Overclock.net › Forums › Benchmarks › Overclock.net HWBot Team › [Updated 31 Jan 13] SystemCompute v0.5.7.2
New Posts  All Forums:Forum Nav:

[Updated 31 Jan 13] SystemCompute v0.5.7.2 - Page 6

post #51 of 142
Quote:
Originally Posted by borandi View Post

Methinks two reasons:
a) AMD single core performance looks like it really hits hard
b) I think PCIe lanes are a big factor here - if you're running CF and it's restricting the card to PCIe x8, then there's a chance that the drop in speed is affecting the score.

But thankyou for your data point, much appreciated!

So I re-ran it with CF disabled, not much change. But some more data for you!

CPU:FX 4100 @ 4.515Ghz
GPU: HD6870 @ 925/1100
RAM: 16GB @ 2006.7Mhz 9-9-9-24-1T

CPU Score = 46401
AMP Score = 43562
Total Score = 43695

Ryzen 912
(16 items)
 
Black n Blue
(33 items)
 
Blue Beetle
(21 items)
 
  hide details  
Reply
Ryzen 912
(16 items)
 
Black n Blue
(33 items)
 
Blue Beetle
(21 items)
 
  hide details  
Reply
post #52 of 142


i7 920 @ 3508MHz
CPU:174033
GTX 460 @ 850 core. Stock memory.
GPU:71252
Total:73420
Ram: 6GB CL8 @ 1336Mhz

Without hyperthreading the CPU score is 150k.

I'm probably going to be the only one posting the results of a four year old chip in this thread tongue.gif. Its fun to see how much processors have improved since then.
Edited by Derp - 11/21/12 at 12:27pm
post #53 of 142
CPU:FX 8120@ 4.013Ghz
GPU: GT440 @Stoc
RAM: 6GB @ 1600Mhz cl 9
CPU Score = 81004
AMP Score = 35881
Total Score = 36909

Evolved
(25 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 5820K X99A RAIDER (MS-7885) NVIDIA GeForce GTX 980 Ti asus Poseidon Corsair  
RAMRAMRAMHard Drive
Corsair  Corsair  PNY Electronics  PCI revodrive 3 
Hard DriveOptical DriveCoolingCooling
ssd samsung  LG BLu-ray. XPC water block XPC 360 Radiator 
CoolingCoolingOSMonitor
XPC 240 Radiator Swifttech Pumb Windows 10 PHillips 40'' 4k Monitor 
KeyboardPowerCaseMouse
Corsair k70 Mecha AX 1500i Corsair 900D Corsair m65, Razer naga, Logitech g700 
Mouse PadAudioOtherOther
steel series mousepad Soundblaster Z, Yamaha rx-v677 Infinity Speakers X 7, Jbl Subwoofer, vanther &... Audeze LCD-2 
Other
FIIO K5 headphone AMP 
  hide details  
Reply
Evolved
(25 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 5820K X99A RAIDER (MS-7885) NVIDIA GeForce GTX 980 Ti asus Poseidon Corsair  
RAMRAMRAMHard Drive
Corsair  Corsair  PNY Electronics  PCI revodrive 3 
Hard DriveOptical DriveCoolingCooling
ssd samsung  LG BLu-ray. XPC water block XPC 360 Radiator 
CoolingCoolingOSMonitor
XPC 240 Radiator Swifttech Pumb Windows 10 PHillips 40'' 4k Monitor 
KeyboardPowerCaseMouse
Corsair k70 Mecha AX 1500i Corsair 900D Corsair m65, Razer naga, Logitech g700 
Mouse PadAudioOtherOther
steel series mousepad Soundblaster Z, Yamaha rx-v677 Infinity Speakers X 7, Jbl Subwoofer, vanther &... Audeze LCD-2 
Other
FIIO K5 headphone AMP 
  hide details  
Reply
post #54 of 142
Thread Starter 
Quote:
Originally Posted by GENiEBEN View Post

Thanks, I wanted to make it calculate CPU & AMP subscores too but I got lost in your crazy algo sad-smiley-002.gif
Here is what I could figure out of it, tho I got lost when I couldn't figure who v48 is: Warning: Spoiler! (Click to show)
Code:
v11 = SUBTEST1
v16 = SUBTEST2
v21 = SUBTEST3
v29 = SUBTEST4
v35 = SUBTEST5
v39 = SUBTEST6
v52 = CPU SCORE
v51 = AMP SCORE
Total score = 1.0 / (0.949999988079071 / v52 + 0.05000000074505806 / v52)
----------------------------------
v50 = 8.388608 / (v50 / 1000.0) * (sub_403470(v50))
v52 = 0.524288 / ((v50 + 1) / 1000.0) * (sub_401CA0(v50 + 4))
v54 = 0.131072 / (v51) * (sub_4019F0(v51 + 4))
v53 = 0.131072 / (v48 / 1000.0) * (sub_402F30(v48))

CPU = ((v52 * 0.33 + v53 * 0.25 + v54 + 4) * 0.45) * 1000.0
AMP = ((v51 * 0.37 + v53[8] * 0.26 + v50 + 4) * 0.37) * 100.0


That calculation looks vile, nothing like what I've coded biggrin.gif The Total Score one is right, but the rest look bad. I didn't use any of the sub_40xxxx functions you list. Each subtest has a simple weighting for the CPU and AMP scores.
Quote:
Originally Posted by GENiEBEN View Post

Quote:
Originally Posted by borandi View Post

Hopefully the CPU size loves cores and HT.  The GPU side loves single core speed though - preferably BCLK over multiplier.  At least that's what I'm seeing, but I think there may be an issue on the C++ AMP coding beyond what I've done.  I'll try and rework it for v0.5.

Don't worry, so far works great wink.gif

PS; As a (personal) favor can you make an alternate compile with MinGW (+/- cpu optimisations) for the sake of comparison? (Include AMP+ locally as MinGW might NOT have it installed). Sorry if it's too much work.

Cheers!

EDIT: Now you made me curios about AMP, started some basic functions and so far I noticed that compiling with /fp:fast favors the AMP code quite a bit, you should give it a try (ofc if the computation isn't very dependent on the values returned lol).
[/quote]
C++ AMP is a part of the VS2012 IDE. I come from a coding background of just using visual studio, not worrying about linking or anything major, so I don't really know what MinGW is or what to do with it :/ Sorry

I'll try with the AMP flag, though I think most of my matrix calcs are doubles for a reason. I'll see what happens when I degrade them all to floats (probably boost the scores quite a lot).
Quote:
Originally Posted by Majorhi View Post

Quote:
Originally Posted by borandi View Post

Methinks two reasons:
a) AMD single core performance looks like it really hits hard
b) I think PCIe lanes are a big factor here - if you're running CF and it's restricting the card to PCIe x8, then there's a chance that the drop in speed is affecting the score.

But thankyou for your data point, much appreciated!

So I re-ran it with CF disabled, not much change. But some more data for you!

Thanks for the score! But even if you disable CF, the presence of the second card installed still drops the PCIe lane to PCIe 2.0 x8, as it says in GPU-Z. To get a better score, you'd actually have to physically remove the card, boot up, and run. It's a limitation I'm trying to get around in code atm, and it would probably boost scores quite a lot (!).

It'd be interesting to see the difference of PCIe 2.0 vs. PCIe 3.0 if there is any if anyone has an decent Ivy system on hand and a 7-series card.

Also thank you to everyone else who has given scores smile.gif
post #55 of 142
Score for you, going for the low score title.

AMD A4-3400 and AMD HD 6670, all stock! guiltysmiley.gif

    
CPUMotherboardGraphicsRAM
Intel Celeron G540 Asus Maximus Gene IV Sapphire Radeon HD 4890 G.Skill DDR3 PC3-10666 
Hard DriveOptical DriveOSMonitor
Seagate Barracuda 1.5TB LG 22x SuperMulti Windows 7 Ultimate LG 23" W2353V 
KeyboardPowerCaseMouse
Ducky OCN Mechanical  Seasonic 550W Thermaltake Lanbox Lite Razer Diamondback 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel Celeron G540 Asus Maximus Gene IV Sapphire Radeon HD 4890 G.Skill DDR3 PC3-10666 
Hard DriveOptical DriveOSMonitor
Seagate Barracuda 1.5TB LG 22x SuperMulti Windows 7 Ultimate LG 23" W2353V 
KeyboardPowerCaseMouse
Ducky OCN Mechanical  Seasonic 550W Thermaltake Lanbox Lite Razer Diamondback 
  hide details  
Reply
post #56 of 142
Thread Starter 
Quote:
Originally Posted by Pavix View Post

all stock. What are your plans for future releases? Nice work btw thumb.gif

Thanks smile.gif I have a few things on the list:

a) Make it multi GPU aware, but you have to use the same GPUs (no mix/match or hybrid)
b) PCIe speed seems to be a big limitation, as well as single core speed. I think this is linked to some read/write segments of the code I need to optimise/adjust.
c) Add in an N-body type simulation
d) Verifiable output + appropriate website to deal with uploaded scores.

Scores will probably not be comparable across versions. But there'll be enough improvement between released versions rather than ditching scoreboard xyz and starting over. I ran this one across as many configs as I had setup at the time before release just to make sure it wasn't FUBARed smile.gif New versions will be similarly tested until we're all happy biggrin.gif
post #57 of 142


thumb.gif
post #58 of 142
We are using nvidia N-body simulation in a loop and GPU linpack to stress our tesla cards in linux thumb.gif

n-body simulation in graphics mode on multiple cards



More pix in my gallery biggrin.gif
Edited by dhenzjhen - 11/21/12 at 4:43pm
Athena
(14 items)
 
Achilles
(9 items)
 
 
CPUMotherboardGraphicsHard Drive
[Intel] i5 2500k  [ASUS] p8p67 Evo B3 [EVGA] GTX 680 [Samsung] 1TB 
CoolingOSMonitorKeyboard
[Noctua] NH-D14 [Micro$oft] Win7 64-bit/Linux [HP] 2311x [Logitech] G510 
PowerCaseMouse
[XFX] 650W [Cooler Master] HAF X [Logitech] G400 
CPUMotherboardGraphicsRAM
990X 6.5Ghz EVGA X58 3 way SLI classified XFX GT 240/modded 8800GTS/8800GTX/5870/4980/FIR... Corsair Dominator PSC  
CoolingOSMonitorCase
F1 Extreme Dark, DICE and LN2!! XPeeeee and Win 7 Old CRT Open 
  hide details  
Reply
Athena
(14 items)
 
Achilles
(9 items)
 
 
CPUMotherboardGraphicsHard Drive
[Intel] i5 2500k  [ASUS] p8p67 Evo B3 [EVGA] GTX 680 [Samsung] 1TB 
CoolingOSMonitorKeyboard
[Noctua] NH-D14 [Micro$oft] Win7 64-bit/Linux [HP] 2311x [Logitech] G510 
PowerCaseMouse
[XFX] 650W [Cooler Master] HAF X [Logitech] G400 
CPUMotherboardGraphicsRAM
990X 6.5Ghz EVGA X58 3 way SLI classified XFX GT 240/modded 8800GTS/8800GTX/5870/4980/FIR... Corsair Dominator PSC  
CoolingOSMonitorCase
F1 Extreme Dark, DICE and LN2!! XPeeeee and Win 7 Old CRT Open 
  hide details  
Reply
post #59 of 142
Thread Starter 
N-body seems like a good way to go. I have some C++ AMP Source code with a GUI to do it, I just need to strip out the GUI then port it to my IDE making sure I have all the right dependencies thumb.gif That was stressful enough with libcpuid so I hope it goes smooth.
post #60 of 142
My $1200 alienware laptop from 2009....sigh, sucks so bad. Oh, running win8

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Overclock.net HWBot Team
Overclock.net › Forums › Benchmarks › Overclock.net HWBot Team › [Updated 31 Jan 13] SystemCompute v0.5.7.2