[56k warning] 1055T Thuban benchmarked to hell and back

Mothergoose729

Senior member
Mar 21, 2009
409
2
81
A few days ago I got the new thuban processor from AMD, a hexa core k10.5 arch with 6mb L3 cache and 3mb combined L2 cache. The break down from AMD's website:


Processor AMD Phenom™ II X6
Model 1055T
OPN Tray HDT55TFBK6DGR
OPN PIB HDT55TFBGRBOX
Operating Mode 32 Bit Yes
Operating Mode 64 Bit Yes
Revision E0
Core Speed (MHz) 2800
Voltages 1.125-1.40V
Max Temps (C) 62
Wattage 125 W
Virtualization Yes
L1 Cache Size (KB) 128
L1 Cache Count 6
L2 Cache Size (KB) 512
L2 Cache Count 6
L3 Cache Size (KB) 6144
CMOS 45nm SOI
Socket AM3
AMD Business Class No
Black Edition No
http://products.amd.com/en-us/DesktopCPUDetail.aspx?id=641


This is AMD's first mainstream hexa core (6 core) processor. It is AM3 and AM2+ compatable, features new technologies like AMD Turbo Core, and is competitively priced at this time at ~199$.

Overclocking

I used an original Thermalright 120 (TRUE) with a lapped base for cooling. A scythic slipstream fan rated for 110cf was used as intake, and a generic coolmaster fan rated at ~35cfm was used in "pull" or exhaust. Case was an open antec 900 with two front intake fans set to low, top 200mm exhaust fan turned off.

I overclocked my unit to 3.2ghz with 1.25v on my Gigabyte 785g motherboard. OC'ing this processor was very interesting, it was more then willing to run under volted and overclocked with some incredible efficiency. Prime 95 ran stable for more then 9 hours at 3.2ghz with 1.168v under load. However, clock speeds as high as 3.8ghz were very difficult to get anywhere near stable, while a small voltage bump to 1.4v was all it took to get 3.7ghz mostly use able. The feeling I got is that this processor is very heat sensitive and would shine exceptionally well at lower voltages under water cooling. Another peculiarity; the heat sensor on my processor under reported my temperatures by at least 5c. I think it is more accurate under load, but at idle it reported 15c, a about 10c below ambient temperatures!

Turbo core is another interesting thing to play with. In the bios you can adjust how far individual cores will "turbo up", allowing you to adjust the multiplier up to 16.5 for "turbo cores" from the locked 14 for the whole processor. Using AMD's oc tool AMD OVERDRIVE, it is possible to control not only how far the turbo cores up clock but also how many turbo cores (default 3, maximum is 5) will clock up under light load. I chose to disable my turbo cores, because it made testing for stability more difficult and it caused the voltage for the processor to fluctuate more widely. For the person up to the challenge, it can provide the adamant overclocker more control over the performance of their system.

CPU-Z, AMD overdrive, and task manager screenshot

Capture-2.png



*A note, this CPU-Z shows my vcore at 1.168v, which was enough to get prime95 stable. However, I noticed when running my benchmarks that this v core was not enough. Cinebench R10 would not run without 1.2v and sisoft sandra's arithmetic test would occasional fail at anything below 1.225v. This pretty much goes for all processors, but on this one especially use a variety of different tests to determine a systems overall stability. Most benchmarks were run with a v core of 1.225. Clock speeds and system bus speeds where never changed.

The benchmarks:
I have the scores listed by software bench suite. Screenshots and description of the benchmark are spoiled to conserve space.

My system specs included:

1055T @ 3.2ghz
GIGABYTE GA-MA785GM-US2H
HD 4200 IGP (graphics)
2x2gb G Skill PI Black@ 763mhz (3:5)
WD 320gb SE16 Single Platter OS Drive
Windows 7 Ultimate 64bit (build 6.1.7600)

Windows Index
CPU index score 7.5

Capture22.png

This in OS benchmark is used by Microsoft to recommend software based on system performance. Probably the least scientific test here, but fun to include nonetheless.


AMD OVERDRIVE Benchmarks
Overall 8096

Capture11.png


Some built in benchmarks in the AMD overclocking utility


Prime95
Best Iteration time 4096k FFT Lenght 46.620ms (other iteration in screenshot)
Capture3.png


Prime95 is a torture test/benchmark/distributed computing project all in one. Using Mersenne's Prime Search algorithm it searches for fantastically large prime numbers. It is a great multithreaded benchmark that stresses processor cache and measures floating point arithmetic performance.


Linx 0.64 64bit

20 runs - 9m 25s - 49.2140 GFlops peak

Capture20.png

Capture21.png


From benchmarkHD.ru:
LinX is a CPU stability tester based on the Intel Linpack technology. The main point of Linpack is to solve systems of linear equations. It is designed as a benchmark to test the performance of a system in GFlops - billions of floating point operation per second. But it is also the most stressful CPU testing program to date and is a great tool to determine CPU stability. One and the same system of equations is solved repeatedly; if all results match each other - the CPU is stable, otherwise the instability is obvious, since the same sytem cannot produce different solutions. Main features include: support for both Intel and AMD CPUs, both 32 and 64-bit Linpack support, HyperThreading support, Real-time error checking, simple and intuitive interface.


SuperPI
4M digits in 2m 7s (other pi lengths in screenshot)

Capture4.png


from wikipedia:
Super PI is a computer program that calculates pi to a specified number of digits after the decimal point - up to a maximum of 32 million. It uses Gauss–Legendre algorithm and is a Windows port of the program used by Yasumasa Kanada in 1995 to compute Pi to 232 digits.


Cinbench R10 64bit
CPUX 17383

Capture2.png


From Maxon.net:
CINEBENCH is a real-world cross platform test suite that evaluates your computer's performance capabilities. CINEBENCH is based on MAXON's award-winning animation software CINEMA 4D, which is used extensively by studios and production houses worldwide for 3D content creation. MAXON software has been used in blockbuster movies such as Spider-Man, Star Wars, The Chronicles of Narnia and many more.

CINEBENCH is the perfect tool to compare CPU and graphics performance across various systems and platforms (Windows and Mac OS X).



3D Mark Vantage
CPU Test - 15979

Capture5.png


3DMark uses an AI and a CPU computed physics tests to measure processor performance for gaming applications.

From futuremark.com:
The 3DMark score is the overall measure of a systems 3D gaming performance, based on comprehensive individual real-time 3D graphics tests and processor tests. Each 3DMark product produces a 3DMark score as its outcome, and the 3DMark scores are comparable within the same product.


PCMark Vantage
Total Score 6024. (Individual scores in screenshot)

Capture12.png


PCMARK tests the overall system performance in a variety of real world applications. The CPU plays just a part in the outcome of these tests.

From futuremark:
A PCMark score is a measure of your computer’s performance across a variety of common tasks such as viewing and editing photos, video, music and other media, gaming, communications, productivity and security.
 

Mothergoose729

Senior member
Mar 21, 2009
409
2
81
Sisoft Sandra 2010 CPU Benchmark Suite

Arithmetic Test

Dryhstone ALU 72.6 GIPS
Whetstone FPU 53.31 GFLOPS

10-Copy.png


From Wikipedia:

Dhrystone is a synthetic computing benchmark program developed in 1984 by Reinhold P. Weicker intended to be representative of system (integer) programming. The Dhrystone grew to become representative of general processor (CPU) performance until it was superseded by the CPU89 benchmark suite from the Standard Performance Evaluation Corporation, today known as the "SPECint" suite. The name is a pun on a different benchmark algorithm called Whetstone.

The Whetstone benchmark is a synthetic benchmark for evaluating the performance of computers.[1] It was first written in Algol 60 in 1972 at the National Physical Laboratory in the United Kingdom and derived from statistics on program behaviour gathered on the KDF9 computer, using a modified version of its Whetstone Algol 60 compiler. The program's behavior replicated that of a typical KDF9 scientific program and was designed to defeat compiler optimizations that would have adversely affected the accuracy of this model. The Whetstone Compiler was built at the Atomic Power Division of the English Electric Company in Whetstone, Leicestershire, England,[2] hence its name.

The Whetstone benchmark primarily measures the floating-point arithmetic performance. A similar benchmark for integer and string operations is the Dhrystone.


Processor MultiMedia
Mutli-media int - 215.6Mpixel/s
Multi-media float - 130.4Mpixel/s

Capture9.png


From Sisoft Sandra's User Manual:
Benchmark the (W)MMX(2), SSE(2/3/4), AVX processor units. Shows how your processors handle multi-media instructions and data in comparison to other typical processors.

Such operations are used by more specialised software, e.g. image manipulation, video decoders/encoders, games.

The test involves the generation of Mandelbrot Set fractals that are used to realistically describe and generate natural objects such as mountains or clouds. By using various multi-media extensions better performance is achieved.


Multi-Core Efficiency

Intercore bandwidth 4.46GB/s
Intercore latency 91ns
Capture8.png


From Sisoft Sandra's User Manual:
Benchmark the multi-core efficiency of the processors. Shows how efficient the processor cores and their inter-connects are in comparison to other types to other typical processors.

The ability of the cores to process data blocks and pass them to another core for processing (producer-consumer paradigm) of different sizes and different chain sizes is measured. The efficiency of the inter-connect between cores is thus benchmarked; however, the number of cores (and processors) also counts as more data buffers can be processed simultaneously (aka "in flight").

True multi-core processors that have shared L2/L3 caches will thus perform much better than cores that have separate caches and are connected by the traditional FSB.


Power Management Efficiency
ALU Power Performance - 21725MIPS
Power Efficiency - 2.12

Capture6.png


From Sisoft Sandra's User Manual:
Benchmark the power management efficiency of the processors. Shows how efficient the power management of your processors is in comparison to other typical processors.

The ability of the processors to step-down in frequency and voltage at different workloads is measured. The more a processor steps down in both frequency and voltage the better the score at the specific workload. The test stops when the workload is too great the processor even at 100% efficiency.

The ALU/FPU score is a geometric mean based on the whole range of workloads; thus the power of the processors does matter in obtaining a higher score.

The Power Efficiency score is a geometric mean based on the supported workloads only. Thus the power of the processors does not matter.


Cryptography
Cryptographics Bandwidth - 773mb/s
Hashing Bandwidth - 848mb/s

Capture7.png


From Sisoft Sandra's User manual:
Measures the cryptography efficiency of the processor units: encryption, decryption and hashing. Shows how your processors handle cryptographic operations in comparison to other typical processors.
 

Mothergoose729

Senior member
Mar 21, 2009
409
2
81
Everest 5.30.1900 Ultimate Edition Benchmark Suite

CPU Queen
28246

Capture13.png


From Everest Ultimate Edition Manual:
This simple integer benchmark focuses on the branch prediction capabilities and the misprediction penalties of the CPU. It finds the solutions for the classic "Queens problem" on a 10 by 10 sized chessboard

CPU Queen test uses integer MMX, SSE2 and SSSE3 optimizations. It consumes less than 1 MB system memory and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.


CPU PhotoWorxx
19280

Capture14.png


From Everest Ultimate Edition Manual:
This integer benchmark peforms different common tasks used during digital photo processing.

CPU PhotoWorxx test uses only the basic x86 instructions, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.


CPU ZLib
122464KB/s

Capture15.png


From Everest Ultimate Edition Manual:
This integer benchmark measures combined CPU and memory subsystem performance through the public ZLib compression library

CPU ZLib test uses only the basic x86 instructions, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.


CPU AES
33194

Capture16.png


From Everest Ultimate Edition Manual:
This integer benchmark measures CPU performance using AES (a.k.a. Rijndael) data encryption. It utilizes Vincent Rijmen, Antoon Bosselaers and Paulo Barreto's public domain C code in ECB mode.

CPU AES test uses only the basic x86 instructions, and it's hardware accelerated on VIA PadLock Security Engine capable VIA C3 and VIA C7 processors. The test consumes 48 MB memory, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.


FPU Julia
12361
*note, when this test completed Everest gave me a warning saying it wasn't fully optimized for my system. Results seemed fairly well in line with other intel and AMD processors however. All the FPU tests in the everest benchmark suite gave me this warning when completed.


Capture17.png


From Everest Ultimate Edition Manual:
This benchmark measures the single precision (also known as 32-bit) floating-point performance through the computation of several frames of the popular "Julia" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing the appropriate x87, 3DNow!, 3DNow!+ or SSE instruction set extension.

FPU Julia test consumes less than 1 MB system memory, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.


FPU Mandel
7824

newmandel.png


From Everest Ultimate Edition Manual:
This benchmark measures the double precision (also known as 64-bit) floating-point performance through the computation of several frames of the popular "Mandelbrot" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing the appropriate x87 or SSE2 instruction set extension.

FPU Mandel test consumes less than 1 MB system memory, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.


FPU SinJulia
4015

newsinjulia.png


From Everest Ultimate Edition Manual:
This benchmark measures the extended precision (also known as 80-bit) floating-point performance through the computation of a single frame of a modified "Julia" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing trigonometric and exponential x87 instructions.

FPU SinJulia test consumes less than 1 MB system memory, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
 

richierich1212

Platinum Member
Jul 5, 2002
2,741
360
126
Very comprehensive testing. If you can't get 3.8GHz stable, double check your memory timings/settings. You may also need to bump up NB to 2.4GHz, and put CPU-NB to 1.25v
 

Schmide

Diamond Member
Mar 7, 2002
5,745
1,036
126
Ya need to keep your Hypertransport link around 2k try lowering the multiplier to 7x or 8x before trying 3.8ghz+. Make sure your north bridge latch and memory divider are lowered as well.

Mine needed 1.5v to get 4.06ghz my base processor voltage was 1.35. (so +0.15v)
 

Mothergoose729

Senior member
Mar 21, 2009
409
2
81
Well color me skeptic... I messed with the HT and bus links and now I am running at 4.0ghz. I have some stability testing left to do, but I will update and post back more results soon.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,278
16,121
136
Well color me skeptic... I messed with the HT and bus links and now I am running at 4.0ghz. I have some stability testing left to do, but I will update and post back more results soon.

I couldn't believe how easy it was t hit 4.1, and on a cheap motherboard.

But I did get the 1090T for the unlocked multi.
 
Last edited:
Apr 20, 2008
10,067
990
126
I couldn't believe how easy it was t hit 4.1, and on a cheap motherboard.

But I did get the 1090T for the unlocked multi.

When you find out a reputable place to buy a cheap motherboard, let me know. Haven't been able to get my hand on one in a loooong time. :awe:

These Thubans make me want to upgrade. I need to get back to working before I can drop money like that again.
 
Last edited by a moderator:

lopri

Elite Member
Jul 27, 2002
13,314
690
126
Your IGP holds back overclocking via HTT. I'm testing my 1090T on the exact same board as yours (bundle combo for $40) and made sure I turned off the IGP. (using an HD 3850 for the moment)
 

Schmide

Diamond Member
Mar 7, 2002
5,745
1,036
126
I'm on a Gigabyte GA-890GPA-UD3H ver 1.0 (bios F7c) using the IGP and I can get to 4.2ghz although it's only stable OCCT at 4.06ghz.

Are you sure?
 

Mothergoose729

Senior member
Mar 21, 2009
409
2
81
I can't see how they are related TBH. I had a 4890 dedicated card in my system part of the time, but took it out because I decided to sell it. Stability when testing was the same.
 

MangoTBG

Diamond Member
Apr 28, 2003
3,101
0
76
Thank you very much for this post! Definitely showed me a few benchmarks I hadn't tried.

I've got my 1055t at 3.5GHz and it seems that our numbers are fairly close across the board.

I wouldn't mind trying to undervolt for lower temps. Any suggestions on how far I can go?
 

Mothergoose729

Senior member
Mar 21, 2009
409
2
81
Thank you very much for this post! Definitely showed me a few benchmarks I hadn't tried.

I've got my 1055t at 3.5GHz and it seems that our numbers are fairly close across the board.

I wouldn't mind trying to undervolt for lower temps. Any suggestions on how far I can go?

It is rated between 1.225-1.475 volts, but it will do much less then that. I advise turning turbo off, and enabling CE1 which will lower idle voltage and make everything easier to control, and coolnquite of course will lower temps at idle without compromising stability.

I found that the processor seemed the most efficient around 3.0ghz. There is the "magic" clock speed threshold that can give you pretty good performance and incrediably low volts. You have to play with it and find out where it is. Remember, increasing clock speed at the same voltage still increase power consumption.
 
Last edited:
Dec 30, 2004
12,553
2
76
I think you need to overclock the cpu-nb. My current chip scores like 14k in Cinebench. L3 cache very important.
 

Mothergoose729

Senior member
Mar 21, 2009
409
2
81
I think you need to overclock the cpu-nb. My current chip scores like 14k in Cinebench. L3 cache very important.

I was playing with it before. At 2400mhz @4.0ghz I was getting something like 20k. I need to fiddle with it a bit more though. I am getting a new motherboard and going through some hardware, when everything get settled I am going to go after these tests again. I expect the scores will be much higher.
 
Dec 30, 2004
12,553
2
76
Hm, I think you should be able to get higher than that. I'm at 2.6Ghz, which should be pretty easily attainable. 2.8 would be good for scores.
 

lopri

Elite Member
Jul 27, 2002
13,314
690
126
Well, I've just moved my 1090T from a 785G board to a 790X board, and there are some significant performance improvement, especially where memory matters. I don't want to go off-topic here, though, because I'm not sure what the topic has become in this thread.. :confused:

Is it about:

1) Which chipset/board clocks Thuban better?
2) Which chipset/board clocks HTT better in general? (Does IGP affect/hinders overclocking HTT, if it's being used?)
3) Which chipset/board performs better with same CPU and memory?

My opinion was centrally about #2 and #3.
 
Last edited:

Mothergoose729

Senior member
Mar 21, 2009
409
2
81
Well, I've just moved my 1090T from a 785G board to a 790X board, and there are some significant performance improvement, especially where memory matters. I don't want to go off-topic here, though, because I'm not sure what the topic has become in this thread.. :confused:

Is it about:

1) Which chipset/board clocks Thuban better?
2) Which chipset/board clocks HTT better in general? (Does IGP affect/hinders overclocking HTT, if it's being used?)
3) Which chipset/board performs better with same CPU and memory?

My opinion was centrally about #2 and #3.

The topic is about thuban and performance. What did you find out?
 
Apr 20, 2008
10,067
990
126
Thanks for the benches btw. It's always nice to have more user scores rather than just reading of review systems which have no other bottlenecks than the part being benchmarked.