The thing is BD is a server CPU first then consumer CPU second. BD was made to offer more cores than Intel as a solution to Intel's HT. As to why AMD wants to offer more cores is because BD is targeted at the enterprise market that does a lot of virtualization that according to what AMD believes is a good target for them. Even on the server side there are some situations that are suited for Intel and AMD separately. At the enteprise level, having more cores is never a disadvantage, the same applies for Intel.

Canned software or whatever software you buy or illegally download cannot make good advantage of more cores than what they are programmed for but customized software that are not commonly used by average user at all can handle the extra because they are customized to take advantage of it. OpenCL is a good step to further take advantage of multiprocessing but its going to take time for software developers to make them.

If you're strictly a gamer then you can just settle for how many cores the game may support. Having a quad core at least is a better choice than a dual core especially if multitasking. Try using a single core now and you'll know what I mean, have to wait a task is completed before another task can be executed. If you really want a powerful dual core that overclocks like mad then get the Core i3 2120K if it ever comes out.

The price of the Core i3 2120K is not even half of the Core i5 2500K and from what I've heard the Core i3 2120K is just a 4 core binned CPU with 2 disabled to give headroom for the other 2 to overclock. The Core i5 2500K is still the better buy. I'm not too sure how much performance gain IB will be but may not be a lot. Performance/watt will increase no doubt which is really the key focus of Tri Gate. So actually you can be satisfied with the Core i5 2500K already and upgrade something else like a fast SSD if you don't have one currently.

You mentioned that your software doesn't take advantage of quad cores but I'm talking about multitasking which refers to running multiple different programs at the same time that would take advantage of a quad. Lets just say you're running 2 of the programs that you mentioned simultaneously. So if 1 of your programs can only make use of 2 then that means you can run 2 at the same time. As for the case of your Core i5 2500K, if 2 cores are occupied with the program that you run then you have 2 more cores to use for browsing or watch a movie while you wait for it to complete(if it takes hours).

Maybe the stepping is an issue but some people suggest that AMD is production capacity constrained since GloFlo has to manufacture Llano to meet the demand leaving no slot to manufacture Bulldozer.

Why not get a SB now?

You're not going to get a huge turbo just because Intel is selling a quad core as a dual core. You can overclock a Core i5 2500K as much as a Core i3 2120K because all SB processors have a max ceiling. No doubt the Core i3 2120K is marketed for the overclocker but it will still be sold with a lower TDP compared to a Core i5 2500K. The Core i5 2500K has the same TDP, as a Core i5 2500 despite the latter does not have an overclocking feature.

Since Turbo Boost is based on the processors TDP, you're not going to get huge boost (0.8-1GHz) than If you're really serious moving from your current Core i5 2500K to a Core i3 2120K then I would consider that a downgrade rather than an upgrade. Keep your Core i5, it will serve you well.

I can't really explain much about it but FlexFP will give you more flexibility when it comes to programs that want to use it compared to the conventional method like SB has 256bit dedicated per core. With the FlexFP you can have 2 128bit and 2 cores in a module can share those 2 to make a 256bit.

FlexFP will not be the determining factor that will put Bulldozer ahead of SB. There are other parts of the CPU that is more important and tweaking those will give a much higher performance than just having FlexFP. I'm a little bit skeptical about FMA4 because SB still uses FMA3 and the industry only moves to FMA4 when Intel pays tells them.

AMD Llano mobile based devices launching by June 14th

sos

AMD FX Processor
Official page for AMD FX processors. I'm 50/50 with AMD Bulldozer as of now but AMD promo videos are always a 100. Their videos are always full of epic lulz

This post has been edited by dma0991: Jun 8 2011, 11:10 PM

^Not too sure if there will be any results posted by him. Posting any performance figures related to a unreleased product would lead to a breach of NDA with AMD. Of course the consequences would be legal action against him and he may not get any ES chips in the future. Then again it would be nice to see some numbers for the final stepping.

IIANM the rules for a NDA is that the ES chips belongs to the company that gives to the person for whatever purpose the company specifies them to do. It can not be sold to another person as it is a property of the company itself even if he did buy it legally. That means the person that sold the ES to him has already gone against the NDA. Most probably AMD would confiscate the ES chip without compensation to the buyer.

Usually ES chips would run at a much lower base clock than a retail version as they try to work out the bugs. Do post some numbers if he really decided to post some information, which I think is unlikely.

A 4 or 6 core BD together with a motherboard for 1k is definitely possible. Since you're gaming, an 8 core is unnecessary since you can better spend on a 4 core and a decent motherboard with good VRM and overclock the 4 cores.

AMD is not aiming to be in the same price bracket as the LGA1366 or LGA2011 so 1k is definitely possible.

AMD Confirms FX Pricing: Mid-$300 for an 8-Core Processor

The price of a FX-8130P is confirmed to be at $320 or about +- RM1000. Comes in a nice packaging as well.

After looking at so many 'leaked' benchmarks I would usually take it with a grain of salt, especially with Cinebench. I lost count already how many I saw a Photoshopped version.

The RAM issue with the APU is already known for quite some time. It has more to do with the bandwidth that the RAM has rather than speed. For a discrete GPU you have a very big bandwidth compared to what you can find on normal RAM. Only solution is to have a dedicated memory for the GPU on the motherboard which was not implemented maybe due to cost. Another solution is to increase the dual channel memory to triple/quad channel memory but have to wait till AMD decides to actually implement.

Putting the RAM issue aside, the number of pins on the socket could have been increased as well. Now the CPU and GPU have to share the 905 pins that Socket FM1 have which might not be a good thing.

That may change when AMD introduces the Komodo to replace the Zambezi. The Komodo is supposed to be an APU so that might be the point that the Socket AM3+ is no longer useful. If according to the timeline Komodo should come out next year and most likely a new Socket will be introduced as well. After they have transitioned to the newer socket then backwards compatibility would continue from there and not backwards compatible with AM3+.

Of course this is all speculation and might have some truth but if AMD really wants to change then they might as well go all the way by bringing down their Socket C32 or Socket G34 for Komodo which has a lot of contacts points and LGA.

It is sort of a big issue with BD but throughout 2005 there has been a lot of changes in design. What we are looking at now is not what it was back in 2005.

BD is made not made exclusively for enthusiasts. It should focus more on virtualization, cloud, database, web and HPC where the softwares used are not restricted and are able to use the full potential of multicore processing.



Comparison between Llano & Sandy Bridge in a mobile platform. They should have done a physical test with the laptops turned on and tested. Simulated version with the edited background looks so weird.

Even if you do not agree with AMD going with more cores, that doesn't stop the progress method that most semiconductor companies are going with. True the current SB is 4 core and most applications are still within 4 cores/threads but does that mean that the upcoming Haswell with 8 cores that would come after Ivy Bridge in 2013 is useless?

Since Intel is also going with more cores like AMD does that mean Intel is going to a path of failure as well? Of course there will be more and more programs developed to take the advantage of multi core but games are slow to catch on. Doesn't mean that you are very satisfied having 2 cores would mean that the majority of users will. For games I think that most PC gamers should be worried more about games being ported from console to PC rather than the other way around. Due to the fact that games are ported from console, having the most uber powerful CPU or GPU is not going to make your game have lifelike graphics and physics.

Despite Intel having a higher per core performance, for applications like VM having more cores is an advantage. You can cram more cores in a server tower means yo can run more virtual machines. When you are depending on thin clients to run your business, having a powerful core is not going to make a difference when most of the office work that involves word processing are not very CPU demanding. So in practice having 8 cores from AMD allows me to run 4 VM while only 2 VM with an Intel if I allocate 2 cores per system. That is already a 50% increase in the number of people that could use. Powerful single core performance has its place in the server market where more cores are not necessary but virtualization, cloud, database, web and HPC which is only a part of the server market can benefit from more cores.

AMD is aiming at the server space with BD because there is a lot of money to be made in the server space and also they want to regain their dropping market share in the server space. Consider that Intel can sell a Core i7 2600K with a die size of 216mm2 versus a Intel Xeon E7-8870 server CPU with a die space of 513mm2. That is 2.4x the die space but if you compare the price of Core i7 2600K which is $320 and Westmere-EX which is $4616, that is a 14.4x increase in profit when you consider the small 2.4x increase in die space.

AMD A-Series Llano APU Sabine Notebook Platform Review

Llano mobile review is out early by 1 day despite the fact that the NDA is not lifted yet.

CyberLink Products Optimized for AMD A-Series APU Family

The main selling point of an APU, leveraging transcoding jobs to the IGP via AMD APP to compensate for the lower performing CPU.

The bias is because Intel huge compared to AMD, therefore for most review sites to stay alive they have to support the winner which at this point in time is Intel. If AMD controls 80% of the market, I am willing to bet that Anand will support AMD. The article is good, just close one eye to the bias. It is unethical but when you have a company like Intel controlling a huge portion of the market, you can expect this kind of skew.

SB is a good product nonetheless but does not mean that it cannot be hyped out of proportion. As an example, the use of some synthetic benchmark should be changed as they barely represent real world use.

Intel doesn't win just by making good CPUs you know. Even if Intel makes SB 10x faster than their competition but nobody knows it then there is no point. That is where review sites comes in and I have to say that I have yet to read a site that is 100% agnostic with brands that they review. Certainly there will be some review sites that bias with Intel and some that bias with AMD. There are more than one way to win a battle. NYAG

Obviously you did not read the link that I posted. Indeed that Intel makes good CPUs but does not mean that they cannot buy off some companies to use their products only, it happened before it can happen again. These days most PC users have more CPU performance than they could utilize,

I think the majority of mainstream users will hardly tap 50% of the potential of what SB has to offer. I have seen as well some who buys a Core i7 2600 but then most of their tasks does not even require HT.


To say which is a faster CPU and which is a cheaper solution is inaccurate. You can see the percentage of transistors that are being allocated for each portion and AMD has 1/3 of each while Intel has the largest portion of it to its CPU. Simply put you cannot gain something without losing performance somewhere else. That is why you can see that AMD has a lesser CPU and a better GPU versus Intel that has a better CPU and a lesser GPU.

Intel opted for better CPU performance therefore they sacrifice their IGP performance and AMD decided to sacrifice their CPU performance for more GPU performance. No doubt that Stars cores are not good but with a transistor budget like Intel, they can make a CPU that is slightly better than what you are currently looking at Llano. Intel themselves cannot perform a miracle by making a CPU and GPU portion that is both better than AMD under the same 32nm node.

@AlamakLor



Regarding that you want the current 990X because of AMD's policy to not change sockets often, there might be a possibility that AMD will not be using AM3+ next year because Komodo is an APU and most likely according to what I said earlier AMD will introduce a new socket for Komodo (Zambezi replacement). I do think that the socket change is necessary though as they don't have enough contact points for a 10 core CPU + IGP bandwidth.

After they have transitioned to the new socket, AMD's usual backwards compatibility will continue from the new socket instead of AM3+ I suppose. However take it with a grain of salt as this is a rumor only for now.

More info on AMD Turbo Core plus a healthy dose of salt.


Not quite when you consider that Intel and Nvidia made a cross licensing agreement to access each others IP. Not too sure why Nvidia agreed to do this because 1.5 billion a year is not a lot of money and if it is a success it might end up eating Nvidia's GPU sales as well if the revision of Intel's graphics based on Nvidia's CUDA cores have substantially better performance than what they are implementing now.

Then again like I said earlier, you cannot gain something without losing performance somewhere else and if Intel wants a more powerful GPU on their die, they might have to sacrifice a little bit of their CPU performance.

This post has been edited by dma0991: Jun 15 2011, 06:13 PM

Lol it depends. If AMD manages to get software developers to increase the number of OpenCL applications then Fusion will be a success.




This post has been edited by dma0991: Jun 15 2011, 06:38 PM