Surprised no one noticed AMDs test-bench:

[cesrai]

I tried...

Sue them I'm sure you can do that, what about the retailer did you talk to them ?

[Darkman]

Sue them I'm sure you can do that, what about the retailer did you talk to them ?

Amazon wouldn't take the board... Probably because it was out of its return policy. I think we're getting too off topic and I'm going to chill for a bit.

[cesrai]

Amazon wouldn't take the board... Probably because it was out of its return policy. I think we're getting too off topic and I'm going to chill for a bit.

Sue them sue MSI.

[Opcode]

Umm, they better do it soon or face bankruptcy, like 3dfx did years ago.

AMD will never go bankrupt. Principle investors will keep bailing AMD out every time they are about to. Without AMD both Nvidia and Intel will become a monopoly all their own. Which is what will keep AMD afloat for as long as either company exists.

 

Well, Intel did design them in the first place to be just strong enough for business computer needs (drive documents in 4K at 60Hz lol) and compute power in their vie to stay on top of IBM in the server chip race. Only recently started pursuing gaming.

Intel still has no interest in gaming (horrible architecture for such tasks). They're trying to play catch-up in the compute department with Broadwell.

[patrickjp93]

AMD will never go bankrupt. Principle investors will keep bailing AMD out every time they are about to. Without AMD both Nvidia and Intel will become a monopoly all their own. Which is what will keep AMD afloat for as long as either company exists.

 

Intel still has no interest in gaming. They're trying to play catch-up in the compute department with Broadwell.

They're fully caught up and surpassed AMD. All the OpenCL benchmarks put Iris Pro 5200 way ahead of Kaveri.

[CoolaxGaming]

Well I dont know but Lisa Su said that ZEN should come in 2016.

Zen will "Reveive" FX with better processes and stuff.

[Enderman]

well, for my use (not even thinking abot budget) id mch rather a 9590 over a i7 any day. 

i find that i'd probably end up rendering recording streaming and playing all at once. i'd like a few moe cores to handle all that. 

 

''oh, but the intel has better per core preformance, eliminating the need for laods of cores''

i dont give a rats ass. per core doesnt mean shit when you have more of them, multi tasking better. running 4 2 thread programs. i'd end up doing that alot. 

and in game, 9590 would only lose a few frames (99% of the time under 3) compared to a similarly setup comp with a 4770k. 

 

anyway, i apologize if that came off as pissy... or rude... these are topics that usually get me rather riled up as it doesnt matter, cause there is no perfect product, everyone has different needs, because, believe it or not some people have tight schedules and arent able to just devote different time slots for work and play... and some people have different styles of each of those. 

 

who cares if they used the other brands CPU. its more honest to show thier buyings what they can really push, instead of cornering them into only using one brand. 

i'd much rather buy from the company who admits to the other guys beign better at some stuff then the company who only uses their own products. 

 

im sure AMD has some magic planned out for APUs. and im sure intel will eventually jump along once they see how it goes. 

An i7 is better than a 9590 at multithreaded tasks...

AMD's eight "half cores"  are still worse than 8 threads.

 

[cesrai]

An i7 is better than a 9590 at multithreaded tasks...

AMD's eight "half cores" are still worse than 8 threads.

Let not say that shall we AMD has 8 cores Intel has 8 cores, different architectures, let's not sat half cores because it will be a war, also the is expected you are comprising an extreme edition with the 9590.

[Opcode]

They're fully caught up and surpassed AMD. All the OpenCL benchmarks put Iris Pro 5200 way ahead of Kaveri.

Iris Pro 5200 is far behind Kaveri if you check both OpenCL and HSA benchmarks (Intel is not even partially compliant) along with their architecture simply not being as good as AMD's at compute performance in general. Intel still has a few years of playing catch-up before they will ever surpass AMD's offerings when it comes to iGPU's. I personally have my doubts that Intel will ever join the HSA foundation let alone even adopt it. Which sooner than later will put them on the bench regardless if they eventually have a much more powerful iGPU. HSA is about to catch on with the launch of the very first fully HSA compliant microprocessor early next year. The HSA stack was already pushed to GIT for Linux. With a superior and simpler programming model HSA is about to take over in software. By the time Intel pulls their head out of their rear end HSA will be standard in every software that can leverage its performance. Intel's gotta catch up in compute performance, gaming performance, and software standards. OpenCL will remain popular and Intel's iGPU will still be able to leverage it. All tho with that circumstance people would rather leverage a discrete GPU which will be all around superior. There is no denying that Intel is the underdog right now in the APU department which is unlikely to change in the next few years. I personally believe 2016 will be the year of the head to head APU battle between the two. AMD's giving Intel a lot of time to play catch up as well. Especially considering we are likely not seeing desktop Carrizo this time around. Unless they are keeping their doors sealed tight for something.

[Enderman]

Let not say that shall we AMD has 8 cores Intel has 8 cores, different architectures, let's not sat half cores because it will be a war, also the is expected you are comprising an extreme edition with the 9590.

Did you not look at the picture?

How can you not see the picture?

It has the 4790k, 3960x, 4960x, 4690k, etc... there is more than just the 5960x.

[BurgerBum]

Okay lets count look pic at the pic

sig.PNG 

 

 

first line starts with *logitech and ends with AEL*, second line starts with *NEEDS and ends with PC*, third line start with #900 and ends with ble, fourth line see we reached the fourth line not going to continue I hope you get the point it's more than three lines.

zoom the fuck out then. im on a 1600x900p monitor at 125% zoom. its only three lines heere.

[cesrai]

Did you not look at the picture?

How can you not see the picture?

It has the 4790k, 3960x, 4960x, 4690k, etc... there is more than just the 5960x.

Not my point it's known that the i7 will win in a lot of tasks, they aren't half cores is my main point, phrase AMD *insert AMD fanboy song here .

[Opcode]

Okay lets count look pic at the pic

sig.PNG 

 

 

first line starts with *logitech and ends with AEL*, second line starts with *NEEDS and ends with PC*, third line start with #900 and ends with ble, fourth line see we reached the fourth line not going to continue I hope you get the point it's more than three lines.

It's only three lines, HTML output can back that up as well.

 

[cesrai]

zoom the fuck out then. im on a 1600x900p monitor at 125% zoom. its only three lines heere.

Did nothing happened show me a pic, there is no way it's three lines, I'm sorry for calling a lair and please change the color my eye hurts.

[cesrai]

It's only three lines, HTML output can back that up as well.

Why does it appear more than three lines for me ? I used different resolutions no hope.

[Opcode]

Why does it appear more than three lines for me ? I used different resolutions no hope.

If your browser window is shrunken down too small text automatically extends itself to another line. Prime example is my signature text is just one line, shrink the browser window down small enough it will turn into two lines. You must be running a very low resolution because with Chrome @ 1920x1080 everything's fine for me.

[patrickjp93]

Iris Pro 5200 is far behind Kaveri if you check both OpenCL and HSA benchmarks (Intel is not even partially compliant) along with their architecture simply not being as good as AMD's at compute performance in general. Intel still has a few years of playing catch-up before they will ever surpass AMD's offerings when it comes to iGPU's. I personally have my doubts that Intel will ever join the HSA foundation let alone even adopt it. Which sooner than later will put them on the bench regardless if they eventually have a much more powerful iGPU. HSA is about to catch on with the launch of the very first fully HSA compliant microprocessor early next year. The HSA stack was already pushed to GIT for Linux. With a superior and simpler programming model HSA is about to take over in software. By the time Intel pulls their head out of their rear end HSA will be standard in every software that can leverage its performance. Intel's gotta catch up in compute performance, gaming performance, and software standards. OpenCL will remain popular and Intel's iGPU will still be able to leverage it. All tho with that circumstance people would rather leverage a discrete GPU which will be all around superior. There is no denying that Intel is the underdog right now in the APU department which is unlikely to change in the next few years. I personally believe 2016 will be the year of the head to head APU battle between the two. AMD's giving Intel a lot of time to play catch up as well. Especially considering we are likely not seeing desktop Carrizo this time around. Unless they are keeping their doors sealed tight for something.

Intel doesn't have to adopt it. Intel has market dominance. If they add all the features (unified memory, equal-citizen scheduling, unified CPU-GPU SIMD/MIMD instructions) then they can push their standard into the market. AMD will have forced Intel to innovate, no doubt, but the reality is Intel is so far ahead already that all they have to do is keep up with the open/major standards (OpenCL 2.0 fully compliant, OpenGL 5 fully compliant, DX12 fully compliant, etc.). OpenCL is doing 1/2 of HSA, making unified memory access possible and making heterogeneous designs easier to implement.

 

You entirely underestimate the slowness of the market. Going along with HSA locks you into a very tight ecosystem (until a ton of people adopt it, ancd the HSA foundation is pretty small and doesn't have much software out), and it still takes time to learn that software stack. OpenCL has already been here, and 2.0 makes it a whole lot easier to leverage the iGPU for acceleration.

 

If AMD can catch up on compute Intel can simply double their performance at the drop of a hat. To think Intel doesn't have a whole host of features and improvements waiting in the wings is foolhardy. Any time AMD does something interesting, Intel just releases 1 bird from the cagte and in the meanwhile copies AMD's move for the next generation of chip. In the meantime, 4 more birds were added over the course of the year. Zen is AMD's last chance, and let's be honest Skylake is not going to be some pushover architecture. 50% increase in graphics core count over Broadwell, a new generation of GPU architecture, AVX 3.2 512-bit extensions, a whole slew of video encoders/decoders, wireless charging on the order of 90+% efficiency, greater core scaling, TSX-NI re-implemented for even stronger multithreaded support and transactions... AMD has it tough, whether or not you'd like to believe it.

[Opcode]

Intel doesn't have to adopt it. Intel has market dominance. If they add all the features (unified memory, equal-citizen scheduling, unified CPU-GPU SIMD/MIMD instructions) then they can push their standard into the market. AMD will have forced Intel to innovate, no doubt, but the reality is Intel is so far ahead already that all they have to do is keep up with the open/major standards (OpenCL 2.0 fully compliant, OpenGL 5 fully compliant, DX12 fully compliant, etc.). OpenCL is doing 1/2 of HSA, making unified memory access possible and making heterogeneous designs easier to implement.

 

You entirely underestimate the slowness of the market. Going along with HSA locks you into a very tight ecosystem (until a ton of people adopt it, ancd the HSA foundation is pretty small and doesn't have much software out), and it still takes time to learn that software stack. OpenCL has already been here, and 2.0 makes it a whole lot easier to leverage the iGPU for acceleration.

 

If AMD can catch up on compute Intel can simply double their performance at the drop of a hat. To think Intel doesn't have a whole host of features and improvements waiting in the wings is foolhardy. Any time AMD does something interesting, Intel just releases 1 bird from the cagte and in the meanwhile copies AMD's move for the next generation of chip. In the meantime, 4 more birds were added over the course of the year. Zen is AMD's last chance, and let's be honest Skylake is not going to be some pushover architecture. 50% increase in graphics core count over Broadwell, a new generation of GPU architecture, AVX 3.2 512-bit extensions, a whole slew of video encoders/decoders, wireless charging on the order of 90+% efficiency, greater core scaling, TSX-NI re-implemented for even stronger multithreaded support and transactions... AMD has it tough, whether or not you'd like to believe it.

Intel has no dominance in the market as both companies took a different direction years ago. The problem Intel will face by pushing their own standards is by the time it makes it to the market HSA would indefinitely be showing promising dominance (especially between now and 2016). While OpenCL is helping close the bridge with heterogeneous computing you still need an architecture designed with it in mind which Intel currently doesn't have. They will finally support unified memory with Gen8 of their architecture while they act as if they pioneered it. Gen8 still lacks critical components for safe and efficient heterogeneous computing that AMD has had for nearly a year now.

 

While OpenCL has a wider range of adoption over HSA. It's also widely adopted as a standard as it has existed for many years now where as HSA is not even fully available yet. You cant fully utilize HSA at this point in time due to there not being a fully compliant architecture on the market. With that being said we don't know exactly where and how far HSA will go.

 

Intel is the one who needs to catch up in the world of compute as they don't even support the basics required for heterogeneous computing. You think Intel can double their performance at the drop of a dime? The only way possible would be for them to stack more EU's per layer (what they are doing with Broadwell and then Skylake). Intel is indeed playing the catch up game as they know if they don't do something they will ultimately lose all competition with AMD. Heterogeneous compute has been taking Intel's flagship enthusiast chips to the cleaner for a while now. That all being because Intel fell behind and thought their superior IPC in serial processing is what ultimately matters. This is why I consider AMD to be the market leader as they are far ahead in heterogeneous computing which is where the market is going thanks to AMD (Intel's trailing AMD's innovation). Give me a 5960k and a 7850k and I will personally show you how to triumph the Intel monster over and over again. I personally don't think AMD fears whatever Intel has to offer. Excavator will bring forth AVX2, BMI2 and RDRAND instructions so Zen will more than likely support all the same instructions as Intel's offerings. There is no question that Zen will be something superior in comparison to their current lineup of FX microprocessors. The fact that Jim Keller was told to run wild while in charge of the architecture leads me to believe that we may have another Athlon 64 on our hands. In which case as long as AMD retains their dominance in heterogeneous computing Intel may be in for a world of hurt. Intel is kind of going along with the generics basically copying that of what AMD has already done and calling it "innovation". Meanwhile AMD's doing some very impressive new things as of late. You may get a hint of me being an AMD fanboy. Tho in which case I am not. Tho battling an Intel fanboy you need to sort of sound like a fanboy yourself.

[patrickjp93]

Intel has no dominance in the market as both companies took a different direction years ago. The problem Intel will face by pushing their own standards is by the time it makes it to the market HSA would indefinitely be showing promising dominance (especially between now and 2016). While OpenCL is helping close the bridge with heterogeneous computing you still need an architecture designed with it in mind which Intel currently doesn't have. They will finally support unified memory with Gen8 of their architecture while they act as if they pioneered it. Gen8 still lacks critical components for safe and efficient heterogeneous computing that AMD has had for nearly a year now.

What critical components? Also, Intel's gen 7.5 is completely OpenCL 1.2 compliant, so it's not like it's actually missing anything. The big up an coming feature of 2.0 is unified memory and transactions, and no Intel didn't pioneer it. They just jumped onboard since it's obvious the market is headed that way.

 

 

 

While OpenCL has a wider range of adoption over HSA. It's also widely adopted as a standard as it has existed for many years now where as HSA is not even fully available yet. You cant fully utilize HSA at this point in time due to there not being a fullycompliant architecture on the market. With that being said we don't know exactly where and how far HSA will go.

The HSA stack isn't fully developed yet either. Furthermore, OpenCL is what, 8 years old? That's not a lot of time

 

No, Intel dominates compute with existing software suites, hence why the Xeon Phis are the most purchased accelerators post-2013, and while Intel has yet to put a big iGPU on its E5 and E7 Xeons, that begins with Skylake. The demand hasn't been there yet, but they're staying ahead of the market to meet AMD at the hilltop when that demand arrives. You're the only one here thinking AMD's GPU must have stronger compute because it's better for gaming. That is simply untrue. The big difference between OpenCL 1.2 vanilla and HSA is automatic (by pragma) load balancing and scheduling between the CPU and GPU (currently still managed solely by the CPU in Kaveri, becoming equal citizen in Carrizo). By that token, HSA programming may be more fluid/easier. I haven't looked at the stack and APIs to know. That said, anything on that front can be done on the programmers' side with a little effort.

 

You don't compare heterogeneous compute against Intel's flagship CPUs because they're 2 totally separate paradigms. Intel's 12+core Xeons are meant to handle lots of serial computing and manage loads/workfloaws coming in from dozens of difference sources. Accelerators are meant to compute solutions quickly to parallelizable problems. Intel's CPUs are still the best at scheduling those tasks vs Opterons as well. What Intel is now aiming for is adding density of compute beyond the Xeon Phi cards. Xeon Phi are great for throughput, but iGPU is great for smaller tasks which need low latency, or you can split jobs now across more platforms and work groups. The idea is not for Intel's iGPUs to suddenly be better than any accelerator. It's to augment and compliment already existing big iron architectures to make up for some of their shortcomings.

 

Also, try a basic map-reduce problem (fundamental to big data analysis) on Kaveri and then on Iris Pro 5200. Iris Pro will win, taking only 75% the time on almost any workload size.

 

IPC does matter a great deal, but it has an upper limit of 1. Intel is already so close to that all they can do is release more instructions and speed up their already world-leading FPU on the CPU side. Once the majority of Intel's engineers are focused on the iGPU, Nvidia will be glad Intel has no interest in dedicated video cards.

 

Again, to make the 5960X fight in compute would be horrifically stupid. Don't use a screwdriver to bash a nail in. CPUs manage. Accelerators compute. That is the eternal relationship. Put your 7850k up against a 4980HQ and watch the 7850k go up in smoke.

 

Intel has AVX 3.2 (512-bit) coming as well as the resurrected TSX-NI multithreading support (which had a hardware bug in Haswell and Core M Broadwell which wasn't discovered until very late in the game).

 

I never said AMD didn't innovate. I just said Intel already has all the research done. The only reason they ever move is if AMD looks like it might start catching up.

 

Also, for every new instruction AMD has to implement, it's more money going into Intel's pockets. Those license agreements are on a per-instruction basis.

[Ertman]

Your logic doesn't belong in this hen fest, please see yourself to the door.

 

I say good day sir! ....

 

*walks to door and leaves room*