FX + Crossfire HD5770, warm ups - 32M, Heaven, 06, Vantage, 11 results on water/air

[BeepBeep2]

Warming up for DICE...still need to run a few of these to finish up both single card/dual card. Due to the amount of results I've left out screenshots. Specific GPU / CPU / DDR3 info is on each respective HWBOT page along with screenshot or verification link for the results.

The 3DMark06 result is so poor due to GPU performance under FX, just under 22K at the highest speed I can run the CPU test, Phenom II X6 does a consistent 24K. I hit almost 27K with the Phenom II X6 using cold air a long time ago.

Test setup:
AMD FX-8150 8-Core Processor @ 5.01 to 5.3 GHz
Dual Sapphire Vapor-X AMD HD5770 Cat 11.11
4GB GSkill Pi Series DDR3-2133 CL7 @ 2000 6-9-6-24 1T
Seasonic M12 based PSU, 850w 80Plus Silver
Win7 x64 Pro

SuperPi 32M -

15m 25.628s - 1st Place AMD VISION (32nm) non-subzero cooling
CPU Speed - 5281 MHz / 2501 MHz,
DDR3 Speed - 2223 MHz 6-10-6-24-36 110ns 1T
http://www.hwbot.org/submission/2265581_

3DMark06 -

2x HD5770 - 21984 3DMarks #105th Place <---this is how weak FX is in single thread / CPU limited 3D.
CPU Speed - 5017 MHz / 2508 MHz, DDR3-2006 6-9-6-24 1T
GPU Speed - 1020 / 1350
http://www.hwbot.org/submission/2265589_

3DMark Vantage Performance Preset -

2x HD5770 - 22266 3DMarks #28th Place
CPU Speed - 5017 MHz / 2508 MHz, DDR3-2006 6-9-6-24 1T
GPU Speed - 1020 / 1350
http://www.hwbot.org/submission/2265585_

3DMark11 Performance Preset -

1x HD5770 - 3365 3DMarks #34th Place
CPU Speed - 5017 MHz / 2508 MHz, DDR3-2006 6-9-6-24 1T
GPU Speed - 1015 / 1375
http://www.hwbot.org/submission/2265579_

Unigine Heaven HWBOT Benchmark Basic Preset (DX9) -

1x HD 5770 - 1511.63 DX9 Marks #17th Place
CPU Speed - 5077 MHz / 2538 MHz, DDR3-2030 6-9-6-24 1T
GPU Speed - 1015 / 1375
http://www.hwbot.org/submission/2265574_

2x HD 5770 - 2823.05 DX9 Marks #1 Place
CPU Speed - 5077 MHz / 2538 MHz, DDR3-2030 6-9-6-24 1T
GPU Speed - 1005 / 1370
http://www.hwbot.org/submission/2265569_

Unigine Heaven HWBOT Benchmark Xtreme Preset (DX11) -

1x HD 5770 - 568.72 DX11 Marks #31 Place
CPU Speed - 5077 MHz / 2538 MHz, DDR3-2006 6-9-6-24 1T
GPU Speed - 1015 / 1375
http://www.hwbot.org/submission/2265576_

2x HD 5770 - 1123.02 DX11 Marks #7 Place
CPU Speed - 5077 MHz / 2538 MHz, DDR3-2030 6-9-6-24 1T
GPU Speed - 1010 / 1360
http://www.hwbot.org/submission/2265564_

[blossa]

Thanks for sharing. I was actually thinking about to test some 3D with FX, so it is great to have some numbers to compare with. I will probably run with slower mems but hopefully a bit higher NB.

[Hondacity]

team.xs ftw

[demonkevy666]

My bulldozer only mange to beat my Thuban from 4.0ghz with 3.0ghz Nb with 4.8ghz that's about 20% higher clocks and only won by 1% :O FAILDOZER!
although in price I paid $320 v for the Thuban and $220 the bulldozer.
I only bought the chip to mess with.

[BeepBeep2]

My bulldozer only mange to beat my Thuban from 4.0ghz with 3.0ghz Nb with 4.8ghz that's about 20% higher clocks and only won by 1% :O FAILDOZER!
although in price I paid $320 v for the Thuban and $220 the bulldozer.
I only bought the chip to mess with.

Apparently I was told by a BD fanboy that 3.7 is a reasonable number for Thuban and 4.8 is reasonable for BD overclocks.
Thuban was competitive when it came out. BD isn't. I payed $275 for my FX-8150, and they should be priced around $190 with the 8120 at $170.

Performance per clock (I got major trolled by BD fanboy seronx on TPU because apparently "IPC increased on BD" and I'm a dumbass) is 20-25% less on Bulldozer than Phenom II.
That means:
4.6-4.7 GHz BD = 3.7-4.0 GHz Phenom II X6 depending on the program running. Power is awful too.

Of course someone will throw up a graph of Llano and FX-4100, and show that power consumption is 5% higher on Llano than BD at stock. However the Llano APUs can be undervolted 30% while the BD CPUs undervolt 5-10%.

PD should be what BD should have been. It took a thousand people talking about how bad BD sucked for AMD to push more into PD's design though

[Solid Snake]

WOW, great scores.

Xtremesystems Team - FTW !!!

[BeepBeep2]

I'll throw this out there too.

DDR3-2661 @ 2400 divider

http://www.hwbot.org/submission/2266...ram_1330.4_mhz #86th Place DDR3 memory clock ...just behind InfraRed @ 81st place and 2688

My CPUs IMC is really weak though, this was done on slot 2, slot 3 doesn't even do 2400.

[I.nfraR.ed]

Something similar here, but with BBSE. OS is tuned for SuperPi, no wazza.
The cpu has a weak IMC, these stick were verified to run 1130+. Cores are not very good as well.
Still waiting for my FX-8150 replacement. This was the first time ever I run 32M on Bulldozer. Seems not many people care about running it, lol.

FX-8120 @ 5235
G.Skill Trident @ 1110 7-9-6-20-28 1T 90ns
15m 26.734s

http://i.imgur.com/QbOLG.png

[BeepBeep2]

That's really impressive efficiency, especially without waza.
My OS isn't really ideal...I just tried to bench everything I could. I couldn't cut ram usage enough on Win 7 to make maxmem useful. One of my main drives in my RAID array is failing too which is causing some issues.

By the way, I found that on this board the 1600 divider was the fastest compared to 1866/2133. Not sure if that applies to others, but I thought I'd share.

[blossa]

2133 divider (on Gigabyte FXA990-UD5) worked best for me in SP32M, but I guess it might be because of my crappy sticks. With CAS8 and around 2100MHz i got 15m05s with water cooling. I didn't save it though, as I was eager to get below 15m. :S

[BeepBeep2]

2133 divider (on Gigabyte FXA990-UD5) worked best for me in SP32M, but I guess it might be because of my crappy sticks. With CAS8 and around 2100MHz i got 15m05s with water cooling. I didn't save it though, as I was eager to get below 15m. :S

What CPU frequency?

[blossa]

54xxMHz something. Decent NB but bad mem-clock. 1.58v vcore if i remember correctly...

[BeepBeep2]

I wanted to update two scores...

3DMark Vantage Performance Preset -
2x HD5770 - 23000 3DMarks #19th Place
CPU Speed - 5277 MHz / 2638 MHz, DDR3-2110 7-10-7-27 1T
GPU Speed - 1020 / 1365
http://www.hwbot.org/submission/2267185_

Up from 27th Place. I was really surprised with CPU test scaling (26.8k vs 25.2k)...I ran it 3 times and each result was about the same. 5277 MHz for GPU tests, 5176 for CPU test vs 5017 before..

SuperPi 32M -
15m 6.112s - 1st Place HWBOT AMD VISION (32nm) non-subzero cooling
CPU Speed - 5458 MHz / 2729 MHz,
DDR3 Speed - 2182 MHz 6-10-6-24-34 90ns 1T
http://www.hwbot.org/submission/2267...min_6sec_112ms

The story behind these results are simply ~20lbs of icy water and a large bucket. The SuperPi run though isn't going to break any records for efficiency.

[blossa]

Nice! Soon sub 15m.

[BeepBeep2]

Nice! Soon sub 15m.

Yeah, but my temperatures were 20c lower than yours if you were using normal water cooling

[BeepBeep2]

Single card 5770...

To get an idea of how poor this is, Flanker got 18284 3DMarks by using a 965 BE @ 4.064 GHz and 5770 clocked at 915/1350. His CPU score was just under 6K.
http://img716.imageshack.us/img716/2...leneck5770.jpg

[FlanK3r]

Old benchmarks are not idea for FX CPU :-/...Did u used XP or win7? In XP u can get higher scores of 3D Marks (without Vantage and 3D11)

[BeepBeep2]

Old benchmarks are not idea for FX CPU :-/...Did u used XP or win7? In XP u can get higher scores of 3D Marks (without Vantage and 3D11)

I know you can get higher 3DMark06 with XP but all I have are Vista 32 bit and Win7 here right now.
I just think it is kind of ridiculous, i7 2500K at 4.8 in 3DMark06 would get about 20000 3DMark with single card.

My 1100T is back in.

First score was 19043 with the X6 @ X4 4.3 GHz

[FlanK3r]

but example in 3d11 is fx good, try it ....

[BeepBeep2]

but example in 3d11 is fx good, try it ....

I already did?

[BeepBeep2]

I think I found my OS for 3DMark06, was having trouble with 05 and 03, 05 cut to desktop at CPU test loading and 03 froze at the exact same frame 3 times. I also wanted to note that single card with Phenom II X6 at sub ambient has the potential to beat out FX in crossfire mode:
http://img14.imageshack.us/img14/556...0354single.png

20354 with GPU @ 1040/1360, CPU was at 438x for 3D and 4354 for CPU test. I had hypers installed and was too lazy to swap to PSC to run @ 2050 6-9-6/2200 7-10-7. I doubt it would have helped anyway.

I was able to hit 25959 too with dual card and CPU @ 4370. I bet 30K+ is doable with CPU under DICE or colder.


Also tried 32M once again with Phenom II, forgot to save screenshot but result was 14m 53.74x seconds off the top of my head. A solid 2-3 seconds off my M4A89TD but I didn't run waza this time. Secondary timings were more aggressive, but same clocks as before. I think after waza, 14.4x should be possible under 4.5 GHz.

[BeepBeep2]

Tomorrow

Goals for victim(s):
1500/3000 MHz+ DDR3 valid
6.4/7.5 GHz CPU-Z valid
PiFast, SPi 32M, 1M, wPrime 32M/1024M at ~6/7 GHz
3D runs with 5770 both single card and crossfire

Slayer (missing due to 24/7 use ...and hopefully not slaying anything!)


Victim

[BeepBeep2]

Okay, a lot more to come from this. CPU is probably a 6.5 GHz chip, I just barely squeaked out 6.46. Maybe if I can get everything working perfect tomorrow I'll have time to take another stab at CPU-Z.

BTW, this will get its own thread when the time has come.

http://img855.imageshack.us/img855/1...singleblur.jpg

[Barr3l Rid3r]

[BeepBeep2]

Okay?

CPU has almost nothing to do with 3DMark11, which is why AMD FX does so great...

From water > LN2 I gained 100 pts total score.