Quad FX VS Intel QX benches......

mzs_biteme said:
With some QX overclocking...: http://hardware.thgweb.de/2006/11/30/amd_4x4_athlon_64_fx74_quadcore_vierkern_cpu/page23.html. They managed to OC FX by an astounding 125MHz... :eek:
rofl2.gif
, but I digress...
horse.gif

Yeah, that cracked me up!
 
Astrogiblet said:
Wow.. according to that my FX-62 out performans two fx chips in 4x4. Wow.

Yes but if that's the case then my E6600 does too LOL!
 
Astrogiblet said:
Wow.. according to that my FX-62 out performans two fx chips in 4x4. Wow.

Which shows, at least to me, how much this platform needs to mature before it's a really good solution...

It's too bad, I really like AMD...
 
Donnie27 said:
Yes but if that's the case then my E6600 does too LOL!


Yea, if you overclock it enough. :p


Arcygenical said:
Which shows, at least to me, how much this platform needs to mature before it's a really good solution...

Yea.. its really quit shitty isnt it? I mean.. two chips that are individually much more powerful then the fx-62 (or should be) are getting out performed by the single, less powerful chip.

I'm thinking they have some bandwidth or throttling issues.
 
Astrogiblet said:
I'm thinking they have some bandwidth or throttling issues.

One issue they have is the double memory controllers, which means it's NUMA (Non-Uniform Memory Architecture).
This means there no longer is a single pool of system memory, but each CPU has its own pool, connected to its local memory controller.

This can be both an advantage and a disadvantage.
The advantage is that each CPU can access its own memory without having to share any resources with the other CPU.
The disadvantage is that one CPU has to 'borrow' the other CPU's memory controller (locking the other CPU out at the same time), if it has to access the other memory.

For servers, it is generally an advantage, because you let each CPU handle its own independent process.
For desktops, you generally just want to run one process as quickly as possible, and leave the rest idle in the background... So it's a disadvantage, because if you want to use both CPUs for that process, they have to share.

This means that the OS and software has to be NUMA-aware, in order to try and keep the sharing of memory controllers to a minimum. Since the quad FX is the first NUMA-system ever for the desktop, there's basically no NUMA-aware desktop software at all.

As a sidenote, it basically shows how strong Intels FSB+memory controller really are. In Intels case, the cores ALWAYS have to share the bus and memory controller, whether they're in the same physical CPU or not. Apparently this is more efficient than AMD's HTT and on-die controllers, which were often praised.
 
LOL, they used the node interleave memory layout. How lame. No wonder performance is so low. Ask anyone with an Opteron system and they can explain why this is so important. :rolleyes:
 
visaris said:
LOL, they used the node interleave memory layout. How lame. No wonder performance is so low. Ask anyone with an Opteron system and they can explain why this is so important. :rolleyes:

Pray tell, why?
 
I really wonder what AMD's reps and CEO's are thinking right now..

It could be one of many things:

1) We tried our best. We'll do better next time.

2) Our buyers are fools, if we brought out a half as good system they would still buy it!

3) What did we do wrong? We did everything RIGHT and it didnt work out good..

4) They will see, when quad father comes out..

Or if they are completely ignoring the benchmarks and thinking "Man, our setup is FAR superior to Intels."
 
Astrogiblet said:
Yea, if you overclock it enough. :p

Yea.. its really quit shitty isnt it? I mean.. two chips that are individually much more powerful then the fx-62 (or should be) are getting out performed by the single, less powerful chip.

I'm thinking they have some bandwidth or throttling issues.

We tested mine and another E6600 vs Two FX-62 and there wasn't a hill of beans difference between the 4 systems. The switch running all kinds of apps slightly faster than eachother. And of Course the only overclock need was 200 more MHz to clean the FX-62's clock! That's after overclocking it to 3GHz HEHEHE! With both running at 3GHz, it was a blowout for the E6600.

I don't know any of my AMDroid friends still buying anything AMD right now. Yup Quad-FX is shit alright.

http://techreport.com/reviews/2006q4/quad-fx/index.x?pg=5
 
Can someone kinda sum it up for me? Like ____ kicks ____ ass or just sum it up. It was kinda confusing....thx
 
I've always been a double agent so moving to Intel will be easy.
It's time for me to move to Intel QC as soon as they get cheaper.
My price range is around $300-$600, no more, so I still got some time.



wait
can someone link me to what cheaper QC Processors Intel is releasing.
 
This is the Intel forum, topics about Quad FX have no place here. Trolls.
 
http://www.gamepc.com/labs/view_content.asp?id=qx6700&page=1

GamePC said:
The QX6700 will match the performance levels of two Xeon 5150 (2.66 GHz) processors, each of which sell for around $800 currently, which makes the QX6700 processor look like a "bargain" in comparison. The performance levels of this chip in professional level applications is un-rivaled in the single-processor space, and it looks like it will be quite some time before AMD has a suitable competitor against Intel's "Kentsfield" based processors. While some may call Intel's implementation less than optimal, the truth is the chip simply works now, delivering amazing performance in applications which can truly take advantage of it.

Are folks still wondering why AMD kept so quiet about it? Many of us said if they had something, just like with Intel, it would been leaked. If K8L is kick-ass, we'll see a few show up at places like Xtremesystems.org:)
 
Back
Top