on it:
http://www.tweakguides.com/UT3_5.html

Screen Percentage: This slider controls the amount of screen space to render the game world in, and the lower the screen percentage, the smaller the game image will be on the screen. However, what actually happens is that the game takes this smaller image and by default upscales it to fill your screen (if UpscaleScreenPercentage=true in your UTEngine.ini file - see the Advanced Tweaking section). The result is that the in-game image becomes noticeably more blurry the lower the screen percentage setting is taken, and in return you gain more performance. In practice this is very similar to running the game at a specified proportion of your current resolution. For example a 50% screen percentage on a 1280x1024 resolution gives the visual equivalent of running at half the resolution.

I don't know why, but on both the G965 and Clarkdale the default setting on UT3 is 50%, meaning it can be confusing why it looks so bad and consequently perform higher. I'm thinking because the game engine itself doesn't recognize Intel GPUs properly and sets everything to low as possible, but that's just a theory.

[DavidC1]

I am a bit lost with that setting. I am always using "Fullscreen" mode, which would be 100%. Again, I can only guess what you mean here, but it makes sense because in "windowed" mode, you create a resource conflict between the 2D and the 3D engine, which you can eliminate by setting fullscreen mode, which completely turns off the 2D mode.

There's a difference between "Screen Percentage" and "Fullscreen Mode". Here's a good article on it:
http://www.tweakguides.com/UT3_5.html

Screen Percentage: This slider controls the amount of screen space to render the game world in, and the lower the screen percentage, the smaller the game image will be on the screen. However, what actually happens is that the game takes this smaller image and by default upscales it to fill your screen (if UpscaleScreenPercentage=true in your UTEngine.ini file - see the Advanced Tweaking section). The result is that the in-game image becomes noticeably more blurry the lower the screen percentage setting is taken, and in return you gain more performance. In practice this is very similar to running the game at a specified proportion of your current resolution. For example a 50% screen percentage on a 1280x1024 resolution gives the visual equivalent of running at half the resolution.

I don't know why, but on both the G965 and Clarkdale the default setting on UT3 is 50%, meaning it can be confusing why it looks so bad and consequently perform higher. I'm thinking because the game engine itself doesn't recognize Intel GPUs properly and sets everything to low as possible, but that's just a theory.

[M_S]

This is kind of funny, there was an entire email chain going back and forth between Epic and Intel telling me all kinds of things I already knew but apparently carefully skirting this subject. What you say makes perfectly sense.

I checked and as you probably know, the advanced setting for the '661 are
FOV: 90
Screen Percentage 50%
Texture Detail: 1

I had never really looked at those settings since most of the time I am running high-end video cards where it doesn't matter but that's what I noticed when I said something along the lines of only the first level of textures being loaded.

With the sliders adjusted to 100/100/5, it does look noticeably better than before. However, even at the max settings, the LOD is a far cry of what you get with the 785, it looks like it is loading 3 layers of textures even when 5 are specified in the settings.

Anyway, thanks for pointing this out, this was very helpful for understanding some of the stuff that is going on there.

[KTE]

It's been many months after launch but I still haven't seen thorough data for Clarkdale; a CPU I wanted to replace my AIO CPU/IGP with... has anyone matched the CPU power to these Clarkdales and then compared the iGFX to AMDs IGP offerings?

G35/G45 were abysmal in comparison. They looked very good in synthetics and when you paired them with a 30-50% faster CPU in low-end game benchmarks, but when the tilted setup was balanced, they performed no where close to even HD 3200 from 2007. I know these new CPUs do some of the game processing themselves to help their iGFX performance but I'm very skeptical about the iGFX being anywhere close to 880G/890GX level performance. It would be a huge jump for Intel to accomplish that so fast.

Since I'm about to competitvely play a Quake 3 port that is about 10 years old as well as CS:S, I'm hoping these CPU+iGFX combos can run them properly?

I haven't seen appropriate, controlled data to make up my mind yet. Just for the record, all AMD IGPs, all of Intels previous IGPs, HD 2400, HD 4350 and HD 5450 can't even run the Q3 port at 800x600, everything low/off, windowed, without averaging 40-60 FPS with massive regular dips down to 5-15 FPS.

[M_S]

I haven't tried Quake3, but I can set up the system again if I finally get a break and run a few quick tests.. it is too funny, though, just about 2 weeks ago I finally deleted some of the old Quake3 patches and benchmarks but I might have some backups somewhere.