Overclock.net banner

1 - 20 of 44 Posts

·
Registered
Joined
·
3,467 Posts
Discussion Starter #1
Foreword: The intent of this test is not for microarchitecture to microarchitecture, So please refrain for any fan boy isim from either side, I do not wish this effort do be invalided and derailed by such discussions. That being said, If you wish to contribute with different chips and uArchs please adhere to my methodology so that results will be as comparable as possible. Please keep in mind though, the mind space on this test is to Maximize what you might own and i just happen to own this FX processor. This is NOT a what you should buy thread. My Intent is to help other FX users because that is what I am limited too.

I am testing using items that I possess. If you have to ask why isn't X aspect tested, the Answer is.... It either will not be tested due to either not owning required items or the set up requires an additionally prohibitive amount of set up time
(IE switching out GPU for a GPU in stock configurations. as it is being used and the other card will likely not fit into the second case without a day worth of re organizing)
Or I am already testing it, and I just haven't finished with those results yet.
As will be described in my methodology in greater extent, i did my best effort to normalize my attainable setting into something tangible so that they are easier to understand and compare.

Disclaimer: I am by NO means what so ever, an academic. There might be errors and inefficiencies in my method but this is the best that occurred to me to do.

Test Hardware: My rig Beast Schlampe
CPU: AMD FX 8370-e

COOLING: Deepcool Captain 360 push-pull

MB: ASUS Crosshair V Formula Z

Ram:G.Skill 2400mhz TridentX 10-12-12-31-43

GPU: GTX 780 ti (CUSTOM BIOS 324w TDP 1267mhz core 1850mhz (7400mhz) Ram Boost disabled)

COOLING: NZXT G10 + Corsair H110 in push config

PSU: Seasonic X-850 Gold

OS: Windows 8.1 pro 64-bit

Methodology:

So my baseline is as follows. I used the base frequency with boost disabled and all power savings features disabled.
the base defaults for my set up are as followed

Cpu. 3.3 ghz, Northbridge 2400mhz HyperTransport 2600mhz.

The memory speed and timings that were used were the STOCK timings in the DDR3 Gskill TridentX line that my set came from, So 1600mhz-2400mhz (beyond is silly for AMD but is not unheard of, i've not gotten that far tweaking this chip)

1600mhz 1.5v 7-8-8-24-31
1866mhz 1.6v 8-9-9-24-33
2133mhz 1.6v 9-11-11-31-42
2400mhz 1.65v 10-12-12-31-43

other speeds and timings I intend on trying but no idea weather my kit can do it.

Ripjaws 2000mhz 1.65v 9-10-9-28-37 (this timing will require me to used a non 200fsb overclock to achieve and up until now my cpu hasn't been to fond of doing such things)

Going forward past this base line I intend on testing notable NB and HT frequency combinations (2600/2600 +2500*/2700*+2700*/2900+*)

*= IF ATTAINABLE IN A STABLE MANNER.

Notable Core frequencies intended on testing. 3.3GHZ, 3.5GHZ, 3.7GHZ, 4GHZ, 4.3GHZ, 4.6GHZ, 5.0GHZ and what ever FSB setting my chip will let me get away with.

All voltages in Bios are set to allow slight Vdroop, certain voltages have been slightly bumped up as to maintain VID within Windows.

GPU INFO: I gimped this thing as best i could, I know its a Very highly overclocked sample as is, but i could be arsed to flash the stock bios back (not to mention i'd have to go look for it on external back ups)

This bios Idles at 927mhz core 1850mhz (7400mhz) memory @ .95v Underload this bios does 1267mhz core 1850mhz (7400mhz) memory @ 1,212v

it does a fair job at emulating a mildly overclocked 980 and a low stock 980 ti score. so its representative of the approximate score that can be expected from the upper end of current Maxwell cards

EVGA precision was used to enable the 108% maximum power range allocated within the cards bios (325w) however as of up until the end of 1866 speed testing only every used 104.6% of its 108% max (not even sure if this is useful info or not)

LACK OF GRAPHS: They will come, but i am choosing not to do them until i've finished the testing and have all my data. you are free to make your own and contribute. they will be linked AND thanked within the OP even once i've done my own graphs

SAMPLE SIZE: 5 runs of each test are being done per configuration then averages reported. (i may get around to linking results within 3dmark but no promises, there will be ALOT)

WHAT THE GABE.N Does this API Test DO? : It tests your CPU's ability to manage API draw calls in GAMES

how it does this is by starting your CPU at a rather low amount of draw calls. so you start with a VERY VERY high FPS, this doesn't matter however due to how the benchmark finishes.

The benchmark has a staggered stop point. The draw call amount will stop once the bench mark reaches 30fps. however this is not what the score tells you.
the score calculation begins at that point. i am still a little gray on but this is what i'm going with.

the score is indicative of the amount of draw calls received in the Second in time succeeding the point which the bench mark gets lower than 30 fps. while the Draw call amount you see in the benchmark is the draw call per frame.

why i didn't do fps testing on this: i don't see how its relevant, every benchmark ends at the same FPS and its score is bearing on the draw calls per second rather than the frame rate per second. the beginning of the test is irrelevant for FPS because no game beside 2D side scrolled will have that low of draw calls per frame, and at the lowest fps range I don't think Assassins creed even got remotely close to that range of draw calls. but then again i don't know for sure due to know owning that game.

RESULTS: 3.3ghz 2400mhz 2600mhz 1600mhz 7-8-8-24-31 1.5v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,253,253

DX11 Multi Threaded score: 2,169,369.5

FireStrike

Score: 8434

Graphics: 13786

Physics: 6445

Combined: 2446

RESULTS: 3.3ghz 2400mhz 2600mhz 1866mhz 8-9-9-24-33 1.6v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,255,925

DX11 Multi Threaded score: 2,239,916.5

FireStrike

Score: 8455

Graphics: 13715

Physics: 6579.5

Combined: 2452.5

RESULTS: 3.3ghz 2400mhz 2600mhz 2133mhz 9-11-11-31-42 1.6v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,296,267

DX11 Multi Threaded score: 2,290,991

FireStrike

Score: 8465

Graphics: 13809.5

Physics: 6506

Combined: 2454

RESULTS: 3.3ghz 2400mhz 2600mhz 2400mhz 10-12-12-31-43 1.65v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,320,251

DX11 Multi Threaded score: 2,342,068.5

FireStrike

Score: 8423

Graphics: 13698

Physics: 6503

Combined: 2445.5

3.3ghz 2600mhz 2600mhz 1600mhz 7-8-8-24-31 1.5v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,271,407.5

DX11 Multi Threaded score: 2,208,695.5

FireStrike

Score: 8616

Graphics: 13829.5

Physics: 6612

Combined: 2557

RESULTS: 3.3ghz 2600mhz 2600mhz 1866mhz 8-9-9-24-33 1.6v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,296,521

DX11 Multi Threaded score: 2,250,445.5

FireStrike

Score: 8592

Graphics: 13819.5

Physics: 6568.5

Combined: 2547

RESULTS: 3.3ghz 2600mhz 2600mhz 2133mhz 9-11-11-31-42 1.6v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,299,535

DX11 Multi Threaded score: 2,283,125.5
FireStrike

Score: 8612

Graphics: 13725

Physics: 6582

Combined: 2586

RESULTS: 3.3ghz 2600mhz 2600mhz 2400mhz 10-12-12-31-43 1.65v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,341,462

DX11 Multi Threaded score: 2,370,347

FireStrike

Score: 8571

Graphics: 13711.5

Physics: 6561

Combined: 2557.5

Updates and more scores will be posted in the current thread conversation as well as This Post will also be updated as I finish portions of the test,







 

·
Premium Member
Joined
·
10,167 Posts
Should be interesting - Sub a dub dub
thumb.gif
 

·
Registered
Joined
·
4,392 Posts

·
Registered
Joined
·
3,467 Posts
Discussion Starter #10
3.3ghz 2600mhz 2600mhz 1600mhz 7-8-8-24-31 1.5v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,271,407.5

DX11 Multi Threaded score: 2,208,695.5

FireStrike

Score: 8616

Graphics: 13829.5

Physics: 6612

Combined: 2557

RESULTS: 3.3ghz 2600mhz 2600mhz 1866mhz 8-9-9-24-33 1.6v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,296,521

DX11 Multi Threaded score: 2,250,445.5

FireStrike

Score: 8592

Graphics: 13819.5

Physics: 6568.5

Combined: 2547

RESULTS: 3.3ghz 2600mhz 2600mhz 2133mhz 9-11-11-31-42 1.6v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,299,535

DX11 Multi Threaded score: 2,283,125.5
FireStrike

Score: 8612

Graphics: 13725

Physics: 6582

Combined: 2586

RESULTS: 3.3ghz 2600mhz 2600mhz 2400mhz 10-12-12-31-43 1.65v

DX 11 API overhead test. AVERAGE.

DX11 Single Threaded score : 1,341,462

DX11 Multi Threaded score: 2,370,347

FireStrike

Score: 8571

Graphics: 13711.5

Physics: 6561

Combined: 2557.5

seems like the API scales better on low clock speed. This is enough for the day, I will pick up again tomorrow.
 

·
Premium Member
Joined
·
10,167 Posts
Random test at 5ghz 2400 ht 2600nb 2600 mhz cl 12 T1
 

·
Premium Member
Joined
·
10,167 Posts
Quote:
Originally Posted by FlailScHLAMP View Post

Quote:
Originally Posted by cssorkinman View Post

Random test at 5ghz 2400 ht 2600nb 2600 mhz cl 12 T1
Thanks man!.. what resolution did you run it at?

i just realized i've been running my tests @ 1440p.
rolleyes.gif
I just ran the API test on that particular run http://www.3dmark.com/3dm/10147888

Here is the subsequent full Firestrike run http://www.3dmark.com/fs/7107263
Stock 780ti

EDIT: 2600ht 2600nb 2600 brought the physics score up about 200 points over the 2400 HT - didn't expect that.
http://www.3dmark.com/fs/7117148
 

·
Registered
Joined
·
3,467 Posts
Discussion Starter #15
so before I jump back into testing today (i've finished up to 4.3ghz 1866mhz mem) I thought i should post the averages I've finished with.

API:
ST: 1,350,320
MT: 2,343,160.5

FireStrike
Score: 8980.5
Graphics: 13814.5
Physics: 7384
Combined: 2722.5

API:
ST: 1,348,545
MT: 2,363,380.5

FireStrike
Score: 8937
Graphics: 13841.5
Physics: 7301
Combined: 2692

API:
ST: 1,349,926
MT: 2,379,110

FireStrike
Score: 8983.5
Graphics: 13812
Physics: 7380
Combined: 2726.5

API:
ST: 1,384,819.5
MT: 2,403,728

FireStrike
Score: 8896.5
Graphics: 13708
Physics: 7393
Combined: 2674

API:
ST: 1,463,362.5
MT: 2,454,690

FireStrike
Score: 8938.5
Graphics: 13806
Physics: 7316.5
Combined: 2700

API:
ST: 1,671,821
MT: 2,481,663.5

FireStrike
Score: 8918
Graphics: 13818
Physics: 7337.5
Combined: 2673.5

scaling on the api appears to be pretty smooth across all speeds, while its fairly evident that FS wants a minimum clock core speed and the majority of its score is dependant on that.

5ghz testing today should show a better comparison
 

·
Registered
Joined
·
3,467 Posts
Discussion Starter #16
Quote:
Originally Posted by cssorkinman View Post

Quote:
Originally Posted by FlailScHLAMP View Post

Quote:
Originally Posted by cssorkinman View Post

Random test at 5ghz 2400 ht 2600nb 2600 mhz cl 12 T1
Thanks man!.. what resolution did you run it at?

i just realized i've been running my tests @ 1440p.
rolleyes.gif
I just ran the API test on that particular run http://www.3dmark.com/3dm/10147888

Here is the subsequent full Firestrike run http://www.3dmark.com/fs/7107263
Stock 780ti

EDIT: 2600ht 2600nb 2600 brought the physics score up about 200 points over the 2400 HT - didn't expect that.
http://www.3dmark.com/fs/7117148
biggrin.gif
well i've never played in the 2600mhz range for ram, maybe i should try it with this chip.

but watching all these bench marks happen, the multi threaded gains are VERY VERY linear in a very gradual slope as to my HT not differing test to test, single threaded performance seems to be boosted by NB frequency and it only shows gains at certain FSB(ht):ram ratio. while your initial runs in the 83XX thread show multi-threaded gains from a 200mhz bump from 2400ht to 2600ht

very interesting observations so far.
 

·
Registered
Joined
·
3,467 Posts
Discussion Starter #18
So looking at these graphs i'm not going to include the Graphics score any more for the reasons that follow.

scores from 3.3ghz -5ghz are bloody stable.they don't change outside the variance from run to run.

I think I can safely say without any arguable doubt that a FX8 regardless of its base clock can drive a high end graphics card. end of petty argument that should have never existed.

@CSSorikinman Still have that Devils Canyon rig? i'm curious if there is a difference in graphics score with the same card and same gpu clocks and in your FX system.
 

·
Premium Member
Joined
·
10,167 Posts
Quote:
Originally Posted by FlailScHLAMP View Post

So looking at these graphs i'm not going to include the Graphics score any more for the reasons that follow.

scores from 3.3ghz -5ghz are bloody stable.they don't change outside the variance from run to run.

I think I can safely say without any arguable doubt that a FX8 regardless of its base clock can drive a high end graphics card. end of petty argument that should have never existed.

@CSSorikinman Still have that Devils Canyon rig? i'm curious if there is a difference in graphics score with the same card and same gpu clocks and in your FX system.
Yes, but the 780ti is back in the FX rig, might have a chance to try it soon though. My cpu temps aren't what they should be on the FX rig - thinking about tearing it down again.
 
1 - 20 of 44 Posts
Top