| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Community
Forum: Chat Room
Thread: Can you do more with less? 134.5 Floating Point MIPS/W.... still tweaking |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 158
|
|
| Author |
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges: 
|
----------------------------------------  |
||
|
|
GeraldRube
Master Cruncher United States Joined: Nov 20, 2004 Post Count: 2153 Status: Offline Project Badges:
|
Darn i was really looking forward to that   |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges: 
|
Hope they mean this July
 Not a bad thing if they can get better performance; I would rather wait another month if it means a 10 or 20% performance hike. |
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
On the present SB's, Using CPUID HWMonitorPro I noticed that while the TDP of an i7-2600K is 95W, in reality it only uses between 60 and 65W when crunching at 100% (HT on). Are you sure about that? I did testing on several 2600k chips and at stock settings they pulled 130-135 watts at the wall at 100% with HT on according to my meter. Not sure; still can't find my meter, but HWMonitorPro says 60 to 65W and I would go along with that, given the heat and fan speed. Turbo off, but clock tweaked up. Anand say's 86W, but that's with turbo left on, and Turbo a big power user. Here it's suggested that the 2600K only uses 87W at 4.5GHz. Here is a shot of an i7-860 95 watt cpu crunching at 100% HT on. Full load, 120 watts from cpu. 24 watts at idle. The load at the wall is 155 watts. The non pro version of CPU-Z doesn't give the same stats for a 2600k but looking at these numbers makes me wonder about the 87 watt draw at 100% with HT on. JMHO 
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
  |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges:
|
No mention of your CPU's frequency. Stock or overclocked?
----------------------------------------95W may indicate the CPU usage (or perhaps just TDP) while the 119.98W indicates the CPU and other processor related power usages (chipset). There might be variation between what is displayed for each CPUID tool edition and what it can show for different CPU types. Either that or it's being especially taxed or overclocked. Whatever the case, a 45nm CPU is not a 32nm processor with new architecture. [Edit 1 times, last edit by skgiven at Jun 6, 2011 11:34:18 PM] |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges:
|
Found my Plug-in Power Monitor, so I can measure the power requirements of this SB system (i7-2600K). System has 8GB DDR3, 2 SATA drives, 1 blue-ray drive and a GTX470, PS2 Keyboard, Trackball, a front panel USB/SD adapter. I’m not including the monitor.
----------------------------------------For ref. the idle power usage of the GTX470 is 33W (at which point the shaders drop to 51MHz) and it supposedly has a TDP of 215W (shaders are at 607MHz). After booting and before I logged on (the best place to get a base line reading) the power usage was 79W. After booting the power draw hovered around 82W (AV, some startups, FF and Word open). Had the CPU at almost ref clocks, not quite fixed from playing (3434MHz, some very slight core voltage bumps), HT on, Turbo off. Started running WCG tasks one by one: 1 task, 103W (C4CW) 2 tasks, 115W (C4CW + FA@H) 3 tasks, 117W (C4CW + 2x FA@H) 4 tasks, 123W (C4CW + 2x FA@H + HCMD) 5 tasks, 131W (C4CW + 3x FA@H + HCMD) 6 tasks, 136W (C4CW + 4x FA@H + HCMD) 7 tasks, 138W (C4CW + 4x FA@H + HCMD + HPF2) 8 tasks, 139W (C4CW + 4x FA@H + HCMD + 2x HPF2) - just working through a small backlog b4 returning to CEP2. CPU temps around 52degC. If you take off the 33W for the GPU, that would be 106W for the entire system (less the GPU). The board would use about 30W and the other components about 10 or 15W, so the CPU itself is probably only using around 60 to 65W. While Cinebench may push it over 80W these tasks don't. Added the GPU and the system jumped to 252W. When the GPU temps rose to their peak the total power usage rose to 265W. As that is not a reliable config I dropped the CPU usage back down to 6threads and the power usage fell slightly to 263W. [Edit 1 times, last edit by skgiven at Jun 19, 2011 4:39:05 PM] |
||
|
|
kateiacy
Veteran Cruncher USA Joined: Jan 23, 2010 Post Count: 1027 Status: Offline Project Badges:
|
Interesting crystal-ball gazing from the AMD Fusion Developers summit:
----------------------------------------http://www.theregister.co.uk/2011/06/14/amd_trinity_and_elephants/ I don't know whether AMD will be putting 10-Teraflops power into notebooks by 2020, but I'll bet somebody will.  |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges:
|
At 3604MHz (1.068V) the system is using 153W, without the GPU running. Take off the 33W for the GPU and that is 123W.
Boinc Benchmarks: 3408 floating point MIPS (Whetstone) per CPU 11391 integer MIPS (Dhrystone) per CPU So the SB can do 3408 MIPS x 8 / 123W = 221 MIPS/W When the GPU is running (most of the time) the GPU app speed is over 48 GFlops/sec (@ 98% GPU utilization and the system power usage going over 290W). |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
I knew there was a reason I set out some rules for this thread
---------------------------------------- This is the power taken at the power socket by a crunching "box" all up CPU, Mobo, Mem, GPU and HD - the minimum system for crunching - including the losses in the PSU. If it's a laptop it will include the power taken by the screen Without HT and using the BOINC CPU benchmark for Floating Point MIPS Take the BOINC CPU Benchmark result multiplied by the number of real Cores and then divide by the power taken at the socket by the whole box or laptop The result is the benchmark to beat -- Then AMD threaten to bring out a CPU with a shared (2*128bit) FPU between 2 x86-64 Cores just to confuse us all - mainly the boinc single threaded benchmark  (The integer units being allocated solely to their own cores) -- Let's hope the FX-8110 arrives this year http://www.cpu-world.com/CPUs/Bulldozer/AMD-FX-Series%20FX-8110.html For the same TDP as the 1055T 6 Core at 2.8Ghz you get 8 Cores at 3.6Ghz I wonder if it will get, all in, under the 133W at the socket [Edit 1 times, last edit by David Autumns at Jun 18, 2011 6:49:10 PM] |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges:
|
Hi David,
The reason I excluded the power used by the GPU is that it is a Very power hungry GPU (215W TDP). Obviously I could use a sub 10W GPU, and I do have several lying about. The GPU is still the main cruncher for me, and I can't see that changing any time soon. That's why I showed it's performance. You can't differentiate between cores and threads unless Boinc's Benchmark does, and it doesn't; it just looks at the thread count. When the BD turns up that will be treated equally in it's many forms. It is worth noting that if I select to crunch on say 2 CPU threads the benchmarks are Much higher. So the benchmark is quite accurate for these SB processors. Perhaps less so for some previous generation i7's. The worry I have about BD is the RAM; dual channel will probably not cut it for some projects, especially with 8 fuller cores, so we might have to mix and match. Anyway I do look forward to having one. |
||
|
|
|