AMD Shanghai/Deneb Review Thread

[iocedmyself]

3.6ghz Intercore

SiSoftware Sandra

Benchmark Results
Inter-Core Bandwidth : 4.70GB/s
Results Interpretation : Higher index values are better.
Inter-Core Latency : 71ns
Results Interpretation : Lower index values are better.

Performance vs. Speed
Inter-Core Bandwidth : 1.60MB/s/MHz
Results Interpretation : Higher index values are better.
Inter-Core Latency : 0.02ns/MHz
Results Interpretation : Lower index values are better.

Detailed Benchmark Results
Processor Affinity : CPU0-CPU2 CPU1-CPU3
Inter-Core Bandwidth @ 2x 8kB : 4.77GB/s
Inter-Core Bandwidth @ 4x 8kB : 4.76GB/s
Inter-Core Bandwidth @ 2x 32kB : 4.95GB/s
Inter-Core Bandwidth @ 4x 32kB : 4.90GB/s
Inter-Core Bandwidth @ 16x 8kB : 4.74GB/s
Inter-Core Bandwidth @ 2x 128kB : 5.23GB/s
Inter-Core Bandwidth @ 4x 128kB : 4.71GB/s
Inter-Core Bandwidth @ 16x 32kB : 4.81GB/s
Inter-Core Bandwidth @ 64x 8kB : 4.76GB/s
Inter-Core Bandwidth @ 16x 128kB : 5.19GB/s
Inter-Core Bandwidth @ 64x 32kB : 4.87GB/s
Inter-Core Bandwidth @ 64x 128kB : 3.12GB/s

Performance Test Status
Run ID : AMD Phenom(tm) II X4 940 Processor (Quad-Core, 3.01GHz, 4x 512kB L2, 6MB L3)
Platform Compliance : Win64 x64
Buffering Used : Yes
NUMA Support : No
SMP (Multi-Processor) Benchmark : No
Total Test Threads : 4
Multi-Core Test : Yes
SMT (Multi-Threaded) Benchmark : No
System Timer : 14.32MHz

Processor
Model : AMD Phenom(tm) II X4 940 Processor
Speed : 3.01GHz
Model Number : 9024
Cores per Processor : 4 Unit(s)
Type : Quad-Core
L2 On-board Cache : 4x 512kB, ECC, Synchronous, Write-Back, 16-way, 64 byte line size
L3 On-board Cache : 6MB, ECC, Synchronous, Write-Back, 48-way, 64 byte line size, 4 threads sharing

Performance Tips
Notice 5008 : To change benchmarks, click Options.
Notice 5004 : Synthetic benchmark. May not tally with 'real-life' performance.
Notice 5006 : Only compare the results with ones obtained using the same version!
Tip 11 : Use the 'Switch Chart Type' button to switch between Detailed and Combined charts.
Tip 2 : Double-click tip or press Enter while a tip is selected for more information about the tip.

[SoF]

super Pi is not for stable testing

xtremesystems.org is not for stable testing usually also I did a little prime at around 3.6 with 1.4V yesterday for about 30min fine. Think 3.8 should be no problem with average air cooling. scales good with voltage.

[iocedmyself]

AMD Phenom(tm) II X4 940 Processor
CPU speed: 3608.79 MHz, 4 cores
CPU features: RDTSC, CMOV, Prefetch, 3DNow!, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 48
L2 TLBS: 512
Prime95 32-bit version 25.6, RdtscTiming=1
Best time for 768K FFT length: 12.826 ms.
Best time for 896K FFT length: 15.363 ms.
Best time for 1024K FFT length: 17.573 ms.
Best time for 1280K FFT length: 21.910 ms.
Best time for 1536K FFT length: 26.959 ms.
Best time for 1792K FFT length: 32.044 ms.
Best time for 2048K FFT length: 36.146 ms.
Best time for 2560K FFT length: 47.483 ms.
Best time for 3072K FFT length: 57.918 ms.
Best time for 3584K FFT length: 69.354 ms.
Best time for 4096K FFT length: 78.383 ms.
Best time for 5120K FFT length: 106.893 ms.
Best time for 6144K FFT length: 129.516 ms.
Best time for 7168K FFT length: 155.793 ms.
Best time for 8192K FFT length: 179.298 ms.
Timing FFTs using 2 threads.
Best time for 768K FFT length: 8.663 ms.
Best time for 896K FFT length: 11.567 ms.
Best time for 1024K FFT length: 13.117 ms.
Best time for 1280K FFT length: 16.420 ms.
Best time for 1536K FFT length: 19.897 ms.
Best time for 1792K FFT length: 23.754 ms.
Best time for 2048K FFT length: 27.026 ms.
Best time for 2560K FFT length: 35.748 ms.
Best time for 3072K FFT length: 42.690 ms.
Best time for 3584K FFT length: 50.396 ms.
Best time for 4096K FFT length: 57.087 ms.
Best time for 5120K FFT length: 71.616 ms.
Best time for 6144K FFT length: 88.316 ms.
Best time for 7168K FFT length: 107.699 ms.
Best time for 8192K FFT length: 122.878 ms.
Timing FFTs using 3 threads.
Best time for 768K FFT length: 7.151 ms.
Best time for 896K FFT length: 12.434 ms.
Best time for 1024K FFT length: 13.562 ms.
Best time for 1280K FFT length: 15.674 ms.
Best time for 1536K FFT length: 17.912 ms.
Best time for 1792K FFT length: 20.610 ms.
Best time for 2048K FFT length: 22.946 ms.
Best time for 2560K FFT length: 33.583 ms.
Best time for 3072K FFT length: 38.143 ms.
Best time for 3584K FFT length: 43.591 ms.
Best time for 4096K FFT length: 48.397 ms.
Best time for 5120K FFT length: 56.122 ms.
Best time for 6144K FFT length: 69.391 ms.
Best time for 7168K FFT length: 85.529 ms.
Best time for 8192K FFT length: 95.301 ms.
Timing FFTs using 4 threads.
Best time for 768K FFT length: 6.646 ms.
Best time for 896K FFT length: 11.919 ms.
Best time for 1024K FFT length: 12.367 ms.
Best time for 1280K FFT length: 14.569 ms.
Best time for 1536K FFT length: 17.115 ms.
Best time for 1792K FFT length: 19.788 ms.
Best time for 2048K FFT length: 21.897 ms.
Best time for 2560K FFT length: 31.301 ms.
Best time for 3072K FFT length: 35.879 ms.
Best time for 3584K FFT length: 41.123 ms.
Best time for 4096K FFT length: 45.414 ms.
Best time for 5120K FFT length: 52.534 ms.
Best time for 6144K FFT length: 62.155 ms.
Best time for 7168K FFT length: 74.891 ms.
Best time for 8192K FFT length: 84.444 ms.
Best time for 58 bit trial factors: 3.269 ms.
Best time for 59 bit trial factors: 3.264 ms.
Best time for 60 bit trial factors: 3.261 ms.
Best time for 61 bit trial factors: 3.269 ms.
Best time for 62 bit trial factors: 6.000 ms.
Best time for 63 bit trial factors: 5.997 ms.
Best time for 64 bit trial factors: 6.002 ms.
Best time for 65 bit trial factors: 5.966 ms.
Best time for 66 bit trial factors: 5.971 ms.
Best time for 67 bit trial factors: 5.932 ms.

[justapost]

Thanks for all the sandra results, wast that 2009.SP1 or SP2? I received new WC parts so i can not run comparisons atm. Undervolt looks really good. 1.1V for 2GHz NB is another nice detail. Also ACC seems to work on your board!

[Macadamia]

SoF's chip and some others say that 3.8-3.9 is the sweet spot.

iocedmyself's chip says that everything can be sweet.


Decisions decisions, lol.

[iocedmyself]

heh oops actually was using 2008 1.13.12 sandra. I'll reload 2009 edition and rerun benches this afternoon or tomorrow With hard disk death and having to reinstall just trying to keep some info flowing

But yeah imma have to say that 3.8-4.0ghz will prolly be the sweet spot for most retail chips on good air cooling, and i again i have to say i suspect the 4+1 phase power is holding back stability because looking in the bios as the supplied vcore voltage it's anywhere between +.002v and .150 of what the cpu voltage is actually set to and i would thin that under operations that would be even more shakey. I can bench up to 4.0 but it isn't 100% stable and results aren't scaling as well as they should so not really worth posting IMHO.

Yes ACC seems to work so long as it's set to auto, i don't know enough about how it functions to go in and manually set % for any reliable results. But as for the undervolt i actually just grabbed the slider with the mouse and pulled it down to something under 1.25 =o)

Oh btw all other settings aside from Vcore and Cpu are stock voltages/freqs

I'm thinking of finally breaking down and putting this under water sometime in the next week or two along with putting it on a 750SB FX board just to see the limit of the thing. One thing that i must say about this BIOS is that it really doesn't like ANY changes being made. After every change i've made it hangs after saving changes and i have to cycle power again, but after that it will boot fine and stable. Where as when it isn't stable it will actually force me into the bios citing boot failure change setttings or reset to default. So i'm not sure if it's just the BIOS revision or the board itself.So more to come as i sort through my reinstall.

[Jazzman]

Yes ACC seems to work so long as it's set to auto, i don't know enough about how it functions to go in and manually set % for any reliable results. But as for the undervolt i actually just grabbed the slider with the mouse and pulled it down to something under 1.25 =o)

..... After every change i've made it hangs after saving changes and i have to cycle power again, but after that it will boot fine and stable. Where as when it isn't stable it will actually force me into the bios citing boot failure change setttings or reset to default. So i'm not sure if it's just the BIOS revision or the board itself.So more to come as i sort through my reinstall.

Auto ACC is probably changing core settings with each overclock, necessitating a reboot. Some of us have used manual ACC to achieve stable overclocks both in AOD and in the bios. The system must reboot to effect these changes.

BTW, with Phenom I's some us have found that at least +2 to +4 per core is required, sometimes auto ACC won't get you there. It definitely ups the the cpu voltage, in some cases significantly. Just for reference I have set my 9950BE at ACC per core settings of +4 +2 +2 +4.

[JohnJohn]

Oh btw all other settings aside from Vcore and Cpu are stock voltages/freqs

.

Hey! so you are saying that the ht stock frecuency is 2000 indeed???

[iocedmyself]

Well what i was talking about was ANY change i make in the bios regardless of if ACC is enabled or not. Ii've tried every manuel setting on all cores but benifits aren't apparent or seemingly consistaint. But i'll keep that in mind and see where i can get using +2 and +4 to all cores.

Btw the only definitive thing i've gleaned about ACC is that reboot is required when changing the ACC settings in over drive, so i've only bothered changing the settings in BIOS usually after i've pushed the chip a bit to far and gotten a BSOD.

Yes 2000mhz is the default clock for for HT, if the boad supported the new 3.0 or 3.1 it would default to that,but as the bios only allows HT selection up to 2200mhz. I have however overcloed HTT to 2600 @ 260x10 with cpu at 14 multi so seems HTT is more pliable like the older 939 chips were.

[Progerien]

do you know the edit function? by the way I don't like sandra benches at all
€dit: finally I have 4 gig 32m running but 4.1 crashes 1m ^^ still no good ram clock even with current bios (P05).

Hello,

Is it me ? Or why can't I see SuperPI result of Phenom II ? Amd doesn't want ? Or it sucks so much ?

At 4Ghz, a penryn can terminate the loop2 less than 1 minute. The phenom II doesn't appear to do this with no optimised ram

Nice work from AMD, I wish Phenom II will roxxx !

Sincerely,

[SoF]

wait for results with 230 htt and some cold - it will do a lot faster but nobody is allowed to show performance yet this is simple air testing and I am more not that good on air Pi never was something you should judge AMD's performance with but I want to compare it to AMD results until now - and for that it really rocks!

[Oldguy932]

Intel vs AMD in super pi isn't what we're trying to look at. Times are blacked out because the parts are still under some sort of nda and they are only allowed to show clocks and voltages at this point it seems.

[iocedmyself]

well i'm not under any NDA and was not told specifically that i couldn't show performance, just erring on the side of caution. Plus screenies are a PITA when you're running 4 monitors

Super Pi is largely dependent on cache sizes and doesn't really speak much for real world performance in most things, especially when you're talking of benchmark performance.

[Progerien]

Okay...

It's just that I have seen some results in games and others benchmarks. But no Superpi (a really good bench from my point of view )

But, if it can't do sub10 (with high frequencies and real cold) I think that AMD will stay in retreat against Intel. And continue the war of price....

Wait and see...

Sincerely,

[iocedmyself]

at 1066 ram, 200 x 10 HTT and 3.4ghz cpu, chip will complete 1mb Super PI in 18-19 seconds from what i've tested

[Progerien]

Thank for your answer

It's a good result for an AMD CPU, but not better than an old, old Q6600...

I wish it will be better in games... and cheaper than Q6600.

Sincerely,

[iocedmyself]

well actually i did point out earlier that the 3.0ghz performance of the deneb was more then double that of a 3.4ghz 9850be in GTA IV which actually does utilize all four cores, and did so with double the draw distance and near 20% less cpu load and memory use

[azraeel101]

Thank for your answer

It's a good result for an AMD CPU, but not better than an old, old Q6600...

I wish it will be better in games... and cheaper than Q6600.

Sincerely,

SPi works much better on Intel CPU-s because of architectural differences, but it doesn't reflect real world performance. Games are a totally different story, I expect Deneb to be very competitive in that area...

[charged3800z24]

Thank for your answer

It's a good result for an AMD CPU, but not better than an old, old Q6600...

I wish it will be better in games... and cheaper than Q6600.

Sincerely,

well it won't be cheaper, and also shouldn't be. It is 45nm tech, uses a ton less power, less heat, and looks to OC a lot better...Thats enough to ask more for it I believe.

[ajaidev]

iocedmyself do run some gameing benchs like GTA4, WIC, etc with screens if possible......

This processor seems more and more of an economic saver after considering i was going to buy a i7 set up and not buy a laptop...!!!

[SoF]

I am close to sub 17s 1m and think there is a lot room for improvements with cold.

[charged3800z24]

I am close to sub 17s 1m and think there is a lot room for improvements with cold.

well...then get 'er cold then ... By the time I get one of these...all the best scores will have been achived LOL...keep it coming..

[Oliverda]

iocedmyself

Could you please run Everest's Cache and Memory bench?

http://img220.imageshack.us/img220/6...hememocpk9.png

Thank you in advance!

[Dami3n]

I am close to sub 17s 1m and think there is a lot room for improvements with cold.

Sub 17s..... this is serious business for green side. That phenom seems to have a lot of headroom to improve .

[iocedmyself]

Sigh been a long time since i used Everest, sad to say i forgot how you ran that benchmark and ended up running all the other benches and saving screenies before it dawned on me