will coming future games will require 8 core processore?

[cveer]

As ps3 and xbox one are releashing with 8 core processor! so will future games will require 8 core processor as common ?

Is it true intel is releashing 8 core desktop processors, haswell-e?
what will be its expected release date?
i am going to buy a new Pc.
should it will be worth waiting for the new intel 8 core processors for future proof gaming!
Please really help!

 

[JOOK-D]

Okay, I have nothing to do, so I'll leave a comment before I go to bed and might respond tomorrow.

Tek proved nothing. Tek is a "hipster" or whatnot. For a period of time he was the big "ANTI-Intel", and can you guess what CPU is in his system to this day?

You seems to not understand why Intel product are more expensive than their opponents(AMD)s.
It is not because Intel is overpricing their products. Intel in manufacturing their own CPUs, which are extremely costly, which AMD cannot do, because they simply doesn't make money enough. AMD pay another company to do this(Just like nvidia). Intel also have way more marketing than AMD could ever be dreaming of, and are playing a very aggressive game(Actually coming with new architectures, providing neccecary ISA), where AMD have been sitting on piledriver for so long, and a generally stuck with the cluster-cores(modules). AMD have been playing a rather safe game, compared to Intel(In the CPU department).

Piledriver is not better than Haswell, which are also a major reason why they have put their FX line on a stop(also because of their shitty marketing, which they are trying heavyly to fix), and been focusing on lower-end systems. AMD knew they cannot keep fighting a fight they would eventuelly lose, so they targeted a new area, where they would stand a bigger chance.

in reality they cant put out new tech and keep up with demand for the Wii U, PS4, and XB. BF4, Thief and soo Watch Dogs will push gaming heavilly toward AMD and Mantle since they are built around AMD instructions and not Intel. basically Tek proved that if you use a radeon card for real world testing with AMD it keeps up with intel just like they promised. So if ou like Nvidia your out of luck with AMD, but AMD all in means top notch performance all around. by November this year all new AAA titles will be built for AMD 8 core CPUs and GPUs. in any logical manner it all makes sence. I am not saying Intel is bad i am saying Tek only shows that in real world use your not losing performance either way. In another words dont waist money if you not just looking for bragging rights.

No.

 

[airborn824]

Okay, I have nothing to do, so I'll leave a comment before I go to bed and might respond tomorrow.

Tek proved nothing. Tek is a "hipster" or whatnot. For a period of time he was the big "ANTI-Intel", and can you guess what CPU is in his system to this day?

You seems to not understand why Intel product are more expensive than their opponents(AMD)s.
It is not because Intel is overpricing their products. Intel in manufacturing their own CPUs, which are extremely costly, which AMD cannot do, because they simply doesn't make money enough. AMD pay another company to do this(Just like nvidia). Intel also have way more marketing than AMD could ever be dreaming of, and are playing a very aggressive game(Actually coming with new architectures, providing neccecary ISA), where AMD have been sitting on piledriver for so long, and a generally stuck with the cluster-cores(modules). AMD have been playing a rather safe game, compared to Intel(In the CPU department).

Piledriver is not better than Haswell, which are also a major reason why they have put their FX line on a stop(also because of their shitty marketing, which they are trying heavyly to fix), and been focusing on lower-end systems. AMD knew they cannot keep fighting a fight they would eventuelly lose, so they targeted a new area, where they would stand a bigger chance.

in reality they cant put out new tech and keep up with demand for the Wii U, PS4, and XB. BF4, Thief and soo Watch Dogs will push gaming heavilly toward AMD and Mantle since they are built around AMD instructions and not Intel. basically Tek proved that if you use a radeon card for real world testing with AMD it keeps up with intel just like they promised. So if ou like Nvidia your out of luck with AMD, but AMD all in means top notch performance all around. by November this year all new AAA titles will be built for AMD 8 core CPUs and GPUs. in any logical manner it all makes sence. I am not saying Intel is bad i am saying Tek only shows that in real world use your not losing performance either way. In another words dont waist money if you not just looking for bragging rights.

No.

rofl you prove me right

 

[airborn824]

hey dont use Toms bias benchmarks they arent real world or even accurate

 

[JOOK-D]

So airborn824 can be passed off as a troll then.

 

[airborn824]

So airborn824 can be passed off as a troll then.

i hate trolls and i hate fan boys thats all. i believe Tek is right and i believe consoles will help and on blind fans cant see it coming. i will buy whatever is good for what i need at the time.

 

[JOOK-D]

So airborn824 can be passed off as a troll then.

i hate trolls and i hate fan boys thats all. i believe Tek is right and i believe consoles will help and on blind fans cant see it coming. i will buy whatever is good for what i need at the time.

You're the blind fan "AMD Loyal Gamer". If you truly believe that Tek Syndicate is right and every other resource on the internet is wrong, then you're an out-and-out fanboy.

I am not an "Intel bafoon". In fact, check the best answer I provided a long time ago, I also don't appreciate you resurrecting old threads just to make an unsupported argument.

I also don't understand how the benchmark I provided proves your point in any way.

 

[airborn824]

So airborn824 can be passed off as a troll then.

i hate trolls and i hate fan boys thats all. i believe Tek is right and i believe consoles will help and on blind fans cant see it coming. i will buy whatever is good for what i need at the time.

You're the blind fan "AMD Loyal Gamer". If you truly believe that Tek Syndicate is right and every other resource on the internet is wrong, then you're an out-and-out fanboy.

I am not an "Intel bafoon". In fact, check the best answer I provided a long time ago, I also don't appreciate you resurrecting old threads just to make an unsupported argument.

I also don't understand how the benchmark I provided proves your point in any way.

to bring you up to reality its the first and only time i have see realtime benchmarks and not synthetic so its quite nice to see real testing. I am loyal yes. i buy AMD because of Intel Business practices not any other reason. not to long ago thay we investigated for price fixing. i have since gotten over that and am looking for a good upgrade. tek has turned my decision upside down and now i must wait.

 

[palladin9479]

If people don't start calming down it'll be vacation time.

If you truly believe that Tek Syndicate is right and every other resource on the internet is wrong, then you're an out-and-out fanboy.

I actually reproduced his findings. The key had to do with what streaming / recording software your using. Image compression is a very parallel friendly task, it can use anywhere from 1 to infinite (as many as the program supports) cores and split the work evenly across them with near to linear results. The i5's utilized four cores with 3 general purpose ALU's each, the fx8350 had eight cores with 2 ALU's and 2 AGU's, for the vast majority of code it's the ALU's that run the game. Intel has four 256-bit SIMD FPU's while the FX8350 has eight 128-bit FPU's that can combine into four 256-bit FPU's when necessary. The Intel FPU's can process multiple simultaneous instructions if the code is optimized for it. Most games utilize 2~4 cores worth of work, this leaves 50~75% of the fx8350's computing power unutilized while the i5-2500K would lose 0~50%. That unutilized portion is where the streaming software comes in, the AMD chip simply has more excess computing power to spare that's not being used by the game. So unless they restricted the streaming software to a single thread's performance or run it at low priority, it would naturally consume computing power that would otherwise go to playing the game.

This changes a bit with the i7 (the i7 is not more expensive then the i5 to produce, they are the exact same chip, i5's are just i7's with a single flag disabled). The i7 allows what would normally be underutilized CPU resources (the extra ALU or FPU per core) to be used for other things, this would somewhat lesson the impact the streaming software had on the game itself. Unless a program is heavily optimized for a specific uArch there is always times when an ALU or two isn't being used. Haswell changed this further as it introduced a fourth ALU into each core and thus each Haswell core has 2x the integer processing resources as a PD core. You won't really see much difference with the i5 as the exposed four cores limits the amount of traffic that can be simultaneously scheduled onto it at once. With Haswell i7's on the other hand you now have significantly more underutilized CPU resources then you had before, two threads scheduled onto the same core would each have 2 ALU's to work with vs 1.5 on SB/IB and the 4 per module (2 per core) of PD. Haswell i7 essentially does away with any competitive advantage AMD's PD design had in multi-threading but not on the i5.

Please remember Teksyndicate was comparing the 8350 to the i5-2500 and i7-2600. At that time the fx8350 was better at streaming then the i5-2500 about about equal to the i7-2600 due to the above stated reasons. These findings actually mirrored many other places findings but were worded in different ways. Gaming + streaming is easily a heavily multi-threaded workload, something PD was already acknowledged as being great at.

 

[vmN]

@airborn824

Spoiler

in reality they cant put out new tech and keep up with demand for the Wii U, PS4, and XB. BF4, Thief and soo Watch Dogs will push gaming heavilly toward AMD and Mantle since they are built around AMD instructions and not Intel. basically Tek proved that if you use a radeon card for real world testing with AMD it keeps up with intel just like they promised. So if ou like Nvidia your out of luck with AMD, but AMD all in means top notch performance all around. by November this year all new AAA titles will be built for AMD 8 core CPUs and GPUs. in any logical manner it all makes sence. I am not saying Intel is bad i am saying Tek only shows that in real world use your not losing performance either way. In another words dont waist money if you not just looking for bragging rights.

Please stop this nonsense

@palladin9479

Spoiler

If people don't start calming down it'll be vacation time.

If you truly believe that Tek Syndicate is right and every other resource on the internet is wrong, then you're an out-and-out fanboy.

I actually reproduced his findings. The key had to do with what streaming / recording software your using. Image compression is a very parallel friendly task, it can use anywhere from 1 to infinite (as many as the program supports) cores and split the work evenly across them with near to linear results. The i5's utilized four cores with 3 general purpose ALU's each, the fx8350 had eight cores with 2 ALU's and 2 AGU's, for the vast majority of code it's the ALU's that run the game. Intel has four 256-bit SIMD FPU's while the FX8350 has eight 128-bit FPU's that can combine into four 256-bit FPU's when necessary. The Intel FPU's can process multiple simultaneous instructions if the code is optimized for it. Most games utilize 2~4 cores worth of work, this leaves 50~75% of the fx8350's computing power unutilized while the i5-2500K would lose 0~50%. That unutilized portion is where the streaming software comes in, the AMD chip simply has more excess computing power to spare that's not being used by the game. So unless they restricted the streaming software to a single thread's performance or run it at low priority, it would naturally consume computing power that would otherwise go to playing the game.

I think it is rather important that you name what architecture you are talking about, and not just the manufacturer.
Also encoding images is multi-core friendly, no doubt. But as any other instruction stream it benefit with nice warm registers, so coders would have a harder time pulling it to more than a certain amount of cores before actually losing performance(or not gaining any performance increase, depends on how you look at it).

Haswell had a greater change on their back-end which made their i5 competitive on gaming and streaming.

This changes a bit with the i7 (the i7 is not more expensive then the i5 to produce, they are the exact same chip, i5's are just i7's with a single flag disabled). The i7 allows what would normally be underutilized CPU resources (the extra ALU or FPU per core) to be used for other things, this would somewhat lesson the impact the streaming software had on the game itself. Unless a program is heavily optimized for a specific uArch there is always times when an ALU or two isn't being used. Haswell changed this further as it introduced a fourth ALU into each core and thus each Haswell core has 2x the integer processing resources as a PD core. You won't really see much difference with the i5 as the exposed four cores limits the amount of traffic that can be simultaneously scheduled onto it at once. With Haswell i7's on the other hand you now have significantly more underutilized CPU resources then you had before, two threads scheduled onto the same core would each have 2 ALU's to work with vs 1.5 on SB/IB and the 4 per module (2 per core) of PD. Haswell i7 essentially does away with any competitive advantage AMD's PD design had in multi-threading but not on the i5.

Well, the 4770k and 4670k is not the same chip. the 4770k have a larger cache.
But, 4770k with a defected cache can and would certainly be sold as a 4670k(with disabled ht).
It is essentially the same thing with the 8350, 6350 and 4350.
Haswell did introduce an improve fetch and decoders also.

 

[palladin9479]

Also encoding images is multi-core friendly, no doubt. But as any other instruction stream it benefit with nice warm registers, so coders would have a harder time pulling it to more than a certain amount of cores before actually losing performance(or not gaining any performance increase, depends on how you look at it).

Encoding scales nearly linearly into infinity. The ability to process in parallel is directly related to how well the work can be broken into segments. You won't use a single instruction stream to compress an image (that is by definition single threaded), you'll use several streams each allocated a segment of work. If your compressing a frame of video capture and want to do it across four cores, you would break the frame into four pieces and assign each to a different worker thread, when all four are done you reassemble the results. The fact that not all four pieces will be equally compressible and that you'll need some control logic is what makes it near linear vs strictly linear. You can thus divide the frame into an infinite number of sub-frames for work distribution until you get to the point where the act of controlling the work becomes more process intensive then actually doing the work, that typically doesn't happen until you hit a few thousand nodes. Major production companies utilize render farms composed of hundreds of separate nodes that all work together in this distributed model to render CGI for our current movies.

By that way that's exactly how GPU's rasterize a 3D image into a 2D picture. They divide up the 2D picture into pixels then assign different pixels to different graphics processing elements (shader cores / texture units / ect.). The final result is then put together inside the frame buffer and exported to the screen for you to see.

So yes the fx8350 worked better then i5-2500K and about the same as the i7-2600K for playing a game while streaming the results via real time video compression. This is predominately a result of the fx8350 having more of it's CPU processing power exposed to the software for utilization. It's really not that hard to understand.

 

[vmN]

We are basically agreeing, but my point is, that the general person would be streaming with 30/60FPS. This can be divided over many cores, that is truly correct, but there will be a point where you will divide it to more cores where you wont get any performance increase. If it wasn't streaming where you were 'limited' by a certain amount of frames per second, then yes, it would be the best scalable process.

I never argued about the PD been stronger than SB(Or the other way around).

EDIT:You are comparing an SB i5 to PD, meanwhile I'm comparing an haswell i5 to PD.

 

[palladin9479]

We are basically agreeing, but my point is, that the general person would be streaming with 30/60FPS. This can be divided over many cores, that is truly correct, but there will be a point where you will divide it to more cores where you wont get any performance increase. If it wasn't streaming where you were 'limited' by a certain amount of frames per second, then yes, it would be the best scalable process.

I never argued about the PD been stronger than SB(Or the other way around).

EDIT:You are comparing an SB i5 to PD, meanwhile I'm comparing an haswell i5 to PD.

The TS article the above poster referred to, and the one you criticized, was fx8350 vs i5-2500K and i7-2600K, not vs Haswell.

The refresh rate has absolutely zero to do with how effectively you can divide up the work. The only thing that matters is what is the smallest unit of division and whether the algorithm utilizes dynamic or static scaling. A static scaling model would divide up the total work into equal segments and then dispatch then to the worker threads, dynamic models would do some heuristics and history analysis to determine if some area's of the work are more intensive then others and then try to divide up the work according to total processing need instead of equal unit distribution.

Because streaming has real time requirements they utilize the same model that VTC processors use in that the compression quality is dynamically scaled based on past rendering times. If a frame is taking too long it will get dropped and the next frame will be rendered at a lower Q value, if a frame is done early then the next frame will be rendered with a slightly higher Q value. Some algorithms will utilize a form of pre-rendering analysis to quickly determine optimum Q values to get the frame done within the allocated time frame. In all cases the work is easily divided up amongst a near infinite number of targets because of how large the screen your talking about is. Assuming 16x16 segments are used (256 pixels of work per unit) a 1080 resolution would have 8,100 units worth of work to be distributed. If you use 32x32 segments (1024 pixels) you get 2025 units or work to be distributed. You can keep going bigger and bigger but even at 512x512, which is f*cking huge, you get 8 units of work to be distributed. And so having four to six additional cores of unused processing (most games are two heavy cores and 1~2 worth of random extra threads) power is easily a big boost for additional rendering.

So what you experience is that, under normal play, the Intel i5 (any of them) would be rendering higher FPS then the fx8350 due to it's advantage in single thread performance accelerating those two heavy threads. Turn on the streaming software and enforce a static Q value with the algorithm set to render on four threads and you'll see some interesting events taking place. The fx8350 slows down only slightly, due to the 10~20% penalty it takes when both cores on a module are active, the i5 takes a big hit as it's got to share those four cores with both the compression software and the active game. Most benchmarks we use today leave large portions of the CPU underutilized, this is evident when you see i3's scoring similarly to i5's (same clock rate) even though the i3 is, quite literally, half the CPU of an i5.

 

[vmN]

@palladin9479

Spoiler

The TS article the above poster referred to, and the one you criticized, was fx8350 vs i5-2500K and i7-2600K, not vs Haswell.

He was the one dragging it in. It started with haswell.
My comment:

I do believe the 4670k, is better than the fx 8350 at streaming.

@palladin9479

Spoiler

So what you experience is that, under normal play, the Intel i5 (any of them) would be rendering higher FPS then the fx8350 due to it's advantage in single thread performance accelerating those two heavy threads. Turn on the streaming software and enforce a static Q value with the algorithm set to render on four threads and you'll see some interesting events taking place. The fx8350 slows down only slightly, due to the 10~20% penalty it takes when both cores on a module are active, the i5 takes a big hit as it's got to share those four cores with both the compression software and the active game. Most benchmarks we use today leave large portions of the CPU underutilized, this is evident when you see i3's scoring similarly to i5's (same clock rate) even though the i3 is, quite literally, half the CPU of an i5.

That certainly depends on the game then. As haswell SIMD are superior than PD, and shouldn't have the huge fallback.
Games rely mostly on ALU (As you previously mentioned). And piledriver sharing their FPU could be a holdback.

 

[JOOK-D]

If people don't start calming down it'll be vacation time.

If you truly believe that Tek Syndicate is right and every other resource on the internet is wrong, then you're an out-and-out fanboy.

I actually reproduced his findings. The key had to do with what streaming / recording software your using. Image compression is a very parallel friendly task, it can use anywhere from 1 to infinite (as many as the program supports) cores and split the work evenly across them with near to linear results. The i5's utilized four cores with 3 general purpose ALU's each, the fx8350 had eight cores with 2 ALU's and 2 AGU's, for the vast majority of code it's the ALU's that run the game. Intel has four 256-bit SIMD FPU's while the FX8350 has eight 128-bit FPU's that can combine into four 256-bit FPU's when necessary. The Intel FPU's can process multiple simultaneous instructions if the code is optimized for it. Most games utilize 2~4 cores worth of work, this leaves 50~75% of the fx8350's computing power unutilized while the i5-2500K would lose 0~50%. That unutilized portion is where the streaming software comes in, the AMD chip simply has more excess computing power to spare that's not being used by the game. So unless they restricted the streaming software to a single thread's performance or run it at low priority, it would naturally consume computing power that would otherwise go to playing the game.

This changes a bit with the i7 (the i7 is not more expensive then the i5 to produce, they are the exact same chip, i5's are just i7's with a single flag disabled). The i7 allows what would normally be underutilized CPU resources (the extra ALU or FPU per core) to be used for other things, this would somewhat lesson the impact the streaming software had on the game itself. Unless a program is heavily optimized for a specific uArch there is always times when an ALU or two isn't being used. Haswell changed this further as it introduced a fourth ALU into each core and thus each Haswell core has 2x the integer processing resources as a PD core. You won't really see much difference with the i5 as the exposed four cores limits the amount of traffic that can be simultaneously scheduled onto it at once. With Haswell i7's on the other hand you now have significantly more underutilized CPU resources then you had before, two threads scheduled onto the same core would each have 2 ALU's to work with vs 1.5 on SB/IB and the 4 per module (2 per core) of PD. Haswell i7 essentially does away with any competitive advantage AMD's PD design had in multi-threading but not on the i5.

Please remember Teksyndicate was comparing the 8350 to the i5-2500 and i7-2600. At that time the fx8350 was better at streaming then the i5-2500 about about equal to the i7-2600 due to the above stated reasons. These findings actually mirrored many other places findings but were worded in different ways. Gaming + streaming is easily a heavily multi-threaded workload, something PD was already acknowledged as being great at.

We are basically agreeing, but my point is, that the general person would be streaming with 30/60FPS. This can be divided over many cores, that is truly correct, but there will be a point where you will divide it to more cores where you wont get any performance increase. If it wasn't streaming where you were 'limited' by a certain amount of frames per second, then yes, it would be the best scalable process.

I never argued about the PD been stronger than SB(Or the other way around).

EDIT:You are comparing an SB i5 to PD, meanwhile I'm comparing an haswell i5 to PD.

The TS article the above poster referred to, and the one you criticized, was fx8350 vs i5-2500K and i7-2600K, not vs Haswell.

The refresh rate has absolutely zero to do with how effectively you can divide up the work. The only thing that matters is what is the smallest unit of division and whether the algorithm utilizes dynamic or static scaling. A static scaling model would divide up the total work into equal segments and then dispatch then to the worker threads, dynamic models would do some heuristics and history analysis to determine if some area's of the work are more intensive then others and then try to divide up the work according to total processing need instead of equal unit distribution.

Because streaming has real time requirements they utilize the same model that VTC processors use in that the compression quality is dynamically scaled based on past rendering times. If a frame is taking too long it will get dropped and the next frame will be rendered at a lower Q value, if a frame is done early then the next frame will be rendered with a slightly higher Q value. Some algorithms will utilize a form of pre-rendering analysis to quickly determine optimum Q values to get the frame done within the allocated time frame. In all cases the work is easily divided up amongst a near infinite number of targets because of how large the screen your talking about is. Assuming 16x16 segments are used (256 pixels of work per unit) a 1080 resolution would have 8,100 units worth of work to be distributed. If you use 32x32 segments (1024 pixels) you get 2025 units or work to be distributed. You can keep going bigger and bigger but even at 512x512, which is f*cking huge, you get 8 units of work to be distributed. And so having four to six additional cores of unused processing (most games are two heavy cores and 1~2 worth of random extra threads) power is easily a big boost for additional rendering.

So what you experience is that, under normal play, the Intel i5 (any of them) would be rendering higher FPS then the fx8350 due to it's advantage in single thread performance accelerating those two heavy threads. Turn on the streaming software and enforce a static Q value with the algorithm set to render on four threads and you'll see some interesting events taking place. The fx8350 slows down only slightly, due to the 10~20% penalty it takes when both cores on a module are active, the i5 takes a big hit as it's got to share those four cores with both the compression software and the active game. Most benchmarks we use today leave large portions of the CPU underutilized, this is evident when you see i3's scoring similarly to i5's (same clock rate) even though the i3 is, quite literally, half the CPU of an i5.

'Scuse my behaviour, I don't respond well to random insults being thrown around. hehe. Also my post was in reference to game performance not the streaming performance, though I probably should have made that clear.

Thanks for the response, it's interesting to hear the technical explanation. I do not understand why this thread is continuing however, OP seems to be far gone and it's a rather old thread spawning not much else other than arguments. I'll be unfollowing.

Cheers.