• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Bulldozer Finally!

Soldato
Joined
18 Oct 2002
Posts
15,861
Location
NW London
... if you thread limit your program to, say, 2,

No way! I've busted a gut trying to get everything working with many many threads. Reducing down to 2 threads would be nigh on impossible now.

You could also consider an i7, which I think would be unarguably faster :D

To go the 2600k route, you are effectively paying extra, just for the HT facility and I'm not 100% sold on HT.

I have a laptop (core i3) which has HT and it performs slower than my Core2Duo machine (when running my program). This makes me think that perhaps HT will not be of much benefit to me. I would rather spend the extra money saved on more RAM.

I think it really would be between the 1100T and i2500k...both over-clocked.
 
Soldato
Joined
19 Dec 2002
Posts
3,745
I have a laptop with an Intel i5-2520M - dual core with hyperthreading. I timed a benchmark using 1,2,3 and 4 threads, and the time to complete for each thread

One instance - 27 secs
Two instances - 30 secs each
Three instances -43 secs each
Four instances - 55 secs each

So much for hyperthreading - the speed of each instance drops almost in half going from 1 thread to 4. Of course with 1 thread it turbos at 3.2Ghz, 2 threads at 3.0Ghz and with 4 threads it is 2.5GHz, so you need to take that into account. But essentially hyperthreading is useless for me
 
Soldato
Joined
26 Jan 2007
Posts
2,541
Location
Leeds
No way! I've busted a gut trying to get everything working with many many threads. Reducing down to 2 threads would be nigh on impossible now.

I only meant for benchmarking purposes ^^; Admittedly if your threads do different things, rather than taking 1/4 of a general calculation, then it's a bugger :p
 
Soldato
Joined
18 Oct 2002
Posts
15,861
Location
NW London
Unfortunately, my program uses separate threads for all sorts of things.
About 20 threads are running throughout the life of the program, using very little computer processor power...but occasionally, some of the threads will come to life (all independent of one another) and take CPU usage to above 80%.

I then have multiple threads used in a single method, which is another way that threads have been used.

To recode the program (which now consists of over 45k lines of code), into a single threaded program would be a nightmare...and besides it would be a downgrade. In general, multi-threaded program is damn difficult to code, but has major speed benefits...going backwards would be a little perverse.

I have created a benchmarking program which creates X clients, all connected to 1 server; those clients then bombard the server with Y messages/queries. The time taken to respond to all those queries, is the final benchmark result. The shorter the time is, the faster the program is running.

I guess the only way to run this program in a single thread would be to use the task manager...however, that would defeat the whole purpose of seeing which CPU can run the program faster.
 
Soldato
Joined
26 Jan 2007
Posts
2,541
Location
Leeds
Under the circumstances it does sound like you've rather gone beyond the complexity where a single thread benchmark is actually informative xD

Wasted my mathcraft :(
 
Soldato
Joined
9 Nov 2009
Posts
24,927
Location
Planet Earth
Associate
Joined
21 Apr 2008
Posts
1,536
Location
Manchester
I have a laptop with an Intel i5-2520M - dual core with hyperthreading. I timed a benchmark using 1,2,3 and 4 threads, and the time to complete for each thread

One instance - 27 secs
Two instances - 30 secs each
Three instances -43 secs each
Four instances - 55 secs each

So much for hyperthreading - the speed of each instance drops almost in half going from 1 thread to 4. Of course with 1 thread it turbos at 3.2Ghz, 2 threads at 3.0Ghz and with 4 threads it is 2.5GHz, so you need to take that into account. But essentially hyperthreading is useless for me

I think core parking is to blame, some programs see benefits, while others see little to no improvement.

Try this it may work for you link
 
Soldato
Joined
9 Nov 2009
Posts
24,927
Location
Planet Earth
I'm sure somebody here was crowing about how such an amazing amount of transistors had been packaged into such a small area. Well... http://www.anandtech.com/show/5176/amd-revises-bulldozer-transistor-count-12b-not-2b
So, much, much larger than Intel 32nm parts of almost equal transistor count. AMD need to get the BD replacement out of the door as soon as they can IMO.

Why are you so concerned with transistor densities??

There are two transistor counts for CPUs. One is the actual number of active transistors and the other the full number of transistors. Some transistors are not active and in other cases groups of transistors are counted as a single active one. The partticulat type of CPU design also affects density too. Sandy Bridge has a nominal count of 995 billion transistors;it actually had 1.16 billion transistors in total. Likewise,Llano is still the densest of all designs ATM.

42780.png
 
Last edited:
Associate
Joined
14 Jun 2008
Posts
2,363
Why are you so concerned with transistor densities??

I am? I'm just pointing out that some were making a big fuss about AMD putting out a 2bn transistor part that was only 315mm². When it turns out to be anything but and makes the (in some cases quite scary) heat and power draw issues all that much worse now we know the chip has almost half the number of transistors than what we originally thought.

As for Llano (which IMO is a good bit of kit), that is a completely different design, is not in competition with the majority of BD or Core chips, and this thread is about BD.
 
Soldato
Joined
18 Oct 2002
Posts
15,861
Location
NW London
Believe me, there are lot of people who are not the nerdy types (which we are) and they will buy them regardless, if the price/core makes sense.

AMD can definitely sell more of these, but only if they can make them faster. At present supply is barely satiating demand.
 
Back
Top Bottom