AMD FX-4100 Bulldozer

[crazycheese]

What a shame.

[Medallish]

It actually did better than I expected, yeah it's about 500MHz faster than the Llano in your test, but it was clearly able to follow and beat the Quadcore Llano into multithreading, I think it's fair to say it is a Quadcore based on what I saw JF_AMD write at Semiaccurate on how we technically define "cores" And unlike Hyperthreading you can't really turn it off, it would be quite a waste if you did, and from what I see in the tests of 8150, once it shows it's Multithreaded teeth you will see it outperforming Hyperthreading, as it should since it requires more die. If we see huge improvements in Ubuntu, I might get a 8150 anyway, although currently I think I will sit this one out and wait for Piledriver, maybe I'll get a Thuban for my Crosshair V board.

[schmidtbag]

The facts people are missing is this:
The CPU is a dual core, but has the physical elements making it seem like a quad core, but it really isn't. I believe the way HT works is it's physically 2 cores with each core running 2 threads at the same time, hoping they don't collide. Bulldozer CPUs are 2 physical cores that run 2 threads in physically separate locations.

The next fact is AMD removed the total amount of instructions per clock in order to increase clock speed. Thats why if you were to take a 3.6Ghz single-core Phenom (if that even exists) and have the FX-4100 face off in single-threaded tasks, the Phenom will immensely outperform, even though everything about it is supposedly worse.

I may be wrong about this, but I believe that if a single-threaded task is run on 1 core, the other thread of that core can't be used. Whereas with HT, there is no restriction. This can cause significant slow-downs.

[-Cas-]

I wondered if this image below is from the Anandtech review of the processor helps clear up some of the confusion with how the processor can be optimised. The issue seems to be getting 'turbo core' enabled and directing threads with shared data to the same core.

[PsynoKhi0]

I may be wrong about this, but I believe that if a single-threaded task is run on 1 core, the other thread of that core can't be used.

Hmm do you really mean core? or module? If it's the latter, I think you can have another thread running on the other core of the same module, but then resources have to be shared and some workloads suffer, e.g. FP-heavy stuff.

[blackshard]

this cpu will speed up by 3% for the cache kernel patch and 10% for the scheduler kernel patch and maybe 20% for the compiler patches.
also as i know amd will upgrade the micro code in the bios to fix some speed bugs.

but yes technically this cpu is a dual-core with some extra Integer help units.

If some one need a true quatcore with greater speed from amd he can get the Opteron 6204.
but yes 400€ isn't cheap.

A dual core with 2 Millions transistors? Quite huge for being a dual core...

[pingufunkybeat]

Anyway, it looks comparable to an i3 in terms of price and performance.

I'd like to know more about power consumption -- how do all these extra transistors affect it?

If power consumption is comparable, then this particular processor is OK. Even if there are doubts is the architecture can compare with intel's high-end offerings in the near to mid future.

[Qaridarium]

A dual core with 2 Millions transistors? Quite huge for being a dual core...

it dosn't matter how much transistors it is in FACT a dual-core.,

[Michael]

Anyone happen to know if there's any newer kernel patches than http://thread.gmane.org/gmane.linux..../focus=1170744 for Bulldozer? Google hasn't turned up any others, just checking.

[WorBlux]

it dosn't matter how much transistors it is in FACT a dual-core.,

Sound more like a dual-core-plus-dual-half-cores. Basically they added more of the circuits that are most commonly used, especially under a server type load, but probably not so good for a lot of multimedia competing for the same FP unit.