Bookmark and Share


UPDATE: Based on comments made by DonanimHaber web-site as well as some other sources, the benchmark results are either simulated or are fake.

The first performance benchmark results of AMD FX-series "Zambezi" microprocessors this week finally emerged on the Internet. The engineering sample of AMD's eight-core chip that was supposed to be the flagship offering - the model 8130P - could demonstrate performance that is comparable to that of Intel's flagship quad-core solution.

DonanimHaber web-site has managed to obtain an engineering sample of AMD FX-8130P (eight cores, 3.20GHz clock-speed, 8MB L2 cache, 8MB L3 cache) and run a number of benchmarks using Gigabyte GA-990FXA-UD5 mainboard, 2GB of DDR3 memory (4GB was installed, only two seen by the system) and an Nvidia GeForce GTX 580 graphics card. The chip was an engineering sample and most likely its performance will not reflect performance of AMD's top-of-the-range Bulldozer chip for desktops; still, something about real-world performance of the forthcoming FX-series chips becomes a little more clear.

The obtained benchmark results of AMD FX-8130P are as follows:

 AMD believes that its multi-core Zambezi FX CPUs will allow it to compete head-to-head with Intel's high-end Core i-series "Sandy Bridge" processors that can sell for as much as $300 and more per chip. The model FX-8130P, which may never see the light of the day as AMD is currently redesigning its Zambezi processors to get higher performance, can fully compete against Intel's Core i7-2600K chip (Sandy Bridge) based on the presented results. Unfortunately, AMD has a quad-core chip which is supposed to significantly outperform the rival from Intel's camp.

Given the fact that AMD is reworking its Zambezi these days and that the chips will only become available in late August or September, it is likely that the eight-core microprocessors by the Sunnyvale, California-based developer will be faster than they are now and the company will try to ensure that they offer better speed than Intel's Core i7-2600K. The big question that remains now is whether the new products by AMD will be comparable to Intel's six-core "Extreme" chips (Gulftown) and whether they will be competitive against Intel's forthcoming Sandy Bridge-E microprocessors with six or eight cores, quad-channel memory controller and other improvements.

AMD did not comment on the news-story.

Tags: AMD, Zambezi, Intel, Core, Gulftown, Bulldozer


Comments currently: 47
Discussion started: 07/12/11 08:42:21 PM
Latest comment: 09/13/11 10:37:34 AM
Expand all threads | Collapse all threads


Ivy bridge will come out soon then Amd will be behind again as always.
4 4 [Posted by: 3Dkiller  | Date: 07/12/11 08:42:21 PM]
- collapse thread

Don't be so sure, it's only an engineering sample (B1). In a few weeks, B2 will enter the stage
4 1 [Posted by: Rares  | Date: 07/12/11 10:18:20 PM]
I've read the article and using google translator I understand that the sample they used has a problem with L3 cache. So there is certainly room for improvement. Always assuming this sample has problems with L3 cache.
On a side note, they have a screenshot of CPU-Z version 1.57.1 while support for AMD Bulldozer was added on version 1.58.
2 2 [Posted by: Filiprino  | Date: 07/13/11 12:27:40 PM]
Ivy Bridge has been pushed back to Q2 of 2012.
3 1 [Posted by: SteelCity1981  | Date: 07/12/11 11:54:06 PM]
Smart move by Intel. Why must they ship on time when there's no competitor? BD absolutely have to beat SB. If they do, It won't hurt SB much because Intel already "sold it out", on contrary if they don't, then it will hurt Fan boy the most.
2 0 [Posted by: jpunk  | Date: 07/13/11 07:08:10 PM]
Do you remember a toaster called Pentium 4 and the Athlon XP and Athlon 64? Do you remember the Pentium 3 and the original Athlon?
4 0 [Posted by: Raptor  | Date: 07/13/11 03:14:40 AM]
but soon after that the second generation bulldozer will out.
in fact IB is much farther out then bulldozer next gen is to IB.
2 1 [Posted by: Countess  | Date: 07/14/11 03:00:16 AM]
They Is A Time 4 Everything Just Want You 2 Know That
0 0 [Posted by: blazzin  | Date: 09/13/11 10:37:34 AM]

show the post
2 7 [Posted by: ericore  | Date: 07/13/11 12:31:48 AM]
- collapse thread

AMD 8 core chip is in reality an 4 core chip. It is basically as much an 8 core chip as the 2600K. Every BD module runs 2 threads but most of it is shared and only the ~5% integer part is duplicated. HyperThreading also increases the transistor number, but probably not as much...
8 core chip is only a marketing thing. It is a 4 core chip!
2 3 [Posted by: Raptor  | Date: 07/13/11 03:22:05 AM]
If you correct people raptor you really should be correct with that correction.

The top end bulldozer is a 4-module chip. Each module contains 2 integer execution area's each with its own instruction scheduler, 2 ALU's, 2 AGU's and a shared L1D cache. This is what AMD are calling a 'core'. Shared between each 'core' within the module shares a front end (fetch/decode), the L1I cache, L2 cache, and 2 128bit FPU's. The 2 FPU's can either have 1 scheduled to each 'core' or be combined for a single 256bit instruction.

AMD are calling each module 2 'cores' because what they have created is far closer to two cores than 1 core with something similar to hyperthreading. Hyperthreading purely operates on the front end allowing for 2 threads to be dispatched to a set of execution units allowing higher utilisation of those units. AMD's is creating a bigger front end, duplicating the Int execution units and sharing the FP executions units. In reality this make each module far closer to being 2-cores than to being a single core.

One of the big reasons this is only a small increase in die area to get the extra 'core' is that even if it increases the size of a module by say 25%, there is still the L3 cache and I/O taking up space, which means the total increase won't be 25%.

Now you may disagree with AMD calling each module 2 cores, but that doesn't in any way make it a 4 core chip 'in reality' as you stated. It's a complex issue so please don't try to just run roughshod over that by saying "8 core chip is only a marketing thing. It is a 4 core chip!"
3 2 [Posted by: dejavu  | Date: 07/13/11 04:48:33 AM]
Of course that BD module is not a regular core, but I believe that you would agree that it is far from being 2 regular cores. So, when people compare a 4 core / 8 thread intel CPU and say that BD is very inefficient because it is an 8 core processor, it is completely unfair.
This module architecture is AMD's response to HT, because it uses very little die area and should perform a lot better than HT.
When the CPU is released, my guess is that single-threaded performance should be very competitive and also multi-threaded, but to a 4c/8t CPU but not to a 8c/16t CPU.
Now you are right in the way that, now it is a lot more difficult to say what is a core, but in my opinion, one module is one core, whatever AMD says about it. I even think that they made a mistake of calling it 2 because of discussions and comparisons like these.
7 1 [Posted by: Raptor  | Date: 07/13/11 07:02:53 AM]
BD Module = 2 physical integer cores with 0 hyperthreading.

Intel Core = 1 physical core with 1-2 threads

Since the common definition of a cpu core is a physical core on the chip. BD has 4 2 core modules making it a 8 core cpu, where intel cpus have at most 6 cores with a bunch of threads.

The AMD module is nothing like the intel cpu structure in that it is all physical cores. Thus saying the BD module is equal to one intel core is completely incorrect because they are two completely different core architectures.

2 4 [Posted by: veli05  | Date: 07/13/11 08:07:14 AM]

0 2 [Posted by: Green lantern  | Date: 07/13/11 10:03:37 AM]
No ... AMD has stayed away from hyperthreading for a while now. To say that the BD module is reverse hyperthreading is not a great way to describe it. AMD instead went the route of trying to pack as many physical cores into the dies as possible to provide the consumer with true multi-process computing solutions, which has been relatively successful. AMD's module architecture is completely different than any previous die architecture to date, so comparing it to hyperthreaded architectures is impossible because the architecture is in a class of its own.
2 2 [Posted by: veli05  | Date: 07/13/11 11:51:09 AM]
reverse hyperthreading is a myth. it doesn't exist.

the concept behind bulldozer is taking 2 cores, integrating them as much as possible, sharing parts when its efficient to do so, and removing all redundant and inefficient parts to achive as much performance as possible with a few transistors as possible.
4 1 [Posted by: Countess  | Date: 07/14/11 05:03:28 AM]
I'd have to agree with you. They should have just called it a core for clarity sake. However, it won't be long before Intel also has modules.
1 3 [Posted by: sedaine  | Date: 07/14/11 03:26:19 AM]
"but which ARM later revealed it was developing pretty much the same technology." Absolute bullshit. You obviously have no idea what you are talking about.
2 1 [Posted by: lol123  | Date: 07/13/11 08:56:13 AM]
What are you smoking? A 20% performance increase based on the same arch is a solid improvement. You say it's pathetic then I guess you also must think that the Phenom II's 20% performance increase over the first gen Phenoms was pathetic to right...
3 0 [Posted by: SteelCity1981  | Date: 07/13/11 02:25:41 PM]

This is something that AMD promised.
2nd generation is to be scheduled for 2012. AMD currently refers to this as Enhanced Bulldozer

Desktop Performance market (Corona platform): Zambezi will be replaced by Komodo (8 cores)
Mainstream market (Virgo platform): The Stars-based Llano Fusion APU line will be replaced by Trinity, Weatherford, and Richland
1 0 [Posted by: Blackcode  | Date: 07/13/11 12:50:07 AM]

This looks pretty close to what AMD has been promising. I'm looking forward to seeing more in-depth testing of release chips.

One minor correction for the article:
most likely its performance will not reflex performance of AMD's top-of-the-range

Should be:

most likely its performance will not reflect performance of AMD's top-of-the-range

2 0 [Posted by: Prosthetic_Head  | Date: 07/13/11 01:50:31 AM]

Interesting. Core i7 2600k runs at 3'4Ghz and Zambezi 8130-P at 3'2Ghz, still it's faster and on well threaded apps like x264 is as fast as a 990X running at 3'46Ghz, a CPU that has 6 cores and 12 threads. 8 full threads > 12 shared threads.
The problem will surface when Intel launchs its 6 core sandy bridge for LGA2011 on Q3-Q4 too, although on X264 if the encoder implements XOP code paths I think there would be no difference at all, even AMD would have better performance on that particular app or any other that benefits from XOP or the other new instructions it supports.

AMD should up the number of cores to equal the number of threads of the competition.
0 2 [Posted by: Filiprino  | Date: 07/13/11 05:10:23 AM]
- collapse thread

Compare 1 BD module with 1 Intel core with hyperthread.8130P in fact have 4 BD module.
BD have diffrent approach to multithread than Intel.
1 0 [Posted by: Blackcode  | Date: 07/13/11 05:50:06 AM]
Well, if we stick with AMD naming scheme, they're promoting this as 8 cores CPUs without hyperthreading.
Of course, if we count modules instead of integer cores, yeah, 4 Intel cores are bested by AMD's 4 modules, and even last generation Intel 6 cores at same frequency could be beaten by AMD's 4 modules where today's Intel Sandy Bridge 4 cores can't get anywhere near of the 990X performance with stock frequencies.
If only AMD had 6 modules going out of the gate ASAP.
0 1 [Posted by: Filiprino  | Date: 07/13/11 08:45:16 AM]
Actually, on x264, AMD's 4 modules are faster clock for clock than 990X. If we look at Core i7 970 which is the same but running at 3.2Ghz (same as Zambezi 8130) instead of 3.46Ghz, i7 970 gets 44.4FPS while 8130P is reported to get 45.4FPS. And believe me, 1FPS on x264 is a meaningful difference. To get 2FPS more on my Q6600 I had to increase the clock speed by 200Mhz, in percentage it was a linear gain (old clock/old FPS versus new clock/new FPS).
0 1 [Posted by: Filiprino  | Date: 07/13/11 10:33:12 AM]
The module, described as two cores, can be contrasted with a single Intel core with HyperThreading. The difference between the two approaches is that Bulldozer provides dedicated schedulers and integer units for each thread, whereas in Intel's core all threads must compete for available resources, except for the individual thread state information.
0 1 [Posted by: Blackcode  | Date: 07/13/11 12:30:36 PM]
SB High OC is the greatest reason why I bought it. So why must I compare it @stock Clock? If BD can beat SB with their each Highest (Stable daily system) Clock, then without a doubt I'll return to AMD's Side.
1 0 [Posted by: jpunk  | Date: 07/13/11 07:19:44 PM]
BD has been already proved to clock at least as high as SB on air cooler.
0 2 [Posted by: Filiprino  | Date: 07/14/11 06:11:27 AM]
Can You give the me Link? I Used SB Daily just @4.6Ghz - @4.8Ghz 1.35V with Frio OCK. If BD can run @4.6Ghz Daily (Doesn't matter if it run 4 Cores), I'll buy it for sure.
1 0 [Posted by: jpunk  | Date: 07/14/11 05:45:33 PM]
Sandy-E delayed till next year... aka 2011 will be released in 2012
1 1 [Posted by: Patriot  | Date: 07/13/11 08:02:33 AM]
As far as I know 6 cores will be launched on Q4 2011 and 8 cores on Q1 2012.
LGA2011 is to be launched on Q4 2011 and LGA 1356 on Q1 2012.
LGA 1155 will stick with 4 cores, which in my opinion sucks big time.
1 1 [Posted by: Filiprino  | Date: 07/13/11 08:48:11 AM]

show the post
0 4 [Posted by: 63jax  | Date: 07/13/11 11:02:31 AM]
- collapse thread

They do call them modules ...
1 1 [Posted by: veli05  | Date: 07/13/11 11:39:56 AM]
i mean to call them ONLY Modules and get rig of the Cores.
1 2 [Posted by: 63jax  | Date: 07/14/11 12:29:59 AM]

AMD has proved that they could do amazing jobs, but nowadays the nanometer is the real battle for them -I think- much more than past years and not even just in the mobile industry but in every aspect of creating chips it is a much more serious battle indeed. Intel seems to be a giant monster in this battle and AMD works so hard with respectable innovations in the chip architecture.
So, It looks they do it again. In past few years AMD was in a place that even SAMSUNG interested to acquire AMD. To me with respect to Korean company, It was a shame! But today AMD is still fighting for good and if just they could do some more in High-K and nanometer techs more quickly there is no doubt to me that AMD is winner again.
And one more thing, absolutely NO WONDER if hearing about Intel's "SOMETHING TECHNOLOGY" that will not be anything but just a copy of AMD's beautiful work against Intel's unlikely HT just like on-chip memory controller in the past. Please if you want to Reply to me do that without the word 'fanboy' or etc...
These words was my opinions anyway.

2 1 [Posted by: Pouria  | Date: 07/13/11 01:21:06 PM]
- collapse thread

Intel copied AMD by putting an intergrated memory controller in the CPU. They will most likely copy AMD again in changing their chip design to a modual approuch, The thing is with AMD and Intel, they have a cross lic agreement allowing each company to copy tech from each other. The only thing AMD lacks is the Fabrication that Intel has... This lack will change in the next few years after the 3 NEW fabs GF has building come online, the NewYork Fab8 will be the most advanced fab in the world... things dont look to bad for AMD going into the future its almost at at a solid fotting in the market give it another few years we may see a 50/50 cpu market place .. here's hoping it would do wonders for the industry.
3 0 [Posted by: vid_ghost  | Date: 07/13/11 04:38:57 PM]
Llano is 32% more dense then Sandy Bridge fabrication tech lacking


Llano and Sandy Bridge are both 32nm
and Zambezi has a theoretical max of 1.61~ BILLION Transistors
(Die size/picture calculation it might be smaller than this)
0 1 [Posted by: seronx  | Date: 07/13/11 08:32:57 PM]
I though Intel has the First Intergated memory COntroller but cancel due to some reason i think that was 1999 or 2000
0 1 [Posted by: xentar  | Date: 07/13/11 11:52:39 PM]
I though Intel has the First Intergated memory COntroller but cancel due to some reason i think that was 1999 or 2000

Oh, you didn't recall Itanium?
Intel was the first to create 64bit Processors also.
They even had a partnership with HP for that.
So why Microsoft added AMD64 for x64 binaries into Windows since? Instead of IA-64....
You can invent things. Everybody can. But you have to make them real out of your mind to make the innovation.
1 0 [Posted by: Pouria  | Date: 07/14/11 02:04:24 AM]
they put in there OS because the AMD version is backward compatible and itanium isn't so ytou could run 64bit and 32 bit on the older amd chips and anly 64bit at the itanium one.
0 0 [Posted by: massau  | Date: 07/15/11 12:12:01 PM]

Intel will be on 22nm tri-gate tech come the start to mid next year Q2 2012 AMD will be on 22nm planer ( flat older tech ) come Q3 2013 and wont be on tri-gate until maybe 2014-2015 when they move onto 14nm-11nm

Lacking they are by ALLOT! .. Unless Global foundries can speed up their timelines.
1 1 [Posted by: vid_ghost  | Date: 07/13/11 10:42:30 PM]
- collapse thread

AMD (or more accurately global foundries) will use FD-SOI for 22nm instead of 3d transistors. it offers many of the same benefits, and has advantages and disadvantages compared to intel 3d transistor.

and 3d transistors actually aren't all THAT great. they offer advantages in terms of power usage OR clock speed(not at the same time).
but the later only at low voltages (sub 1 volt) as compared to the previous 32nm.

at voltages used in CPU's (the voltages we care about) the advantages as compared to a normal die shrink aren't very significant.
0 1 [Posted by: Countess  | Date: 07/14/11 05:12:02 AM]

Hybrid products due to come out should be interesting. 2012 looks like would be interesting for AMD since we get Bulldozer APU's with faster GPU's then current Llano.

I suspect performance will be somewhere between 6850 - 6950 for GPU's. Have you any idea what the implications of that are?

If AMD has a 20% market share for desktop and assume not many people install discrete graphics - then 20% of desktop users can start doing interesting things like watch 3D at 1080P. In fact, you could watch content at greater than 1080p with ease.

For the average user, this would be the desktop to get. Much like Llano is the best thing to get for an average Laptop user that probably uses their laptop for e-mail, browsing, watching videos and encoding.
0 0 [Posted by: sedaine  | Date: 07/14/11 03:47:16 AM]


OBRosky sent photoshopped results to Donanimhaber

OBRosky also photoshops everything on his blog and is an Intel Evangelist

Zambezi performance unknown
1 0 [Posted by: seronx  | Date: 07/14/11 03:55:14 PM]
- collapse thread

He did not claim photoshopped he claimed to have generated the data on a massively over clocked CPU... Performance remains unknown but it appears less than encouraging.
0 0 [Posted by: jumpingjack  | Date: 07/15/11 08:36:21 PM]


letter from the site about the fakes.
3 0 [Posted by: RtFusion  | Date: 07/14/11 09:12:05 PM]

Shmintel fan boys not so smart would pay so much more just to get around 20% more power for more then 1g lol not too bright now. Must have failed math
0 0 [Posted by: Atprod99  | Date: 09/04/11 06:52:14 PM]


Add your Comment

Related news

Latest News

Wednesday, November 5, 2014

10:48 pm | LG’s Unique Ultra-Wide Curved 34” Display Finally Hits the Market. LG 34UC97 Available in the U.S. and the U.K.

Wednesday, October 8, 2014

12:52 pm | Lisa Su Appointed as New CEO of Advanced Micro Devices. Rory Read Steps Down, Lisa Su Becomes New CEO of AMD

Thursday, August 28, 2014

4:22 am | AMD Has No Plans to Reconsider Recommended Prices of Radeon R9 Graphics Cards. AMD Will Not Lower Recommended Prices of Radeon R9 Graphics Solutions

Wednesday, August 27, 2014

1:09 pm | Samsung Begins to Produce 2.13GHz 64GB DDR4 Memory Modules. Samsung Uses TSV DRAMs for 64GB DDR4 RDIMMs

Tuesday, August 26, 2014

10:41 am | AMD Quietly Reveals Third Iteration of GCN Architecture with Tonga GPU. AMD Unleashes Radeon R9 285 Graphics Cards, Tonga GPU, GCN 1.2 Architecture