| 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: 
|
skgiven
----------------------------------------The BOINC Benchmark is the number crunching power of a single Core It's a single thread benchmark on each of the cores so HT is redundant HT just virtualises the threads so that the real cores have more work thrown at them so the latencies in fetching info are reduced and the potential for the CPU to just twiddle it's thumbs waiting for the next thing to turn up for it to do is reduced but under the covers you can only run 4 threads at the same time on a 4 core processor - only 4 FPU's Bulldozers design certainly muddies the waters on this distinction The best way to describe it would be hardware supported hyperthreading Can you see the problems it will bring when it comes to measuring the performance against another cpu using BOINC's current benchmark? The only way to know for sure how efficient it is will to compare its points per day allocation to today's CPU's According to BOINCstats over the last month I am getting 3120 BOINC points/day out of my 1055T ( 21840 WCG points/day ) for just 133W at the socket (0.2346GFlops/W) When the WCG started I had a 3200+ Barton Cored Athlon when CPU's had just one core  This had a TDP of 89W and must have been taking at least 150W at the socket probably much more. I know it produced 3.447GFlopsEven being generous it would have produced just 0.023GFlops/W 7 Years later I am producing over 10 times as much work per watt for the WCG That's progress  [Edit 1 times, last edit by David Autumns at Sep 14, 2011 11:16:18 PM] |
||
|
|
KWSN - A Shrubbery
Master Cruncher Joined: Jan 8, 2006 Post Count: 1585 Status: Offline |
They're attempting to make a scaleable architecture that will work for more than a couple years. The idea of splitting up the tasks of a CPU isn't a new concept. The core does the bulk of the work, and FPU typically sits idle waiting for a task from the core. By pairing one FPU with two cores, it makes the entire operation more efficient, takes up less space, and uses less energy.
----------------------------------------David is right, this should put out significantly better end results than hyperthreading which is merely a logical core. The bottleneck at the FPU (if there is any at all) won't hold it back as much as calling a few spare cycles an extra core.  Distributed computing volunteer since September 27, 2000 |
||
|
|
retsof
Former Community Advisor USA Joined: Jul 31, 2005 Post Count: 6824 Status: Offline Project Badges:
|
So the forthcoming Bulldozer CPU's will have 2 'Cores' per Module  That sounds very much like 2 threads per Core! Sounds like AMD are attempting to rename Core to Module  Eh, not buying that... It is sort of / kind of like hyperthreading, since each bulldozer module of 2 cores has only one floating point register. A module has two CPUs in it, similar to what Intel has been doing.
SUPPORT ADVISOR
Work+GPU i7 8700 12threads School i7 4770 8threads Default+GPU Ryzen 7 3700X 16threads Ryzen 7 3800X 16 threads Ryzen 9 3900X 24threads Home i7 3540M 4threads50% |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
Let's hope this doesn't turn into vapourware
----------------------------------------http://www.cpu-world.com/CPUs/Bulldozer/AMD-F...8120%20FD8120WMW8KGU.html 8 Cores (open to debate - see above) 3.1Ghz/core all cores flat out 95W TDP Earlier greener AMD CPU's vanished into thin air That said I do have the 905E and the 1055T with 65W and 95W TDP's running in the machines beside me so every now and again they do make it into reality This CPU is more efficient than the quicker FX-8150 (8 cores 125W but clocked at 3.6Ghz all cores flat out) that will be released at the same time Maybe released on the 12th October but this has been a long old wait |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
Tomorrow
---------------------------------------- |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
http://www.overclockers.co.uk/productlist.php...mp;catid=6&subid=1942
----------------------------------------nearly there As expected the 3.1Ghz 95W TDP 8 core is not among them The benchmarks out there are not looking too promising It's beating the i7-2600 though with it's hyperthreading enabled  |
||
|
|
krakatuk
Advanced Cruncher Germany Joined: Oct 3, 2008 Post Count: 141 Status: Offline Project Badges: 
|
Honestly saying, benchmarks are looking very disappointing.
---------------------------------------- And it seems to be not a candidate for new MIPS/W records as well - it's very hungry... Another sandbox-bulldozer... So sad...   |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
---------------------------------------- [Edit 1 times, last edit by Former Member at Oct 12, 2011 3:31:14 PM] |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
It's that shared FPU
----------------------------------------Still interesting to know how it will perform multithreaded on real World Community Grid Work Units With a bit of luck I will get my bonus in November  maybe the vapourware (UK spelling ) 95W TDP 8 core 3.1Ghz will have turned up by then |
||
|
|
kateiacy
Veteran Cruncher USA Joined: Jan 23, 2010 Post Count: 1027 Status: Offline Project Badges:
|
Yes, disappointing in both absolute performance and performance per watt.
----------------------------------------I guess I get to "hurry up and wait" some more to see what both Intel and AMD are going to deliver in early 2012.  |
||
|
|
|