vmN :
Tom Tancredi :
Well the only thing I would first correct on was the 'Cores' argument. More physical cores is good but when you go above 3 or 4, the problem is (and reason why AMD falls behind Intel with HALF as many 'cores') how the Cores get in each other way.
No. It is because what AMD and Intel define as a core is different, and what is inside them can vary.
Haswell for an example have 4 ALU per core, meaning piledriver have 4 ALU per module(2 per core).
Meaning in single-threaded application Haswell essestially have 2 extra ALUs to work.
Cores getting in the way of eachother happends to all architectures, only certain task can be stretch to an infinity amount of cores without "getting" in the way of each other. (rendering as an example)
Uhm I think you mistyped that, because according to what your saying "Haswell for an example have 4 ALU per core" would then result in a i3 which has 2 "cores" would have 8 ALUs (same as a i7????). In actuality, Intel broke the processing into 'micro-operations' which is where the Hyperthreading helps on. Crash course on all three generations here:
http://www.anandtech.com/show/6355/intels-haswell-architecture/8
As you will note there are different ALUs we are discussing (Naehelm for example has '5' ALUs technically) but as well Store Data, Vectors, and more to the solution. Mainly though the bigger difference in how the "micro-ops" occur by breaking down the processing into six ports and how to ensure these don't conflict (as I am mentioning below) is where the Threading comes into play. In the case of Intel, they came up with the better solution. IF AMD had come up with a similiar or came up originally with the Hyper threading method and applied it to the current base of physical cores (for example the FX-6xxx line) the performance would be the difference of a Ferrarri vs a Beetle. Unfortunately AMD has not address this issue, and shows in multiple performance test that no matter how many physical cores they 'added' to keep pace with Intel's iCore line of virtual and physical cores (see below) the performance levels didn't match even half the expectations unfortunately (8 core AMD should trounce a 'virtual' 8 core i7 for example).
vmN :
Tom Tancredi :
Think of it this way, if you lifted a set of bricks from one place to another your a Single Core PC. Adding another worker (Dual Core) you can do twice the work and speed up the process. If you add more workers (TriCore, QuadCore), and step back you will tend to notice some workers (cores) are waiting for the other workers to get 'out of the way' so they can 'grab a brick' (grab data from memory) or 'drop a brick' (pass the processed data to the component -Sound, GPU, etc.- to 'do' what the data needs to do) and turn around to go back and get another. This is called THREADING, or how the Cores request data, then pass data; there is a quick point where you will see one core or more doing more work then the other cores, and basically get 'in the way'.
Cores been memoryhungry is certainly one issue with piledriver, but not the greatest. Threading is essentially something completely different than what you descriped. (unless ofc I missunderstood something)
Uhmm no actually it is exactly as I am describing, as you can check yourself (not based on my education, certificated or been working on computers since 1984)
https://www.google.com/search?q=What+is+Hyper+Threading&rlz=1C1VSNC_en&oq=What+is+Hyper+Threading&aqs=chrome..69i57j0l5.5352j0j1&sourceid=chrome&es_sm=122&ie=UTF-8
Some best quotes:
"For each processor core that is physically present, the operating system addresses two virtual or logical cores, and shares the workload between them when possible. The main function of hyper-threading is to decrease the number of dependent instructions on the pipeline."
-
http://en.wikipedia.org/wiki/Hyper-threading
"A technology developed by Intel that enables multithreaded software applications to execute threads in parallel on a single multi-core processor instead of processing threads in a linear fashion. Older systems took advantage of dual-processing threading in software by splitting instructions into multiple streams so that more than one processor could act upon them at once."
-
http://www.webopedia.com/TERM/H/Hyper_Threading.html
"Strictly speaking, Hyper-Threading is best applied to operations and applications where multiple tasks can be intelligently scheduled so there's no idle time on your processor....operations where tasks have to be done in serial, or where one operation has to take place before another can begin, generally don't benefit from Hyper-Threading. Whether you have a single core or a quad core, Hyper-Threading can optimize tasks that can be conducted in parallel so the whole operation is faster"
-
http://lifehacker.com/how-hyper-threading-really-works-and-when-its-actuall-1394216262
So back to my example, pile A moved to pile B, one at a time with multiple cores (like AMD continues to do) / workers to do the task they get in the way of each other (keeping it simple here folks), but with Hyperthreading not only 'faking' a virtual second core, so that 'one core' (worker) picks up two bricks at a time, it appears as though two workers are doing it to the OS (multiple cores) but more importantly when they move to Pile A or to Pile B, they are "intelligently scheduled" so no one gets in the way of the other core. So you improve how often a brick is lifted and put down more rapidly and thus improving performance, even if you had only 'Half' the workers (Intel) doing "twice" the work virtually (each worker picks up 2 bricks instead of 1) because the other workers (AMD) do it one at a time in linear (serial) order.
vmN :
Tom Tancredi :
AMD decided 'more cores good' concept, Intel decided 'better threading good' concept with Hyperthreading. What results is (for normal using PC applications and for gaming - the 95% use of PCs) less core Intel processors outperform AMD systems with TWICE the number of physical cores. So while the FX-8xxx has 8 Cores, a i7 with 4 Cores but 4 HyperThread virtual cores performance TWICE the amount of performance then a FX-8xxx (and why every GPU benchmark you see tests with i7 not FX-8 and why every benchmark test ALWAYS has the i7s at the top).
See my first comment. Also there is nothing as virtual cores. There is threads, and that is it.
To the OS / Application later your wrong. To this layer YES they are virtual cores (see explanations and links above to clarify if necessary) because each core is taking two work loads and as one 'virtual core' is processing one part the other 'core' is also processing the workload, in parallel (NOT SERIAL as your inferring and as AMD does at the moment) so as one workload gets done it doesn't wait on the 'core' it hands it off (hard drive, gpu, sound card, etc.) and go grabs another workload. So yes in all instances it is a virtual core, because you have TWO Processing functions happening inside a single physical core, the Threading comes into the play on the I/O scheduling of the 'work' itself in and out (again see the many example and explanations).
vmN :
Tom Tancredi :
If we drop down to a i5 (as your comparing), then you would think much 'less' performance out of the i5 then the FX-8, BUT what has happened until Haswell (i5-4xxx) is they were Either / OR (either i5 or FX-8) as 'about' the same performance. NOW with the Haswell line, it is i3 VS FX-8 then i5 then i7 repeatedly in the current benchmarks. AMD has decided (2013-2014) NOT to persue competing against Haswell and has completely shifted focus to the 90% other marketspace of the 'common user' needing a "low power, low COST, LOW DEMAND" answer, which they now push the APU line to address. As the FX -8xxx is the only real 'competition' to the low-medium level Intel line, AND any new AMD processor will use a new motherboard / CPU design (see the APUs) which will be INCOMPATIBLE with the FX-xxxx line requiring a Mobo AND CPU replacement, the FX-8xxx does not seem the best solution anymore as it is a 'dead end' when you think about both things.
I3s aren't competing with fx 8xxx, they are basically competing with fx 6xxx. The i5s are currently competing with the fx 8xxx.
I believe AMD already decided back when finishing bulldozer, that it would be their last HEDT (For a certain time ofc, hopefully)
AMD is looking for a new area to explorer, where they stand a bigger chance to compete against intel. AMD is excellent to produce APUs with an exclusiv IGP for a lower pricepoint, and that is were they are aiming their resources at.
Well I could toss at you every BF4 benchmark which shows your wrong, but as well Tom's hardware specifically shows it in other games as well
http://www.bit-tech.net/hardware/2013/11/14/intel-core-i3-4130-haswell-review/5 . Nevermind the review here :
http://cpu.userbenchmark.com/Compare/Intel-Core-i3-4130-vs-AMD-FX-8-Core-Black-8350/1621vs1489 showing the newer i3 Haswell verses the much older Piledriver (which Piledriver was originally design to compete against the established iX-2xxx and directly incoming iX-3xxx cores specs they were aware of) on Single and Dual Core processes (All current consumer applications and games) is beaten by the i3-4xxx, but obviously if serial multicore processing is done the 8 Core FX-8 would do much better then the 2Core+4Threads (NOT as you said 4 to each 1 above) just as much as the i5 would do better (even by Intel's own sales material) than the i3, and of course i7 does better then the i5 for the very same reasons.
The point being, the Haswell chipset has reset the marketspace that the 'lowest' offering by Intel is on par - beats the ONLY highest core AMD has to offer at the moment for Joe/Jane Consumer. Otherwise I agree with the rest of what you said.
vmN :
Tom Tancredi :
That said, I would like to correct your thought process about worrying on the CPU for " video editing (Videostudio Pro X6 atm, probably the adobe suite later on) ". The CPU will not be the MOST important thing you should be considering. While yes a i5 can do video editing, by Intel's own admission a i7 is much better suited for it (more CPU / Thread cores better processing / coding of video). The real issues come from amount of RAM and buying a 'Gaming GPU' than a Workstation GPU. These are MUCH different. 8GB of RAM (common PC) is nothing in Video Editing, and 16GB is the "bare minimum", while commonly 32GB is used with many Amatuer - Pros going with 64GB to as high end as 256GB systems on specialized motherboards to handle those RAM needs.
You are totally overestimating how much memory needs to be in use.
Editing itself isn't the huge memory-hungry beast, rendering it all is.
Rendering is requiring as much memory as possible. (16GB is more than fine for the general consumer).
General consumer yes, but for Video Editing and Rendering, no. I defer again to the numerous posts of OTHER people posting (search Tom's Hardware forum for Workstation as I advised) whom are actually in College or Independent Video Editors, or the several Mom and Pop Video Editing 'companies' that have posted here, and the numbers I used and justification for it were repeatedly used. Again this is about 'time', and yes some did complain on their i7 with 8GB of RAM trying to render 15sec of video at a time (like for example animation sequences) taking HOURS and forget if they wanted to do 30minute long 'shorts' all at once, they didn't want to wait days (which some did try) for it to complete. When discussing just doubling the RAM, this improved, but then the Pros, Independents, etc. weighed in and noted 32GB should be the minimum because using Sony Vegas, CAD, etc. all use both RAM and CPU together for 'crunching' the data before the actual 'rendering' element was handled to the GPU (again noting Quardos and such).
vmN :
Tom Tancredi :
In Video Editing it renders / encodes along with Audio editing / encoding into the outputted file / stream is completely opposite to a Video Game 'Rendering' to a screen for Frames Per Second performance. Video Editing / Rendering takes time, and a Gamer's GPU (the common consumer cards you see) are not designed to do those tasks, but Workstation GPUs like the Quadro (search Tom's Hardware Forum for Workstation and you will see many threads showing you typical and PRO setups) to do the rendering and encoding cuts the time to minutes and a couple hours then for Gaming GPU which can actually take days to render a Video. Conversely while a Workstation GPU is powerful it isn't built for FPS performance, and commonly (see again the forum threads) will performs almost half as well as a equal Gamer's GPU in gaming (say 90FPS in BF4 on a Titan, a Quadro would do 45 for rough example). So you really need to quantify your expectations and decide which is a priority, Video Editing or Gaming, then buy appropriately - is Frames Per Second more important then the HOURS / DAYS to Encode/Render a 20 minute Video?
If your looking for build suggestions numerous Workstation threads are on here with common builds, but the normal price range for Gamer's Desktops is around $1000 or more (typically a i7, 8GB, SSD, 1TB, etc.) but Video Editors START at $2500 upwards (typically either i7s or XEONs, 32GB, 512GB SSD, 4TB, etc.).
The card themselfs are pretty similar, it is more the drivers and which kind of SIMD they have.
You certianly wont be using days to render a 20min movie, unless it is a very special movie in some crazy high format or some shit.
Once again you have overestimated the power of a workstation card. Lower-end workstation cards (Prices similar to high-end "gamer"-GPUs) Generally have worse performance than a high-end gaming GPU.
Yes the 'Normal' is a 1080P video format, so I guess that is 'crazy high format' they all want to do and were doing normally. And NO I am not overestimting the power of a Workstation card which is specifically designed to do this sort and more work, but suck on FPS in gaming and isn't just a 'driver' thing. If that all there was, I am VERY sure the cards would not be selling anymore because several dozen people would have posted 'CAD Drivers" for every GPU card made so they too could 'cheaply' allow everyone to do Professional level Video Editing with the established cards everyone can buy at Best Buy/Fry's Electronics/etc.
Again having supported both Workstations, Business PCs, Consumer PCs, Gamer PCs, and such I have had this experiance first hand of seeing the difference, again you can search the forum and see the posts from others ALSO saying the same thing. Or you can just read the NVidia White Paper http://www.nvidia.com/object/quadro_geforce.html, NVidia's own Legally Binding statements on its website http://www.nvidia.com/object/quadro-desktops-pcs-features.html, the performance difference independent review
http://www.xbitlabs.com/articles/graphics/display/quadrofx-firepro.html or the best walk through I seen by AMD on what is graphics and how they are much different for Professional Workstation Cards as compared to Gamer's GPUs
http://www.amd.com/Documents/49521_Graphics101_Book.pdf
In a nutshell Video Editing, Rendering, CAD, etc. high end video processing is BEST served by Workstation cards but perform lousy in Frames Per Second performance needs as Gamer's cards, Gamer's card push and optimized for FPS demands, but then are deficient powered for Workstation demands in comparison.
Facebook
Twitter
Instagram