Bulldozer

Well three things are wrong with Bulldozer/Zambezi. One is the L1/2 cache are slow. I dont see any way to correct it without a major stepping or not until Piledriver. Second window does not know what a module is. Considering this has been in development for 4 years, thats unbelievable. The third thing is when there is 4 or less threads, a thread should have been able to run on both sets of pipelines on a module.

And a fourth reason.
http://www.xbitlabs.com/news/cpu/di...x_AMD_Engineer_Explains_Bulldozer_Fiasco.html
 
Last edited:
So Deneb and Thuban architectures are better than Zambezi? Is Llano architecture design the same as Thuban that can run off DDR3 1866 memory?
 
Troncoso said:
What did you have your AMD CPU's in? AMD is just as reliable as Intel, it's rarely the processor that's even the issue whenever you have problems.
Click to expand...

My triple core phenom was in a Biostar board if I remember correctly. My 955 was in a Gigabyte MA790XT-UD4P board. The 965 was also in the Gigabyte board.
 
kennebell347 said:
My triple core phenom was in a Biostar board if I remember correctly. My 955 was in a Gigabyte MA790XT-UD4P board. The 965 was also in the Gigabyte board.
Click to expand...

I have had nothing but good experiences with AMD minus one motherboard. It was a cheap Foxconn motherboard that had problems. It had nothing to do with the AMD processor.

I've owned many desktop Advanced Micro Devices processors:

Athlon XP 2200+ Single-Core
Athlon 4000+ Single-Core
Athlon 4600+ Dual-Core
Phenom 9550 Quad-Core
Phenom II 940 Quad-Core
Phenom II 945 Quad-Core
 
2048Megabytes said:
So Deneb and Thuban architectures are better than Zambezi?
Click to expand...

I wouldnt say that. AMD can do improvements to the Zambezi. It just has faults in the design that should have been worked out before it was released. From what it looks like AMD just didnt have the resources in research and development to pull off Bulldozer as they once wanted it to be. The Deneb/Thuban is pretty much tweaked as far as it can go.

Kinda like the same that happen with Phenom I. Where it should have performed like the Phenom II. I would say they can get better performance with Bulldozer with a few bios/Driver/Patch updates. Then probably a new stepping later on. But my guess Piledriver will perform close to what they wanted Bulldozer to be at the start. Kinda funny they will call it Piledriver, did not want it to be a rerun to call it Bulldozer II.


2048Megabytes said:
Is Llano architecture design the same as Thuban that can run off DDR3 1866 memory?
Click to expand...

Its just a die shrink of the Athlon II core with double L2 cache. Not for sure about the memory controller. If its just a updated Deneb controller or a twist of the Bulldozer controller.
 
i just hope they have the updates, and that the profits from the oem sales of the 8150 (although newegg did sell out day 1 so they obviously were selling fast so prophits from that too) and use the money to make a few updates to the 8170 before launch early next year, maybe AMD will let them go in and hand sculpt some of the transistors since apparently automated designing was an issue
 
It would be nice if They would work on getting all these issues fixed, and Piledriver out around CES 2012. It was a real disappointment to me to see the FX 6100 getting wrecked in benches by Thuban. I'm not giving up on bulldozer though, I would rather buy an AMD product than Intel. I think we all know this probably won't happen, but one can hope.
 
Last edited:
StrangleHold said:
Its just a die shrink of the Athlon II core with double L2 cache. Not for sure about the memory controller. If its just a updated Deneb controller or a twist of the Bulldozer controller.
Click to expand...

The AMD A8-3850 Llano Quad-Core Processor marks are pretty impressive to me. The marks are estimated around the overall same processing power as a Core i7 920 Quad-Core Processor and only running at 100 Watts.
 
StrangleHold said:
I dont know. Here the A8 3850 performance is pretty equal to a Athlon II X4 635, both run the same clock speed. Even with the 3850 double L2 cache size.
http://www.anandtech.com/show/4448/amd-llano-desktop-performance-preview/2
Click to expand...

There are people questioning these marks. They are wondering if DDR3 1866 or DDR3 1600 memory was used with the A8-3850 Processor or DDR3 1333. They do not say what memory was used in the tests. DDR3 1866 would definitely push the marks higher on the A8-3850 Processor.
 
mihir said:
Any patch released for Bulldozer yet?
Click to expand...

Havent heard of one yet.

2048Megabytes said:
There are people questioning these marks. They are wondering if DDR3 1866 or DDR3 1600 memory was used with the A8-3850 Processor or DDR3 1333. They do not say what memory was used in the tests. DDR3 1866 would definitely push the marks higher on the A8-3850 Processor.
Click to expand...


The Athlon II can run 1600. It might gain alittle with 1866. Most of the small performance gain is from the die shrink and probably a few tweaks.
 
Last edited:
nothing on updates yet, but anandtech has a piece on the b3 revision in the works, but said since c3 on the phenom ii x4's waited 9 months to not wait for it
 
so the 8170 would be b3, makes sense really, and if they get those damn patches out that thing could be killer.
 
From what I understand the B3 is suppost to clock alittle higher and use less wattage. The L1 and L2 cache probably wont be fixed till Piledriver.
 
we can only hope, god from what i've read why the hell would the stupid a$$es at amd make the techs go to an automated transistor builder or whatever, something said it was 20% slower and took 20% more die space too
 
might be smarter to get a 1100t for the same, or a 1090t for a little less since a simple multiplier change will make it a 1100t
 
You must log in or register to reply here.
Back
Top