The new processors haven't quite hit stores yet, but they are launched now! Performance is a bit less than I was expecting :/ But it isnt as terrible as some people are making it out to be. The processor is a bit better than current generation AMD processors, and the top end processor is about equal to Intel's 2500k. Although it trails behind the 2600k (which most thought it would). Pricing on the top end model is $245, and the lower end 8 thread processor is a little over $200. So they are some pretty good CPUs if your are on a budget. Here is a review on the new processors, I recommend reading through the overclocking page as well as the pages for the game benchmarks (since gaming is what this community is about): http://www.legitreviews.com/article/1741/19/ And a lengthy but very good review from Tom's Hardware: http://www.tomshardware.com/reviews/fx-8150-zambezi-bulldozer-990fx,3043.html Why is performance a bit low? Well from reading through a lot of info about these and comparing results from all over the place I can see 2 things wrong with the processor and a possible 3rd. First is that something seems to have gone wrong in the design of the cache prefetching algorithms. Performance is heavily dependent on memory, which means the cache is missing ALOT. That is not supposed to happen and is a major factor in all modern processor's performance. Most likely this has to do with how tightly coupled the cores are to each other in each module. The only way to fix this is to redesign part of the CPU, and that probably wont happen anytime soon. Piledriver is already being worked on (the 2nd gen Bulldozer) and we may see a fix in this, but IMO it is already too far along and we will probably have to wait to see a fix for the cache problems until the 3rd generation Bulldozer processors in 2013 The second thing I see as wrong is yet again the memory controller is quite slow at stock speeds. This is a big problem because of issue #1. If the memory and the controller were faster, it would help cover up some of the performance problems of the cache misses. And third, it seems a patch was released for Windows to help with this, but Windows 7 and earlier is not designed at all for this type of processor and the scheduling in Windows does not take proper advantage of the CPU configuration for maximizing speed. The patch seems to have fixed this but some people have said the patch bring about other possible performance issues (as well as security). Ill have to look into the patch more to see what exactly is wrong with it, but for now the quick temporary patch seems to be doing an ok job of things. The main issue with power draw is Global Foundries 32nm process. It is currently having a lot of problems and yields are both low and needing a lot of power to function. Later models of this processor will hopefully improve greatly in power consumption as Global Foundries fixes its process. This will in turn also bring higher average overclocks as well. From the reviews, people who reviewed with the latest patches and did some good overclocking saw pretty good gains in speed. Once I get my processor and do some tests Ill let you know just how well having fast memory compares to average speed memory, and how much performance is gained by overclocking the memory controller as far as I can. But my recommendations for now? If you can afford it go with an Intel 2600k. If you cant afford that, then either an Intel 2500k or an AMD FX-8150 will be your next best thing as they are about equal.
I want to see some benchmarks for the x4 and x6 models since those are in my price range. :zzz: Well should be some stuff out by end of November maybe even get some minor price war with 2500k. If Intel drops it by $50 I would take a long hard look at it maybe drop down to a 500g hard drive instead of 1t drive to fit it in my budget. Guess just have to wait and see. :drool::bigeyes:
So the consensus around the interwebs seems to be that these are server parts they tried to shoehorn in to the desktop market to defray some of the RnD cost. I think they are going to have to cut the price a bit though if they want to sell a lot of them. Especially with Trinity coming next year which seems like it's far more focused on the consumer market.
Now some people are saying that most of the reviews where done with engineering samples on Asus MBs. Don't know how much difference that would make. But I do find it a little off that mostly all benchmarks are done with Nvidia cards. Along with some 4x and 6x tests I would like to see some FPS tests with a few AMD cards as well and maybe a few different MBs from different companies.
Nope, people are trying to make excuses. I have watched extensively as all manner of tests are put to these things: Reviews were done with retail samples, earlier engineering samples were actually much slower due to disabled prefetchers. The prefetcher is the problem with these CPUs, AMD could never get it working right so it completely destroys its cache and thus performance is terrible due to many trips to RAM and back. What you are probably hearing that people are misunderstanding is that AMD did not release final AGESA code for support of bulldozer to run properly until a couple weeks before launch. Motherboard makers (only ASRock) were not able to get a stable working bios out on the right code, they were using older (not final) AGESA code for the bios. You can gain 15-20% performance over an FX-4100 in heavily threaded situations by disabling the second core in each module as compared to two modules active with both cores (4 modules, single core vs 2 modules dual core). More performance is gained in lightly threaded situations by having just 2 modules in dual core mode The Windows 7 scheduler doesnt know how to use Bulldozer and a patch is needed to the system to increase performance another 5-10%. Windows 8 scheduler does a better job. Global Foundries, the place these CPUs are made, is having a ton of problems with their 32nm SOI process node, which led to very bad yields on these processors and why they take such huge amounts of power to run when overclocked. Also why the launch speed is so low. AMD had planned to launch these processors at around 4GHz stock with turbos to 5GHz. The plan has always been to decrease IPC (instructions per clock) over the previous generation by 10-20%. At the same time the architecture was designed for very high speeds, so with a 10% IPC decrease we were supposed to see a 40% core speed increase. Thus giving total hgiher performance. But Global Foundries failed big time and the core speeds could not be delivered, thus leaving us with less IPC and only equal speeds. Testing done with AMD graphics cards come out exactly the same as the Nvidia cards And odly enough, Gigabyte motherboards get worse results than ASUS, and ASRock gets much better results than ASUS. So get an ASROCK AM3+ board if you want to run one of these processors. MASSIVE performance increases are seen from both overclocking the northbridge to 2800+ speeds and using DDR-1866+ RAM speeds. and very large speed increases are seen once you have the processors core speed at or above 5.5GHz We wont see any better performance till AMD can fix the prefetchers killing the other core in the modules and GF can fix its 32nm SOI process node. Testing is still going on to see how we can increase performance, but pretty much all the basics are now known in how these processors work in different situations.
here is what happens with a Bulldozer if you do a bit of overclocking and put good memory in it. 4 cores vs 4 cores, BD vs Sandy Bridge. Same core speeds. This is just dirt 2 though, doesnt mean all testing would be the same but it is a good example of things after you get good performing all around system instead of testing with 1333 memory:
Dude doing some real world testing seems to show that the difference in performance in minimal in terms of gaming. whether that's good or bad is up to you I still say they need to drop the prices if they want this to sell well.
Those are the results I have seen from tests elsewhere too in gaming at normal gaming settings. As you overclock the CPU you see a correlation to increased minimum FPS, which is the most important part of gaming. This happens with both the 2500k, 2600k, and bulldozer. When overclocked, the BD actually gets a little bit better minimum FPS in most games than Sandy Bridge. However any money you could save on the whole platform would be offset by higher power consumption.
All I need to know is do I buy or wait longer?? The only reason why I was waiting for these new CPUs is because Guild War 2 isn't here yet? Well they CPUs are here and I'm not reading good things so should I wait?
Or just get a 2500k and call it a day. Revisions for bulldozer are supposed to be out early next year if you really want to wait as well. In that case you could just grab a phenom II to hold you over they work just fine for now.
lol AMDs Battlefield 3 Can You Run It? page says the fx8150 can't play it. (note these things are mostly bullshit marketing gimmicks.)
The FX-8150 can easily play the game. It can play anything the Sandy Bridge processors can*, just at a bit lower minimum FPS if you are running at stock speeds. To say something *cant* in this day and age is pretty dumb. Those sites tools probably arent updated for the latest hardware because it says it recommends a 2GHz dual core and a 4.6GHz quad is below the recommended. Also Nvidia is using their marketing to say the recommended graphics settings are much higher than what the game recommended actually is. If you look at their chart it says "meets medium quality recommendations" just so they can push their more expensive cards. Those "medium quality settings" are the actual recommended despite what Nvidia wants to claim. *apparently not. My statement is no longer true because some very weird bugs were discovered that cause a BSOD on certain games all the time no matter what. No one knows why this happens or how to fix it yet but the problem is being looked into. Games include Dues Ex: HR, Shogun 2, and more.
Heard some very bad news today. AMD is cutting 10% of their staff. Obviously because they need to save money and get people out who fucked up really bad with Bulldozer. But the very odd and bad news is that many of the people being cut are actually from the GPU side of things, not the CPU. much of the driver team was given the boot (WTF???) and much of the marketting team on the GPU side. Additionally, they let go the two most important people for the community and overclocking area of AMD processors. All the people who are part of the community and really get word out there and push the limits of things. This wont lead to any good for AMD Its very sad to see Simon and Brian no longer with AMD It sounds like the new CEO is taking a step away from enthusiast graphics cards and probably PC processors. If thats the case we wont have any competition to Nvidia or Intel.
That's perplexing, because they've seemed to have had the upper hand over nVidia for some time now on the GPU front.