[Various] Ashes of the Singularity DX12 Benchmarks - Page 111 - Overclock.net - An Overclocking Community

Forum Jump: 

[Various] Ashes of the Singularity DX12 Benchmarks

 
Thread Tools
post #1101 of 2682 (permalink) Old 08-27-2015, 08:21 AM
 
Join Date: Feb 2014
Location: between the Carpathian Mountains and the Baltic Sea.
Posts: 5,762
There's one thing that strikes me just now, although it should've stricken me on the 1st time I saw these charts. I know it must have been a serious discussion in some previous posts but I just don't have time to read 110 pages.
Why is everyone cheering AMD so much if 390X DX12 still loses to 980 DX11 in majority of those charts ?
Klocek001 is offline  
Sponsored Links
Advertisement
 
post #1102 of 2682 (permalink) Old 08-27-2015, 08:47 AM
Zen
 
Kpjoslee's Avatar
 
Join Date: Jan 2013
Location: Somewhere in US.
Posts: 955
Rep: 50 (Unique: 33)
Quote:
Originally Posted by Klocek001 View Post

There's one thing that strikes me just now, although it should've stricken me on the 1st time I saw these charts. I know it must have been a serious discussion in some previous posts but I just don't have time to read 110 pages.
Why is everyone cheering AMD so much if 390X DX12 still loses to 980 DX11 in majority of those charts ?

Because it is finally allowing GCN gpus the way it was intended to perform, as it is no longer bound by CPU overhead problem that severely limited their DX11 driver. For DX11, Nvidia's driver seems very well optimized so that is probably the reason why we are seeing good numbers in DX11. For DX12, it could be either how the code is optimized for Nvidia's architecture or problem on Nvidia's DX12 driver, or little bit of both. I don't think their DX12 performance should fall behind DX11's if the code and the driver is not the problem.
It is the game that is still work in progress. I think there is room for improvement on DX12 performance on both sides. It is hard to draw anything conclusive at this point.

My home PC
(15 items)
CPU
AMD Threadripper 1950x
Motherboard
Gigabyte Aorus X399 Gaming 7
GPU
EVGA Geforce RTX 2080 Ti XC Ultra
RAM
G.Skill DDR4 3600 CL16
Hard Drive
Samsung Evo 840 500GB
Hard Drive
Samsung 960 Pro 500GB
Power Supply
EVGA SuperNova G2 1300W
Cooling
Noctua NH-U14S TR4
Case
Corsair Carbide Air 540
Operating System
Windows 10 Pro
Monitor
Dell U2711
Monitor
Samsung 55" 4k
Keyboard
Corsair K70
Mouse
Logitech G502
Audio
Denon AVR-X3300W
▲ hide details ▲
Kpjoslee is offline  
post #1103 of 2682 (permalink) Old 08-27-2015, 09:26 AM
New to Overclock.net
 
GorillaSceptre's Avatar
 
Join Date: Oct 2014
Posts: 2,373
Rep: 334 (Unique: 132)
Quote:
Originally Posted by Mahigan View Post

The difference can be summed up like this:

AMD GCN Asynchronous Compute Solution:
  1. Each ACE fetches commands from R/W L2 Cache, Global Data Share Cache and On board memory (GDDR5/HBM) and form task queues
  2. 64 Total task queues can be formed by the ACEs (prioritizing pending tasks to be executed).
  3. 8 Compute tasks/cycle and 1 Graphics task/cycle are executed in real time
  4. Each ACE can operate independently and complete tasks out of order
  5. Each ACE can synchronize tasks with other ACEs and the Graphics Command Processor in order to operate "in order" but also in parallel
  6. ACEs can communicate through the R/W L2 Cache, Global Data Share Cache and On board memory (GDDR5/HBM) allowing for an incredible amount of memory bandwidth


nVIDIA HyperQ:
  1. Grid Management Unit receives thousands of pending tasks
  2. Grid Management Unit sends 32 tasks to the work distributor (32 compute tasks or 31 compute, 1 Graphics task)
  3. Work Distributor assigns 32 tasks across 32 AWS's.
  4. 32 Compute tasks/cycle or 31 Compute tasks/cycle and 1 Graphics task/cycle are executed in real time
  5. Each AWS cannot operate independently to complete tasks out of order
  6. Each AWS must synchronize tasks with other ACWs and complete these tasks "in order" but also operate in parallel
  7. If one task is dependent on another task, HyperQ receives a pipeline stall (pause)
  8. AWS's can communicate though a single medium, the L2 Cache, which is limited in memory bandwidth as it is shared with all elements of the Graphics and Compute pipeline

Doesn't that actually show that Maxwell 2 is more capable than GCN? headscratch.gif

Edit:

Just ignore that tongue.gif

CPU: 2600k @ 4.8GHZ
MB: Maximus IV Extreme-Z
GPU: MSI 390X
CASE: 650D
AUDIO: Xonar Essence One-AKG Q701
GorillaSceptre is offline  
Sponsored Links
Advertisement
 
post #1104 of 2682 (permalink) Old 08-27-2015, 09:30 AM
New to Overclock.net
 
Themisseble's Avatar
 
Join Date: Oct 2013
Posts: 2,001
Rep: 38 (Unique: 29)
Quote:
Originally Posted by GorillaSceptre View Post

Doesn't that actually show that Maxwell 2 is more capable than GCN? headscratch.gif

Nope:)
Themisseble is offline  
post #1105 of 2682 (permalink) Old 08-27-2015, 09:32 AM
 
Join Date: Feb 2014
Location: between the Carpathian Mountains and the Baltic Sea.
Posts: 5,762
Quote:
Originally Posted by Kpjoslee View Post

Because it is finally allowing GCN gpus the way it was intended to perform, as it is no longer bound by CPU overhead problem that severely limited their DX11 driver. For DX11, Nvidia's driver seems very well optimized so that is probably the reason why we are seeing good numbers in DX11. For DX12, it could be either how the code is optimized for Nvidia's architecture or problem on Nvidia's DX12 driver, or little bit of both. I don't think their DX12 performance should fall behind DX11's if the code and the driver is not the problem.
It is the game that is still work in progress. I think there is room for improvement on DX12 performance on both sides. It is hard to draw anything conclusive at this point.
yeah but I don't see people buryying nvidia if it still performs better on old api. people quoting 60% performace gain for amd are obviously confused. this 60% gain still makes 390X fall behind 980 running on an old api.
Klocek001 is offline  
post #1106 of 2682 (permalink) Old 08-27-2015, 09:39 AM
New to Overclock.net
 
airfathaaaaa's Avatar
 
Join Date: Aug 2015
Posts: 825
Rep: 62 (Unique: 31)
so let me get this straight

ace engines can actually work on the "Background" in a way to prebuff something so when it get asked to do so it will provide it with not latency at all?
but nvidia one in the same position they will have to complete the tasks till one engine gets free so that they could process the call?
airfathaaaaa is offline  
post #1107 of 2682 (permalink) Old 08-27-2015, 09:53 AM
 
Join Date: Nov 2009
Location: Turkiye
Posts: 6,565
Quote:
Originally Posted by GorillaSceptre View Post

Doesn't that actually show that Maxwell 2 is more capable than GCN? headscratch.gif

Edit:

Just ignore that tongue.gif
As much as a 32 core cpu with no cache structure makes competition to a 8 core 8 register per core cpu.
mtcn77 is offline  
post #1108 of 2682 (permalink) Old 08-27-2015, 09:54 AM
New to Overclock.net
 
GorillaSceptre's Avatar
 
Join Date: Oct 2014
Posts: 2,373
Rep: 334 (Unique: 132)
@Mahigan Haven't seen you respond to this?

From razor1.

"Slide 56 tells you what it can do 9 devices, 8+1 compute and graphics respectively and in queue 64+

yeah it has the same limitations as Maxwell 2 when it comes to register pressure slide 34

Quote:
However, the new architecture is more susceptible to: register pressure
using too many registers with a shader can reduce the maximum waves per SIMD

This is a direct quote from that slide

then it gives a table to clarify this

Maxwell 2's AWC's and AMD's ACE's are setup similarly, I think the slides talking about the cache which is like how CPU's cache works is what misled you. The reason for this is because the APU's would need this.

This slide deck should be available on line for everyone.

http://www.slideshare.net/DevCentral...e-by-layla-mah "

"Any case Fury X seems to come in at around the 980 ti performance wise in this early bench from AOS, so unless there is some magic sauce for the r290x, lets put this down to Fiji and maxwell 2 drivers aren't as mature in Dx12 yet. And AMD's Dx11 drivers were just forgotten about."

CPU: 2600k @ 4.8GHZ
MB: Maximus IV Extreme-Z
GPU: MSI 390X
CASE: 650D
AUDIO: Xonar Essence One-AKG Q701
GorillaSceptre is offline  
post #1109 of 2682 (permalink) Old 08-27-2015, 11:24 AM
New to Overclock.net
 
orlfman's Avatar
 
Join Date: Jul 2008
Location: arizona
Posts: 714
Rep: 67 (Unique: 54)
outside of the DX12 benchmarks, how is the actually game its self? is it fun for a rts? is it as fun as supreme commander forge alliance was? is it worth $45-$50 for game value?

ever since EA butchered the command and conquer franchise, there really has been a lack of good RTS as of late. sins of a solar empire was fun but ashes reminds me of SC-FA which i enjoyed.

orlfman is offline  
post #1110 of 2682 (permalink) Old 08-27-2015, 11:26 AM
New to Overclock.net
 
Themisseble's Avatar
 
Join Date: Oct 2013
Posts: 2,001
Rep: 38 (Unique: 29)
@Mahigan

Anything new about CPU performance?
Themisseble is offline  
Closed Thread

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 2 (0 members and 2 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off