LinkBack URL
About LinkBacksI'd prefer you to stay so that I could school you some more. I find that most flaws can be cured with education, granted you'll be a harder task than most.Originally Posted by Bambooz
I'd prefer you to stay so that I could school you some more. I find that most flaws can be cured with education, granted you'll be a harder task than most.
It's pretty much the case. There is no clock headroom left and IPC gains can't go on. New instructions etc are mostly for benchmarking purposes it seems. Look at how often Skylake basically just draws with Haswell in some of them.
Also GPU is where it's going. Basically speaking it's impossible for AMD not to match Intel with Zen, at least in gaming terms. If they can't equal that they might as well give up. With Dx12 and any kind of decent single threaded performance the fps charts with each CPU basically equal except in extreme corner cases. Later on when devs start using the API properly, more cores will matter. Gaming is what is selling desktop CPUs, nothing else.
The end of dual cores is finally coming, with quads set to be the new entry point. Not before time, and not because Intel wanted to push the industry forward, that's for sure.
In some cases, performance scaling with cores is fairly good and in those cases it often makes sense to use a greater number of smaller cores; the smaller cores being generally more efficient, cheaper, etc.
For instance you can get almost four ARM A7 cores in the die size/power budget of a single A15 core. If you're after single-threaded performance it won't do you much good but with highly multi-threaded apps you're laughing. Extracting single-threaded performance from a core and keeping execution resources fed takes a lot of work with diminishing returns as you continue to push, so for embarrassingly parallel applications like graphics rendering we have GPUs which use a huge number of smaller cores (in very simple terms anyway, of course it's more complex than that in reality).
For a given die size/transistor count, the theoretical throughput of a GPU can easily be many times greater than that of a CPU, but of course not everything is as parallelisable as graphics. Amdahl's law is based on this and the relationship of potential performance gains vs how much of your code is inherently serial i.e. if a part of your code can't be spread across cores, at some point it will end up being a bottleneck for further performance increases. We saw something along those lines with DX11 and earlier - they've had some limited multithreading support but as long as some of the code is stuck waiting on a single thread you'll be stuck in terms of how much additional threads help.
Going back a few years Intel were talking about the concept of essentially having larger and smaller cores suited to different tasks: http://www.anandtech.com/show/2580/14
Of course we have something not all that different to that now with 'APUs' i.e. CPU and GPU. A remaining problem is getting software to play nice with such architectures. Standards would help and while AMD is pushing HSA with a few other companies, software support is still catching up and of course it doesn't really help you much having a potentially very powerful architecture if developers either won't or can't effectively make use of it.
Looking at it from a slightly different angle, mobile processors are a good example for having certain common functions performed in dedicated hardware instead of running on the CPU. For instance video encode/decode, ISP, audio DSP, cryptography and so on are well-suited to hardware implementations and can be many times faster and more efficient than software would be.
Cinebench- Almost breaking through the 1,000-mark barrier ....
In what way is it a barrier? It's in your mind.....
Well the issue is in my view that we are stuck with developers often refuse to adopt new tech. Look at how long operating systems devs refused to give us a 64 bit os.
When AMD released the 64 bit cpu which we are using every day, it was hard to find programs which acutally made use of the 64 bit power. True it does not work on every thing available and actually microsoft was dead right about running office as 64bit is an absolute waste.
But those like me who actually had programs running in 64 bit saw huge performance gains.
This is also the issue for 8 core or higher cpu's no the 6700 is not an 8 core its a 4 core with HT
If intel would release a real 8 core you all start crying that it burns so much energy, because thats the main reason why the AMD FX is actually using much more energy.
True the AMD suffers from its shared cache design, because it has been proven that this hold back the cpu. Anyway as long as there is no need to scale up for normal programs we probably not going to see any improvement at all. Saying that its a long term project to develop is true. But we are at a point that intel does not need to upscale as well. They have no competition so besides the quest to get the cpu's use less power for the same performance intel is not going to release anything they have in the pipeline in coming years other than this kinda silly updates.
So either the need has to change in the programmers world where more cores is going to give a performance increase or AMD must find a way to wake the giant again to move its developing muscles. But i am certain they do not need to worry, hell if it goes on like this their only competition is going bancrupt, because they are in red numbers again. If i read all the comments many of you would love to see AMD disappear, but i bet if it happens you will be stuck for a lot longer with no real development. Then all of you can cry why AMd was so stupid to go bancrupt.
Intel Xeon E5-2630v5 would like a word. 8C/16T. 85W TDP.
This seems really expensive for a Mainstream socket cpu, its almost as expensive as a i7 5820K, for maybe €50 more you can get a 5820K and a X99 Mobo that has way more PCIe lanes which is the only new thing Z170 brought.
Have been itching to upgrade from my current 3770K/Maximus V Formula based build. I had figured I'll go for a new build due to the then new LGA1150 platform. I was initially excited at the arrival of 6700K, but am not too sure now. I am still keen on a new build on the latest 1151 platform, but am now left wondering if I should hold out a little while longer and in the meantime settle for a 980 GPU upgrade on my current 3770K rig instead.
Why ? You're looking at about a 5-10% increase in performance for CPU based tasks. I literally cannot think of a single sane reason to upgrade from 3770K. It'd cost you £400 to get such a slither of performance gain. If you're loaded and like having the newest parts then sure go ahead I'd do the same if I had the money tbh, but then I'd go for X99. The upgrade you're talking about is just...I can't wrap my head around it.
Put the money into a GPU, a faster SSD, more storage, a better or second screen, anything except this upgrade. You'll feel the benefits much more than you would otherwise
I totally agree with your reasoning. But I just want the new beasty as it has arrived at an opportune moment for me plus the future-proof factor. In addition to my lame excuse, the truth though is I still need to build a new rig to replace a now struggling maxed-out Dell Inspirion 531 to stop my four geeky kids from queuing and fighting for my current 3770K rig. I considered getting a console at considerably less cost but they cannot understand why I won't let them have my 3770K rig while I keep the suggested console....! Lol!
What future-proof factor? Speed is the future-proof factor - chips have been x86 compatible for years and will continue to be soIf a new chip isn't any faster then it isn't any more future-proof
Well, future-proof in the sense that my maxed out Dell Inspirion 531 has been struggling for a few years now, while my 3770K rig still very capable and has decent mileage left in it yet.What future-proof factor? Speed is the future-proof factor - chips have been x86 compatible for years and will continue to be so
I cannot argue with you there!!If a new chip isn't any faster then it isn't any more future-proof
Ah fair enough. TBH I'd suggest building the kids an AMD equipped system and save yourself several hundred quid. A quad core AMD will save you about 200 quid and the most you'd lose between that and a high end Intel, assuming a discreet GPU is in place, would be about 10fps. Also going to assume as it's kids you have it hooked up to a TV ? So as long as it's not 4K the most they'll need to push is 1080p, meaning an AMD quad core and a GTX960 would give them 60FPS no matter what. Just a suggestion, if you wanna spoil em (and give you an excuse to give yourself an even higher end rig) go ahead ha.
Daheelah (16-08-2015)
Re: Intel Core i7-6700K (14nm Skylake)
I have considered that idea (as the old Dell is AMD based) though not necessarily the same specs. Alas, I concur with your reasoning and your suggested spec is very sound.
And just when common sense was about to prevail, I saw your closing statement.
Yeah my man, now you're talking. You sure get my drift!! You hit the bull smack in the eye with the statement in the brackets
We are all fixin' to get spoilt..! Lol!
What sort of CPU is in that system? A lot of older systems could see a significant uplift in performance by installing a basic SSD as HDD clutter and hundreds of Windows updates tend to drag down performance a fair bit.I have considered that idea (as the old Dell is AMD based) though not necessarily the same specs. Alas, I concur with your reasoning and your suggested spec is very sound.
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
Reply With Quote