User panel stuff on forum
  53 posts on 2 pages  First page12Last page
General Discussion
2020-05-12, 15:08
News Writer
906 posts

Registered:
Jan 2006
Has anyone tried this extremely cheap APU for quakeworld? Only USD$50 and both the CPU and GPU components are unlocked and overclockable.

Would love to see how it performs as it could be built in to a very low cost ultra portable QW LAN PC in something like this:

https://www.in-win.com/uploads/Product/gaming-chassis/chopin/chopin_overview_space_02.png
2020-05-15, 11:17
Member
80 posts

Registered:
Apr 2017
I'm pretty sure that could do 308 FPS easily. I don't know if it could reach the 1000+ fps that you're used to. The question is can you really notice the difference between 308 fps and 1000+ fps?. Have you ever done a proper test?. We should make a script that can randomly switch between 308 and 1000+ fps and then ask the player to guess!.

The bottleneck for Integrated graphics is the memory and so some users will buy the fastest memory possible.. But I'm not sure if this is wise as an APU has a reduced number of shaders to match the reduced memory bandwidth..

When I needed to travel I would place my motherboard with all the components still plugged in, into the box that came with my motherboard. Very compact
2020-05-15, 12:17
Member
340 posts

Registered:
Nov 2006
Maybe it helps as reference: Running Ryzen 5 3400G on the new ezQuake "vid_renderer 1"


Asked for OpenGL 4.3, got X.Org
error: gamma size '0' seems weird, refusing to use it
&c0f0Renderer&r: OpenGL (classic)
Triple-buffering of GL buffers: enabled

OpenGL (classic)
Vendor: X.Org
Renderer: AMD RAVEN (DRM 3.36.0, 5.6.0-1-amd64, LLVM 10.0.0)
Version: 4.6 (Compatibility Profile) Mesa 20.0.6
Textures:
Units: 8
Size: 16384
Preferences
Image Format: 0x1908 (RGBA)
Image Type: 0x1406 (FLOAT)
Lightmaps: RGBA/UBYTE
Supported features:
Shaders: &c0f0available&r
Compute: &c0f0available&r
Framebuffers: &c0f0available&r
Tex arrays: &c0f0available&r
Tex samplers: &c0f0available&r
HW lighting: &c0f0available&r
Video
Resolution: 3440x1440@60hz [fullscreen]
Format: 24-bit color
24-bit z-buffer
0-bit stencil
RAW mouse input initialized


Running "timedemo demo1"


969 frames 1.6 seconds 599.4 fps


@ 640x480:


969 frames 0.5 seconds 2119.7 fps
2020-05-15, 17:02
Member
206 posts

Registered:
Feb 2011
lemonjuiced wrote:
The question is can you really notice the difference between 308 fps and 1000+ fps?. Have you ever done a proper test?.

Dirtbox had a post here (that I can't find) explaining how he can even notice a difference between 1001fps and 1155fps.
2020-05-15, 21:08
Member
48 posts

Registered:
Jan 2010
hah, personally I can easily spot difference between 308, 616 and 1001. Not sure about 1001 vs 1155 as never tried it.
2020-05-15, 21:52
Member
80 posts

Registered:
Apr 2017
Interesting... Does it feel different or look different?, or maybe both? Are the tear lines more noticeable?. Or does it just feel less laggy?.

Personally I can't even get 1000 fps to test it myself...
2020-05-16, 16:02
Member
147 posts

Registered:
Nov 2006
I made a placebo.cfg.
Just change var1 and var2 to the variable you want like cl_earlypackets, sys_yieldcpu, sys_highpriority or whatever may be the case.

Quote:
alias var1 "cl_maxfps 308"
alias var2 "cl_maxfps 1001"
set_alias_str var1text var1
set_alias_str var2text var2
alias placebo "if $rand > 0.5 then set choice 1 else set choice 0; if $choice == 1 then var1 else var2; echo Press 1 for $var1text or 0 for $var2text"
alias count "echo right: $right, wrong: $wrong"
alias correct "echo correct; inc right; count"
alias fail "echo wrong; inc wrong; count"
alias placebo_restart "set right 0; set wrong 0; placebo"
placebo_restart
bind 1 "if $choice == 1 correct else fail; placebo"
bind 0 "if $choice == 0 correct else fail; placebo"
2020-05-16, 17:31
Member
80 posts

Registered:
Apr 2017
I just did a test using your script, Matrix!.

I can't hit 1000 fps but I found that I could still tell the difference between 308 and 616. I noticed that at 308 fps things seemed slightly more blurry than at 616. But what's interesting is that I then tested between 72 fps (my monitor's refresh rate) and 616 fps and I couldn't tell the difference at all.

And in terms of lag/latency all fps settings felt the same to me.

So it seems that if your fps is multiples of your monitor's refresh rate then you won't notice the difference between fps settings. But if your monitor's refresh rate isn't multiples of 77 (QW physics) then things may start to seem a little blurry with lower fps settings.

This does make sense when considering my earlier findings that if I used 77 fps on a 72hz monitor then everything would seem jittery. But if I used higher multiples of 77 then things would start to smoothen out the higher I went. Yet if I used 72 fps or higher multiples of 72 on a 72hz monitor then things would be smooth regardless of how high the fps was.

So the conclusion:

Use multiples of your monitor's refresh rate for maximum smoothness.. But keep in mind that this can potentially result in more noticeable tear lines.
2020-05-19, 09:25
News Writer
906 posts

Registered:
Jan 2006
Tuna wrote:
Maybe it helps as reference: Running Ryzen 5 3400G on the new ezQuake "vid_renderer 1"

Hey Tuna, are you able to do a timedemo on FS vs LA e1m2 4on4 (that comes bundled in nQuake)? that is the best demo to use.

(Edited 2020-05-19, 10:11)
2020-05-19, 09:27
News Writer
906 posts

Registered:
Jan 2006
er wrote:
hah, personally I can easily spot difference between 308, 616 and 1001. Not sure about 1001 vs 1155 as never tried it.

Yeah there is a quite a big difference and very easy to tell.

To answer BloodDog(BigDick) - Look at an item and move your mouse left and right so your crosshair goes either side of an item that you are looking at. The higher the fps, the less blurry the item is. This also goes for enemies that you are aiming at. 616fps is probably the minimum you want to use, 770 and 1001 are also good but the optimal value i've tested is 1155. anything above that yields diminishing returns.

Oh and you can enable the vidlag hud element which shows the delay in how long a frame was rendered. the higher the fps, the lower the delay
2020-05-19, 12:29
Administrator
1859 posts

Registered:
Feb 2006
dirtbox wrote:
Hey Tuna, are you able to do a timedemo on FS vs LA e1m2 4on4 (that comes bundled in nQuake)? that is the best demo to use.

Using the nQuake package is probably the best comparison for benchmarking, but be aware that timedemo will give lower average FPS than you get while playing as VWEP's are enabled during demo playback.
2020-05-19, 17:12
Member
340 posts

Registered:
Nov 2006
dirtbox wrote:

Hey Tuna, are you able to do a timedemo on FS vs LA e1m2 4on4 (that comes bundled in nQuake)? that is the best demo to use.



WQHD

24819 frames 36.6 seconds 678.1 fps


SVGA (-window)

24819 frames 19.2 seconds 1292.1 fps


This is vanilla ezQuake config. Couldn't run on nQuake due to missing correct libjpeg library and I didn't even try messing with configs. All Linux.

Note that the 3400G has 11 GPU CUs while the Athlon 3000 has only 3 I believe. That may make quite a difference. And of course ram speed may have an impact. Never tested whether it makes a significant difference for Quake. I run 3200 Mhz dual channel.
2020-05-22, 11:00
Member
22 posts

Registered:
Sep 2016
Can you try the timedemo again but set r_dynamic 0 and gl_detail 0 first? Maybe also try gl_max_size 32.
2020-05-23, 20:16
Member
340 posts

Registered:
Nov 2006
Sure can do.. i can use a config ttoo if you provide one. Also hope to upgrade to 4700G when its available and if it runs on this board..


24819 frames 31.3 seconds 791.9 fps
24819 frames 13.4 seconds 1857.2 fps
2020-05-27, 12:29
News Writer
906 posts

Registered:
Jan 2006
Tuna wrote:
Sure can do.. i can use a config ttoo if you provide one. Also hope to upgrade to 4700G when its available and if it runs on this board..

My config comes bundled with nQuake or its on the front page of qw.nu to download. Can you load that config and do a timedemo of the e1m2 4on4 demo that comes with nquake? that would be so good. Thanks Tuna!
2020-05-28, 07:16
Administrator
1859 posts

Registered:
Feb 2006
He's running ultrawide 1440p (3440x1440), dunno how to force a specific resolution in 3.5 between systems, but I think the benchmarks would be more relatable in widescreen 1080p (1920x1080).
2020-05-28, 12:53
News Writer
906 posts

Registered:
Jan 2006
Zalon wrote:
He's running ultrawide 1440p (3440x1440), dunno how to force a specific resolution in 3.5 between systems, but I think the benchmarks would be more relatable in widescreen 1080p (1920x1080).

who are you referring to?
2020-05-29, 17:30
Administrator
1859 posts

Registered:
Feb 2006
dirtbox wrote:
who are you referring to?

Tuna, who's doing the benchmarks
2020-05-30, 11:43
News Writer
906 posts

Registered:
Jan 2006
Zalon wrote:
Tuna, who's doing the benchmarks


1857.2 fps at that high-res is pretty decent. would love to know the 1080p and 1440p numbers
2020-05-30, 11:57
Member
80 posts

Registered:
Apr 2017
dirtbox wrote:
Zalon wrote:
Tuna, who's doing the benchmarks


1857.2 fps at that high-res is pretty decent. would love to know the 1080p and 1440p numbers


The 1875.2 was for 640x480.
2020-05-30, 17:59
Member
340 posts

Registered:
Nov 2006
Dirtbox config, new renderer


24819 frames 13.7 seconds 1805.6 fps [640x480]
24819 frames 30.0 seconds 827.2 fps [3440x1440]
24819 frames 23.7 seconds 1047.1 fps [1920x1080]
2020-05-30, 18:22
Administrator
1859 posts

Registered:
Feb 2006
Tuna wrote:
Dirtbox config, new renderer


24819 frames 13.7 seconds 1805.6 fps [640x480]
24819 frames 30.0 seconds 827.2 fps [3440x1440]
24819 frames 23.7 seconds 1047.1 fps [1920x1080]


Is this on the FS vs LA demo included with nQuake?
2020-05-30, 18:46
Member
340 posts

Registered:
Nov 2006
Zalon wrote:

Is this on the FS vs LA demo included with nQuake?


Should be, yes.
2020-06-03, 03:51
News Writer
906 posts

Registered:
Jan 2006
Tuna wrote:
Dirtbox config, new renderer


24819 frames 13.7 seconds 1805.6 fps [640x480]
24819 frames 30.0 seconds 827.2 fps [3440x1440]
24819 frames 23.7 seconds 1047.1 fps [1920x1080]


Thanks, that is perfect. If you are getting that amount of fps on a 3400G then a 3000G definitely won't be good enough for 1155fps solid.
2020-06-09, 00:27
News Writer
906 posts

Registered:
Jan 2006
Tuna wrote:
Sure can do.. i can use a config ttoo if you provide one. Also hope to upgrade to 4700G when its available and if it runs on this board..


What do you know about the Ryzen 7 4700G?

I found this benchmark which says it is 8 cores/16 threads and Base clock 3.6 GHz, turbo 4 GHz.

I ended up building the mini-PC with the Athlon 3000G APU anyway as primarily it is just for a streaming PC but later I also plan to put in a Ryzen 4000 series APU. I just need a mini-PC that is capable of playing PUBG and I really hope that the Ryzen 7 4700G can achieve that later.

I actually had some great success with the Athlon 3000G. I was able to overclock the CPU from 3.5ghz to 3.9ghz and the GPU from 1100mhz to 1600mhz using the stock cooler.
The e1m2 4on4 timedemo really crushes the APU but for 1on1 games it stays above 700fps which means capping it to 616fps will get you stable duels.
2020-06-10, 17:27
Member
340 posts

Registered:
Nov 2006
dirtbox wrote:

What do you know about the Ryzen 7 4700G?

I think all the details leaked + some benchmarks too. Basically CPU wise it will be a big update (twice the performance probably). That just won't help you with games though. It comes with 8 CUs instead of 11 CUs as the 3400G does. They are more optimized which kind of mitigates the 3 less CUs. Looks like it will have only minimal gains in GPU performance - perhaps about 10%.
2020-06-11, 11:40
Member
22 posts

Registered:
Sep 2016
Dirtbox, how much did it cost to make that build?
2020-06-11, 11:40
Member
22 posts

Registered:
Sep 2016
I thought that athlon would do better in qw, since i played a lot on intel hd 4600 inbuilt gpu, and all timerefresh i did to test was over 1000fps. (on 1680x1050 res), and that 3000g should have better gpu.
2020-06-11, 13:34
News Writer
906 posts

Registered:
Jan 2006
Repast wrote:
I thought that athlon would do better in qw, since i played a lot on intel hd 4600 inbuilt gpu, and all timerefresh i did to test was over 1000fps. (on 1680x1050 res), and that 3000g should have better gpu.

The Athlon 3000G is the cheapest AMD CPU available. Maybe even cheaper than all Intel CPU's also. Only USD$50 (AUD$85).

I got a good motherboard with a B450 chipset and the rest of the parts I already have so it didn't cost me a lot. The plan is to drop in a 4700G when they come out which is going to be 8 cores 16 threads + GPU.
2020-06-11, 13:52
News Writer
906 posts

Registered:
Jan 2006
Repast wrote:
I thought that athlon would do better in qw, since i played a lot on intel hd 4600 inbuilt gpu, and all timerefresh i did to test was over 1000fps. (on 1680x1050 res), and that 3000g should have better gpu.

I just realised you said a timerefresh, that is very very different to a timedemo. a timerefresh is not a useful way in measuring a PCs performance.

nQuake comes with a 4on4 on e1m2 demo which is about as intensive as qw gets. Use that for a timedemo to see how a PC stacks up.
  53 posts on 2 pages  First page12Last page