[Expreview] Test for AMD Bulldozer Hotfixes from Microsoft
1 - 20 of 77 Posts
Joined
·
4,766 Posts
thanks for the post. Glad 4 thread performance was up, funny how 8 "core" was down by small margins though, I would say its within margin of error but its very consistent throughout the tests.
I just wish there was more gaming results, either way the games they showed were supposed to be heavy cpu dependent games right? I was hoping for more gains
Quote:
I just wish there was more gaming results, either way the games they showed were supposed to be heavy cpu dependent games right? I was hoping for more gains
Quote:
Yeah....dont worry not all AMD users worship Bulldozer, its just some of us are stuck on amd and want to make the most of our expensive motherboards lol...Originally Posted by raclimja
Lol
Joined
·
75 Posts
They continue to make malicious benchmarks. Do they know how much improve the speed of the memory ram Fx` performance. They go on with the 1600 mgz speed memory; the speed must be least 1860. I own a fx an the procesor run better from better speed. the first bernchmarks were made with apm on. We know how much throttle from it. I think they knew that. but they didnt care.
Joined
·
3,607 Posts
Quote:
I'd be interested in seeing proof. I've not seen any benchmarks highlighting memory frequency scaling.Originally Posted by sorance2000
They continue to make malicious benchmarks. Do they know how much improve the speed of the memory ram Fx` performance. They go on with the 1600 mgz speed memory; the speed must be least 1860. I own a fx an the procesor run better from better speed. the first bernchmarks were made with apm on. We know how much throttle from it. I think they knew that. but they didnt care.
Joined
·
1,550 Posts
Now I've read everything.
Yes. Every website on the internet is intentionally paid off by the evil zionist nwo killuminati head council run by Nvidia & Intel. Despite many AMD biased websites posting benchmarks that are no different than any other review, it's ALL MALICIOUS TO HARM AMD!
Yes. Every website on the internet is intentionally paid off by the evil zionist nwo killuminati head council run by Nvidia & Intel. Despite many AMD biased websites posting benchmarks that are no different than any other review, it's ALL MALICIOUS TO HARM AMD!
Joined
·
10,307 Posts
lol. called it. this update wasnt going to increase performance on piledriver at all. in fact all bulldozer does is bulldoze literal crap.
Joined
·
203 Posts
So, this is to be the level of discourse on OCN these days?
Quote:
Quote:
When titles like Crysis 2 and Left for Dead 2 are seeing around a 10% performance increase, it would be nice to have a broader range of CPU bound, and GPU bound games being tested. Some increases in higher minimums, lightly threaded use, smoother performance, and better multi-GPU use are all welcome.Originally Posted by 66racer
thanks for the post. Glad 4 thread performance was up, funny how 8 "core" was down by small margins though, I would say its within margin of error but its very consistent throughout the tests.
I just wish there was more gaming results, either way the games they showed were supposed to be heavy cpu dependent games right? I was hoping for more gains
Yeah....dont worry not all AMD users worship Bulldozer, its just some of us are stuck on amd and want to make the most of our expensive motherboards lol...
Joined
·
13,796 Posts
The Bulldozer patch was targeted to improve core parking and how task scheduler issues.
Core parking problems occurred when certain cores were parked prematurely reducing performance. There is a performance penalty every time a core is parked and unparked during a task.
Task scheduler fix changed how threads were assigned to cores. It is more efficient for bulldozer after the fix. Now power consumption should be lower for lightly threaded tasks by putting threads on the same module first before moving on to the next module. This allows the Turbo core (dynamic overclocking) to occur properly.
These fixes will not really improve performance for fully loaded 8 threads. Edit: Also because of the more limited scope of scenarios, it was never going to give massive performance improvements across the board.
Core parking problems occurred when certain cores were parked prematurely reducing performance. There is a performance penalty every time a core is parked and unparked during a task.
Task scheduler fix changed how threads were assigned to cores. It is more efficient for bulldozer after the fix. Now power consumption should be lower for lightly threaded tasks by putting threads on the same module first before moving on to the next module. This allows the Turbo core (dynamic overclocking) to occur properly.
These fixes will not really improve performance for fully loaded 8 threads. Edit: Also because of the more limited scope of scenarios, it was never going to give massive performance improvements across the board.
Joined
·
15,008 Posts
Basically it improves performance when using less than 8 cores.
Interesting.
Now bring me a test that loads four threads at 50%. No, not two threads spread along 4 cores. I want 4 different threads running the same workload at 50%.
And let's see how much this improves their ability to share resources.
Interesting.
Now bring me a test that loads four threads at 50%. No, not two threads spread along 4 cores. I want 4 different threads running the same workload at 50%.
And let's see how much this improves their ability to share resources.
Joined
·
2,687 Posts
Quote:
Case and point. If you have a Gulftown and open the windows task manager how many threads do you see? 12 Why because we all know for a fact Gulftown is a 6 core processor w/ HT.
Gulftown is not a 12 core processor.
Because its a less than 8 core processor. I think this proves its a 4 core, 8 threads.Originally Posted by Artikbot
Basically it improves performance when using less than 8 cores.
Interesting.
Case and point. If you have a Gulftown and open the windows task manager how many threads do you see? 12 Why because we all know for a fact Gulftown is a 6 core processor w/ HT.
Gulftown is not a 12 core processor.
Joined
·
11,178 Posts
Quote:
Bulldozer doesn't really fit into the old classifications. It is a four module, eight thread CPU.
One module does not equal one core, but one module does not equal two cores as well.
It does have eight physical cores so AMD marketing isn't lying...but each pair of cores has to share a FPU so there isn't any way that it can function as a truly eight core CPU. No more than my CPU can if I turn on HT. Though the two processes are extremely different and each have their own pros and cons.
I see there is still a little bit of grease and fur left from this horse.Originally Posted by DrFPS
Because its a less than 8 core processor. I think this proves its a 4 core, 8 threads.
Case and point. If you have a Gulftown and open the windows task manager how many threads do you see? 12 Why because we all know for a fact Gulftown is a 6 core processor w/ HT.
Gulftown is not a 12 core processor.
Bulldozer doesn't really fit into the old classifications. It is a four module, eight thread CPU.
One module does not equal one core, but one module does not equal two cores as well.
It does have eight physical cores so AMD marketing isn't lying...but each pair of cores has to share a FPU so there isn't any way that it can function as a truly eight core CPU. No more than my CPU can if I turn on HT. Though the two processes are extremely different and each have their own pros and cons.
Quote:
Then why are you "jolly" and so concerned..? Can you not live by your own creed..?Originally Posted by Chewy
Correct.
But when these products eventually materialize and are genuinely not upto the pre launch hype, including official amd slides. then i see nothing to get jelly about..
Joined
·
1,535 Posts
Quote:
This is a test on Bulldozer not piledriver. Obviously. Piledriver doesnt come out till trinity and thats apu. The piledriver cpu doesnt come out till Q4 2012. So basically you called nothing cause windows 8 will be out before piledriver which its scheduler is a whole lot more complex for Bulldozer architecture than win 7 and its patches.
well you called wrong cause piledriver isnt out yet. If only you knew what you were talking about then maybe you might call right once in a while.Originally Posted by srsparky32
lol. called it. this update wasnt going to increase performance on piledriver at all. in fact all bulldozer does is bulldoze literal crap.
This is a test on Bulldozer not piledriver. Obviously. Piledriver doesnt come out till trinity and thats apu. The piledriver cpu doesnt come out till Q4 2012. So basically you called nothing cause windows 8 will be out before piledriver which its scheduler is a whole lot more complex for Bulldozer architecture than win 7 and its patches.
Joined
·
1,535 Posts
Quote:
Quote:
So you guys think that no one cares about bulldozer because this thread has only been out for 9 hours and since everyone is not on this site the same time you guys are and making posts taht you come to this conclusion? What you are saying makes absolutely no sense. Most members of ocn have yet to see this thread. This is my first time seeing it and i come on here all the time. You obviously have no idea whats going on. Hence, no one takes your comments seriously .Originally Posted by DayoftheGreek
No, Chewy had it right.
Quote:
This is exactly correct. This is what a lot of people dont understand or just cant comprehend. They also dont understand that Bulldozer is the first of its kind and has no optimizations whatsoever. this architecture has barely seen the day of light and yet everyone expects it to jump through hoops right out the gate. Perfection takes practice and practice takes time. These guys with your already optimized cpus comparing a new born architecture and just saying it completely sucks only makes them not look smart. Its like comparing IQ of a college student to an infant and saying the infant is stupid. No one can really say how Bulldozer architecture is until it has been optimized and the scheduler has been optimized which we'll see in windows 8. Wait until then to make your lame comments about how your daddy is better than someone elses daddy. Its like a day care in here trying to teach little kids how things work and get all whinny and throw a fit when they dont get there way. Pentium and phenom were the same when they were first released. The arch wasnt optimized but once they were they turned out to be great cpus. Dont be so quick to judge a book by its cover especially when its in rough draft form. Wait until all the mistakes are fixed and all the optimizations and revisions are made and the final copy is complete before you say it sucks. What will you say if/when the revised bulldozers start coming and they blow away SB or IB? What then? Then you either hide cause your too afraid to take the consequences or you just trun around and say that "I never said bulldozer sucks." Give it a break already. You guys act like you want AMD to fail when in fact if AMD succeeds, you get to benefit from it with better prices and/or better products. So instead of always stating how much you think AMD is worse than Intel, start cheering for them and if they come through, it turns around and you get to reap the benefits. Also, no one really cares about "your" opinions.Originally Posted by Vagrant Storm
I see there is still a little bit of grease and fur left from this horse.
Bulldozer doesn't really fit into the old classifications. It is a four module, eight thread CPU.
One module does not equal one core, but one module does not equal two cores as well.
It does have eight physical cores so AMD marketing isn't lying...but each pair of cores has to share a FPU so there isn't any way that it can function as a truly eight core CPU. No more than my CPU can if I turn on HT. Though the two processes are extremely different and each have their own pros and cons.
Joined
·
565 Posts
Quote:
![thumb.gif]()
yeah. i just saw this thread 10 minutes ago. i generally ignore them and keep going. however, im happy with my new toy. it may not be the fastest but im happy with my gaming and [email protected] experience. i'd never return my processor for an intel processor. ive spent the last 10-15 years going for intel solutions. change is good and new technology is good.Originally Posted by mikezachlowe2004
So you guys think that no one cares about bulldozer because this thread has only been out for 9 hours and since everyone is not on this site the same time you guys are and making posts taht you come to this conclusion?
Joined
·
2,997 Posts
You guys will want to wait for Bulldozer 20h-2Fh if you want any real improvement
Bulldozer 10h-1Fh only adds improved Branch Predictor and TLB Entries and a longer Store Queue and removes 4 cores and replaces them with a VLIW4 GPU
Bulldozer 20h-2Fh adds everything above minus the GPU + one more compute unit and Quad-channel and allows execution of Mov(and more) instructions in the AGLUs
Bulldozer 10h-1Fh only adds improved Branch Predictor and TLB Entries and a longer Store Queue and removes 4 cores and replaces them with a VLIW4 GPU
Bulldozer 20h-2Fh adds everything above minus the GPU + one more compute unit and Quad-channel and allows execution of Mov(and more) instructions in the AGLUs
Joined
·
1,728 Posts
Quote:
What are you talking about? 20-2Fh = Piledriver?Originally Posted by Seronx
You guys will want to wait for Bulldozer 20h-2Fh if you want any real improvement
Bulldozer 10h-1Fh only adds improved Branch Predictor and TLB Entries and a longer Store Queue and removes 4 cores and replaces them with a VLIW4 GPU
Bulldozer 20h-2Fh adds everything above minus the GPU + one more compute unit and Quad-channel and allows execution of Mov(and more) instructions in the AGLUs
1 - 20 of 77 Posts
Join the discussion
