CPU/Mobo AMD Bulldozer Discussion Thread
- Thread starter Hades.
- Start date
- Status
Well I don't see a need of jumping to either bulldozer or sandy bridge if you have a phenom II. The difference in performance is there but not jaw dropping.
In read world applications it hardly matters, shaving couple of seconds or bunch of them by spend k's of money doesn't make sense. Rather spend your money in upgrading components that really deserves ungrading.
Its not like Phenom II performance is puny in comparision to SB and Bulldozer. Overclock its and get a after market cooler or spend in WC kit. Even if bulldozer performs I won't be upgrading soon. have much important things to buy than spending on CPU that I hardly use except in coding,.
In read world applications it hardly matters, shaving couple of seconds or bunch of them by spend k's of money doesn't make sense. Rather spend your money in upgrading components that really deserves ungrading.
Its not like Phenom II performance is puny in comparision to SB and Bulldozer. Overclock its and get a after market cooler or spend in WC kit. Even if bulldozer performs I won't be upgrading soon. have much important things to buy than spending on CPU that I hardly use except in coding,.
When will you put your Phenom II on sale?hector19 said:
You really really don't need to upgrade. Unless you are doing something really really CPU specific. Are you? You don't have to upgrade for gaming at least. Just look at my specs man, I am the one who really does need to upgrade but I am still getting by fine. Have to reduce a bit of settings though.mrcool63 said:
You can upgrade. If you think you will gain from additional cores you should probably wait, otherwise i5 2500K FTWmrcool63 said:dude im running the adobe suite ie photoshop, premiere pro after effects. also avid composer sometimes. also conversion of videos.
so now tell me if i should wait
Zambezi branding and boxes revealed at E3 AMD Officially reveals Slides, Packaging and Live Demonstration of Bulldozer Zambezi-FX CPU's at E3 2011
Black Edition will come in metallic boxes ? :drool:
Black Edition will come in metallic boxes ? :drool:
vpraveenis
Explorer
^^ If they had the performance numbers to beat Sandybridge, they would have definitely leaked the benches to get people to hold off from buying Sandybridge. Since they are still tweaking and trying to improve performance, they are keeping a tight hold on the performance figures I guess.
mihir.khatwani
Contributor
MD claims 80% of two full cores.
Each module has two sets of pipelines (the reason it shows up as two cores) Shares some upper end and the L2 cache. How the performance comes into play is how it will runs threads.
A module will not split up a single thread on both sets of pipelines. So a single thread will run with the full upperend and a set of pipelines and the whole 2mb. of L2 cache (which is a good thing).
But the question is how it will splits up threads. Take two threads, will it allocate the second thread to the same module and use the second set of pipelines (80% of two full cores) or will it send the thread to the second module (so it can use 100% resources of the upperend and L2) of both modules. The later would be better performance.
Say you have a 4 module running 4 threads, it would be better to run one thread on each module, then start using the other set of pipelines when thread 5/6/7 and so on kicks in.
If it just has one thread on a module. It gets to use the whole upperend a set of pipelines and the full L2 cache.
If it running two threads is has to share the Fetch/Decode/FP Scheduler and the L2. The L2 really doesnt matter that much though because each module has a full 2mb. of L2 cache.
Each set of pipelines has their own Integer scheduler/128 bit FMAC and L1 cache.
Thats the reason they say a module has the 80% preformance of two full cores when running two threads. But thats alot better then Intels Hyper Threading.
Each module has two sets of pipelines (the reason it shows up as two cores) Shares some upper end and the L2 cache. How the performance comes into play is how it will runs threads.
A module will not split up a single thread on both sets of pipelines. So a single thread will run with the full upperend and a set of pipelines and the whole 2mb. of L2 cache (which is a good thing).
But the question is how it will splits up threads. Take two threads, will it allocate the second thread to the same module and use the second set of pipelines (80% of two full cores) or will it send the thread to the second module (so it can use 100% resources of the upperend and L2) of both modules. The later would be better performance.
Say you have a 4 module running 4 threads, it would be better to run one thread on each module, then start using the other set of pipelines when thread 5/6/7 and so on kicks in.
If it just has one thread on a module. It gets to use the whole upperend a set of pipelines and the full L2 cache.
If it running two threads is has to share the Fetch/Decode/FP Scheduler and the L2. The L2 really doesnt matter that much though because each module has a full 2mb. of L2 cache.
Each set of pipelines has their own Integer scheduler/128 bit FMAC and L1 cache.
Thats the reason they say a module has the 80% preformance of two full cores when running two threads. But thats alot better then Intels Hyper Threading.
- Status