AMD CPU speculation... and expert conjecture

[The Q6660 Inside]

http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/

This would be absolutely awesome! Sadly none of those specs even come close to matching up (25nm, 6.0Ghz with 75w TDP along with 12 cores)

I think that if this processor was released hafijur, could not troll about how Intel has a huge advantage in power consumption.

someone missed april 1st, but its still quite comical, the funny part being someone actually believing it enough to post the information as if it were possible (wccf).

@gamerk316:

Why do I get the feeling that Linux benchmarks will, once again, be dramatically faster for AMD architecture when all the dust settles?

Oh, that's right...the Linux kernel is light years ahead of Windows kernel...

Spoiler

No, it's because the Linux kernel is much less for graphics than performance. It's all about performance, and works better obviously, with more physical cores, and something about AMD's architecture. I have to keep telling you this. In Real Life (not sure if you know about it), almost every computer that is sold either comes with Windows or OSx. Also, Linux is (at least to me) no more useful than to code and create programs, which, I don't do very much, so there's no point in having it. Plus for the gaming market, there may be a couple games that run much better on it, however, that's only ~ 5% of the whole game populous, so there's no advantage there.

You're doing what people thought I was doing: Talking about stuff that no one cares about and trolling.

Cool, Linux works for you, and it benefits AMD. However, it's not a large enough advantage to use in the real world. Now instead of blabbering about how amazing it is, put up some actual information [in-fer-mey-shuhn] about it, so we can build on it, instead of you starting another useless and random flame war.

I would not say that Linux benefits AMD. There is nothing in Linux designed specifically for AMD. Linux works for essentially processor arch and brand. What happens is that Windows is often optimized for Intel: Wintel.

Linux is used in Real Life / Real World. Linux is used in supercomputers, servers, workstations, laptops, desktops, game consoles... Linux is used by end users and by Governments. Toms has a recent article about the market share of linux and concludes that there are more Linux users than Windows users (Toms counts Android users as well).

I could agree that Linux is used that much (didn't really say against it), however the compatibility for it is really for those scientists and custom coders, I would say. On a realistic basis you wouldn't use Linux on a gaming rig or rendering rig (rendering is arguable, but most likely not). Even there, where there are the performance gains of AMD for Linux, the actual software (mostly games) isn't compatible (unless you sit there and re-code the whole thing), thus making the whole system null.

Some thoughts:

Spoiler

I would say for myself, the high core count CPU's come in handy with ArmA / DayZ, which is my main game, and benefits even more if each processor core is stronger. Thus is why I'm getting a 3930k, since the base core power is much higher than most AMD's. In the cfg you can set how many cores it recognizes, set the affinity, and in the Properties or DayZ Commander, you can set instructions to how many threads it recognizes. The thing that I think is cool though, is that if you set it to more than the actual threads it has, it goes back to thread 0, and basically "using it again". So for example, it already recognizes 1 thread, which is thread 0. The start instructions for extra threads, I believe, is " ex_Thread= X " (something like that) X being the extra thread count.

So with my current 4 core / thread CPU, I would set it to "ex_thread = 3", so then it uses all 4 threads. However, if I set the number to 4, it'll use all the threads, but then use thread 0, twice, and so on with however many numbers you put. For that CPU, it barely helps, and make's it a little hotter, but it's possible.

That's why with the 3930k, I can use the 12 threads, and then double over them to get better performance. I'll test it, but it shouldn't be a problem if it's overclocked to 4.5 Ghz.

STOP MENTIONING THE 3930K GOSHDARNIT, MUST I GET LN2 AND A 980X TO SHUT YOU UP :3 Other than that, I must agree that moar coars is the future TBH.

 

[noob2222]

@juanrga

you missed the point, gaming will be slower on the APU, period.

aside from that, what about Dgpu performance with openCL?

even your chart shows how fast it is with a 7770 vs APU. AMD is pushing OpenCL, but its not striclty limited to IGP of the APU.

 

[GOM3RPLY3R]

http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/

This would be absolutely awesome! Sadly none of those specs even come close to matching up (25nm, 6.0Ghz with 75w TDP along with 12 cores)

I think that if this processor was released hafijur, could not troll about how Intel has a huge advantage in power consumption.

someone missed april 1st, but its still quite comical, the funny part being someone actually believing it enough to post the information as if it were possible (wccf).

@gamerk316:

Why do I get the feeling that Linux benchmarks will, once again, be dramatically faster for AMD architecture when all the dust settles?

Oh, that's right...the Linux kernel is light years ahead of Windows kernel...

Spoiler

No, it's because the Linux kernel is much less for graphics than performance. It's all about performance, and works better obviously, with more physical cores, and something about AMD's architecture. I have to keep telling you this. In Real Life (not sure if you know about it), almost every computer that is sold either comes with Windows or OSx. Also, Linux is (at least to me) no more useful than to code and create programs, which, I don't do very much, so there's no point in having it. Plus for the gaming market, there may be a couple games that run much better on it, however, that's only ~ 5% of the whole game populous, so there's no advantage there.

You're doing what people thought I was doing: Talking about stuff that no one cares about and trolling.

Cool, Linux works for you, and it benefits AMD. However, it's not a large enough advantage to use in the real world. Now instead of blabbering about how amazing it is, put up some actual information [in-fer-mey-shuhn] about it, so we can build on it, instead of you starting another useless and random flame war.

I would not say that Linux benefits AMD. There is nothing in Linux designed specifically for AMD. Linux works for essentially processor arch and brand. What happens is that Windows is often optimized for Intel: Wintel.

Linux is used in Real Life / Real World. Linux is used in supercomputers, servers, workstations, laptops, desktops, game consoles... Linux is used by end users and by Governments. Toms has a recent article about the market share of linux and concludes that there are more Linux users than Windows users (Toms counts Android users as well).

I could agree that Linux is used that much (didn't really say against it), however the compatibility for it is really for those scientists and custom coders, I would say. On a realistic basis you wouldn't use Linux on a gaming rig or rendering rig (rendering is arguable, but most likely not). Even there, where there are the performance gains of AMD for Linux, the actual software (mostly games) isn't compatible (unless you sit there and re-code the whole thing), thus making the whole system null.

Some thoughts:

Spoiler

I would say for myself, the high core count CPU's come in handy with ArmA / DayZ, which is my main game, and benefits even more if each processor core is stronger. Thus is why I'm getting a 3930k, since the base core power is much higher than most AMD's. In the cfg you can set how many cores it recognizes, set the affinity, and in the Properties or DayZ Commander, you can set instructions to how many threads it recognizes. The thing that I think is cool though, is that if you set it to more than the actual threads it has, it goes back to thread 0, and basically "using it again". So for example, it already recognizes 1 thread, which is thread 0. The start instructions for extra threads, I believe, is " ex_Thread= X " (something like that) X being the extra thread count.

So with my current 4 core / thread CPU, I would set it to "ex_thread = 3", so then it uses all 4 threads. However, if I set the number to 4, it'll use all the threads, but then use thread 0, twice, and so on with however many numbers you put. For that CPU, it barely helps, and make's it a little hotter, but it's possible.

That's why with the 3930k, I can use the 12 threads, and then double over them to get better performance. I'll test it, but it shouldn't be a problem if it's overclocked to 4.5 Ghz.

STOP MENTIONING THE 3930K GOSHDARNIT, MUST I GET LN2 AND A 980X TO SHUT YOU UP :3 Other than that, I must agree that moar coars is the future TBH.

1. It's optional to read, that's why I made it a spoiler.

2. I'm using it as an EXAMPLE, not as if it's the best. It's the best example I can bring out.

Now that you know that, if you don't like what I'm doing, then kindly, GTFO or just shut up about it. It's not like I'm saying a Pentium 3 @ 1 ghz is better than an e5 Xeon @ 4 Ghz. (Btw, can't wait for the 10 Core Xeons. )

 

[The Q6660 Inside]

http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/

This would be absolutely awesome! Sadly none of those specs even come close to matching up (25nm, 6.0Ghz with 75w TDP along with 12 cores)

I think that if this processor was released hafijur, could not troll about how Intel has a huge advantage in power consumption.

someone missed april 1st, but its still quite comical, the funny part being someone actually believing it enough to post the information as if it were possible (wccf).

@gamerk316:

Why do I get the feeling that Linux benchmarks will, once again, be dramatically faster for AMD architecture when all the dust settles?

Oh, that's right...the Linux kernel is light years ahead of Windows kernel...

Spoiler

No, it's because the Linux kernel is much less for graphics than performance. It's all about performance, and works better obviously, with more physical cores, and something about AMD's architecture. I have to keep telling you this. In Real Life (not sure if you know about it), almost every computer that is sold either comes with Windows or OSx. Also, Linux is (at least to me) no more useful than to code and create programs, which, I don't do very much, so there's no point in having it. Plus for the gaming market, there may be a couple games that run much better on it, however, that's only ~ 5% of the whole game populous, so there's no advantage there.

You're doing what people thought I was doing: Talking about stuff that no one cares about and trolling.

Cool, Linux works for you, and it benefits AMD. However, it's not a large enough advantage to use in the real world. Now instead of blabbering about how amazing it is, put up some actual information [in-fer-mey-shuhn] about it, so we can build on it, instead of you starting another useless and random flame war.

I would not say that Linux benefits AMD. There is nothing in Linux designed specifically for AMD. Linux works for essentially processor arch and brand. What happens is that Windows is often optimized for Intel: Wintel.

Linux is used in Real Life / Real World. Linux is used in supercomputers, servers, workstations, laptops, desktops, game consoles... Linux is used by end users and by Governments. Toms has a recent article about the market share of linux and concludes that there are more Linux users than Windows users (Toms counts Android users as well).

I could agree that Linux is used that much (didn't really say against it), however the compatibility for it is really for those scientists and custom coders, I would say. On a realistic basis you wouldn't use Linux on a gaming rig or rendering rig (rendering is arguable, but most likely not). Even there, where there are the performance gains of AMD for Linux, the actual software (mostly games) isn't compatible (unless you sit there and re-code the whole thing), thus making the whole system null.

Some thoughts:

Spoiler

I would say for myself, the high core count CPU's come in handy with ArmA / DayZ, which is my main game, and benefits even more if each processor core is stronger. Thus is why I'm getting a 3930k, since the base core power is much higher than most AMD's. In the cfg you can set how many cores it recognizes, set the affinity, and in the Properties or DayZ Commander, you can set instructions to how many threads it recognizes. The thing that I think is cool though, is that if you set it to more than the actual threads it has, it goes back to thread 0, and basically "using it again". So for example, it already recognizes 1 thread, which is thread 0. The start instructions for extra threads, I believe, is " ex_Thread= X " (something like that) X being the extra thread count.

So with my current 4 core / thread CPU, I would set it to "ex_thread = 3", so then it uses all 4 threads. However, if I set the number to 4, it'll use all the threads, but then use thread 0, twice, and so on with however many numbers you put. For that CPU, it barely helps, and make's it a little hotter, but it's possible.

That's why with the 3930k, I can use the 12 threads, and then double over them to get better performance. I'll test it, but it shouldn't be a problem if it's overclocked to 4.5 Ghz.

STOP MENTIONING THE 3930K GOSHDARNIT, MUST I GET LN2 AND A 980X TO SHUT YOU UP :3 Other than that, I must agree that moar coars is the future TBH.

1. It's optional to read, that's why I made it a spoiler.

2. I'm using it as an EXAMPLE, not as if it's the best. It's the best example I can bring out.

Now that you know that, if you don't like what I'm doing, then kindly, GTFO or just shut up about it. It's not like I'm saying a Pentium 3 @ 1 ghz is better than an e5 Xeon @ 4 Ghz. (Btw, can't wait for the 10 Core Xeons. )

It was joking, note that I agree with your post :3 Use the 3970X as an example to avoid this issue :3

BTW: Can't wait for the 32 Core Opterons :3 .

 

[GOM3RPLY3R]

Spoiler

http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/

This would be absolutely awesome! Sadly none of those specs even come close to matching up (25nm, 6.0Ghz with 75w TDP along with 12 cores)

I think that if this processor was released hafijur, could not troll about how Intel has a huge advantage in power consumption.

someone missed april 1st, but its still quite comical, the funny part being someone actually believing it enough to post the information as if it were possible (wccf).

@gamerk316:

Why do I get the feeling that Linux benchmarks will, once again, be dramatically faster for AMD architecture when all the dust settles?

Oh, that's right...the Linux kernel is light years ahead of Windows kernel...

Spoiler

No, it's because the Linux kernel is much less for graphics than performance. It's all about performance, and works better obviously, with more physical cores, and something about AMD's architecture. I have to keep telling you this. In Real Life (not sure if you know about it), almost every computer that is sold either comes with Windows or OSx. Also, Linux is (at least to me) no more useful than to code and create programs, which, I don't do very much, so there's no point in having it. Plus for the gaming market, there may be a couple games that run much better on it, however, that's only ~ 5% of the whole game populous, so there's no advantage there.

You're doing what people thought I was doing: Talking about stuff that no one cares about and trolling.

Cool, Linux works for you, and it benefits AMD. However, it's not a large enough advantage to use in the real world. Now instead of blabbering about how amazing it is, put up some actual information [in-fer-mey-shuhn] about it, so we can build on it, instead of you starting another useless and random flame war.

I would not say that Linux benefits AMD. There is nothing in Linux designed specifically for AMD. Linux works for essentially processor arch and brand. What happens is that Windows is often optimized for Intel: Wintel.

Linux is used in Real Life / Real World. Linux is used in supercomputers, servers, workstations, laptops, desktops, game consoles... Linux is used by end users and by Governments. Toms has a recent article about the market share of linux and concludes that there are more Linux users than Windows users (Toms counts Android users as well).

I could agree that Linux is used that much (didn't really say against it), however the compatibility for it is really for those scientists and custom coders, I would say. On a realistic basis you wouldn't use Linux on a gaming rig or rendering rig (rendering is arguable, but most likely not). Even there, where there are the performance gains of AMD for Linux, the actual software (mostly games) isn't compatible (unless you sit there and re-code the whole thing), thus making the whole system null.

Some thoughts:

Spoiler

I would say for myself, the high core count CPU's come in handy with ArmA / DayZ, which is my main game, and benefits even more if each processor core is stronger. Thus is why I'm getting a 3930k, since the base core power is much higher than most AMD's. In the cfg you can set how many cores it recognizes, set the affinity, and in the Properties or DayZ Commander, you can set instructions to how many threads it recognizes. The thing that I think is cool though, is that if you set it to more than the actual threads it has, it goes back to thread 0, and basically "using it again". So for example, it already recognizes 1 thread, which is thread 0. The start instructions for extra threads, I believe, is " ex_Thread= X " (something like that) X being the extra thread count.

So with my current 4 core / thread CPU, I would set it to "ex_thread = 3", so then it uses all 4 threads. However, if I set the number to 4, it'll use all the threads, but then use thread 0, twice, and so on with however many numbers you put. For that CPU, it barely helps, and make's it a little hotter, but it's possible.

That's why with the 3930k, I can use the 12 threads, and then double over them to get better performance. I'll test it, but it shouldn't be a problem if it's overclocked to 4.5 Ghz.

STOP MENTIONING THE 3930K GOSHDARNIT, MUST I GET LN2 AND A 980X TO SHUT YOU UP :3 Other than that, I must agree that moar coars is the future TBH.

1. It's optional to read, that's why I made it a spoiler.

2. I'm using it as an EXAMPLE, not as if it's the best. It's the best example I can bring out.

Now that you know that, if you don't like what I'm doing, then kindly, GTFO or just shut up about it. It's not like I'm saying a Pentium 3 @ 1 ghz is better than an e5 Xeon @ 4 Ghz. (Btw, can't wait for the 10 Core Xeons. )

It was joking, note that I agree with your post :3 Use the 3970X as an example to avoid this issue :3

BTW: Can't wait for the 32 Core Opterons :3 .

True, even though the 3970X and 3930k performance are basically the same (3970X is just clocked higher really).

And... That's alot of cores....

へ　　　へ
の　　　の
　　も
　　へ

 

[gamerk316]

As I said above performance depends on IPC, number of cores, frequency... If "the CPU is getting its work done", as you say, then adding more cores, or increasing the IPC, or overclocking, or any combination of those will not increase performance. Before, we discussed benchmarks showing how an OC 4770k was obtaining the same FPS than at stock.

Ahh, but here's the point you miss:

Likely GPU bottleneck somewhere around 45 FPS or so. So you wouldn't expect CPU gains to lead to any farther increase in performance past that point.

You can continue negating benchmarks, you can continue negating that the FX-8350 @ 4Ghz is much faster than the FX-4300 @ 3.8 GHz and that the difference in FPS cannot be explained by freq. You can also negate that the i7-3770k @ 3.5GHz is much faster than the i5-3570k @ 3.4GHz and that the difference in FPS cannot be explained by freq.

Remember the Crysis 3 CPU usage for the i5 CPU? The one that showed the cores loaded to about 85% each? Now subtract 15% IPC from those numbers to simulate FX. Oh right, that would get us somewhere around 100% CPU usage, otherwise known as a CPU bottleneck.

In short: The FX-4300, at stock clocks, is CPU bottlenecked in Crysis 3. Hence the FX-8350's performance advantage. Heck, based on the results, it looks like a FX-4300 @ 4.7 is ALSO CPU bottlencecked, though I'd want to look at CPU usage statistics to confirm.

Also note: Intel Quads at stock offer significantly higher performance at a lower clock, so they AREN'T CPU bottlenecked.

Also note 2: The i3 at stock outperforms the FX-4300 at stock. Again: IPC edge of the duo outperforming the quad (and performing about the same as the FX-6300). Remember, 9 worker threads, but two do about 80% of the work. As a result, the higher IPC of the i3 is more important then the extra cores of the FX-4300 or FX-6300.

Hence again, why 15% IPC gains for SR would be HUGE for AMD.

--------------------------------------------------

And note: I am in no way advocating that duo's are a good way to build a gaming system around. I'm simply pointing out that with enough clock/IPC, duo's can still outperform quads in gaming, even as games scale to beyond two processing cores. Same logic applies to quads. Simply "adding more cores" will not improve performance if per-core performance is lacking.

 

[jdwii]

having some basic programming knowledge and being able to look at and understand what benchmarks mean i agree with gamer 100% here

 

[noob2222]

As I said above performance depends on IPC, number of cores, frequency... If "the CPU is getting its work done", as you say, then adding more cores, or increasing the IPC, or overclocking, or any combination of those will not increase performance. Before, we discussed benchmarks showing how an OC 4770k was obtaining the same FPS than at stock.

Ahh, but here's the point you miss:

Likely GPU bottleneck somewhere around 45 FPS or so. So you wouldn't expect CPU gains to lead to any farther increase in performance past that point.

You can continue negating benchmarks, you can continue negating that the FX-8350 @ 4Ghz is much faster than the FX-4300 @ 3.8 GHz and that the difference in FPS cannot be explained by freq. You can also negate that the i7-3770k @ 3.5GHz is much faster than the i5-3570k @ 3.4GHz and that the difference in FPS cannot be explained by freq.

Remember the Crysis 3 CPU usage for the i5 CPU? The one that showed the cores loaded to about 85% each? Now subtract 15% IPC from those numbers to simulate FX. Oh right, that would get us somewhere around 100% CPU usage, otherwise known as a CPU bottleneck.

In short: The FX-4300, at stock clocks, is CPU bottlenecked in Crysis 3. Hence the FX-8350's performance advantage. Heck, based on the results, it looks like a FX-4300 @ 4.7 is ALSO CPU bottlencecked, though I'd want to look at CPU usage statistics to confirm.

Also note: Intel Quads at stock offer significantly higher performance at a lower clock, so they AREN'T CPU bottlenecked.

Also note 2: The i3 at stock outperforms the FX-4300 at stock. Again: IPC edge of the duo outperforming the quad (and performing about the same as the FX-6300). Remember, 9 worker threads, but two do about 80% of the work. As a result, the higher IPC of the i3 is more important then the extra cores of the FX-4300 or FX-6300.

Hence again, why 15% IPC gains for SR would be HUGE for AMD.

--------------------------------------------------

And note: I am in no way advocating that duo's are a good way to build a gaming system around. I'm simply pointing out that with enough clock/IPC, duo's can still outperform quads in gaming, even as games scale to beyond two processing cores. Same logic applies to quads. Simply "adding more cores" will not improve performance if per-core performance is lacking.

while I somewhat agree here, the 4300 isn't a true quad, its more like the i3 than a quad. remember when amd's cmt uses 2 cores on the same module, they fight for resources and the single core ipc drops. this is the downfall of cmt, but sr doubling the dispatcher should help this. i3 doesn't suffer this single core performance drop by utilizing ht, but the overall "quad boost" is quite small in comparison.

If you take the 8350 and force the game to use 1 core from each module, it will significantly outperform the 4300 alone. this is also why the phenom II still outperforms the 4300. this is also why im against a quad APU for a true gaming machine.

kaveri will shed some light on this exact issue when comparing single core to full cpu usage.

 

[palladin9479]

Remember APU's are not designed to be used in high end gaming rigs. Using any dGPU with an APU completely defeats the purpose of the APU, might as well just use a regular CPU for more value. APU's are for SFF and miniature machines, and in that realm they dominate.

 

[juanrga]

@gamerk316:

Why do I get the feeling that Linux benchmarks will, once again, be dramatically faster for AMD architecture when all the dust settles?

Oh, that's right...the Linux kernel is light years ahead of Windows kernel...

Spoiler

No, it's because the Linux kernel is much less for graphics than performance. It's all about performance, and works better obviously, with more physical cores, and something about AMD's architecture. I have to keep telling you this. In Real Life (not sure if you know about it), almost every computer that is sold either comes with Windows or OSx. Also, Linux is (at least to me) no more useful than to code and create programs, which, I don't do very much, so there's no point in having it. Plus for the gaming market, there may be a couple games that run much better on it, however, that's only ~ 5% of the whole game populous, so there's no advantage there.

You're doing what people thought I was doing: Talking about stuff that no one cares about and trolling.

Cool, Linux works for you, and it benefits AMD. However, it's not a large enough advantage to use in the real world. Now instead of blabbering about how amazing it is, put up some actual information [in-fer-mey-shuhn] about it, so we can build on it, instead of you starting another useless and random flame war.

I would not say that Linux benefits AMD. There is nothing in Linux designed specifically for AMD. Linux works for essentially processor arch and brand. What happens is that Windows is often optimized for Intel: Wintel.

Linux is used in Real Life / Real World. Linux is used in supercomputers, servers, workstations, laptops, desktops, game consoles... Linux is used by end users and by Governments. Toms has a recent article about the market share of linux and concludes that there are more Linux users than Windows users (Toms counts Android users as well).

I could agree that Linux is used that much (didn't really say against it), however the compatibility for it is really for those scientists and custom coders, I would say. On a realistic basis you wouldn't use Linux on a gaming rig or rendering rig (rendering is arguable, but most likely not). Even there, where there are the performance gains of AMD for Linux, the actual software (mostly games) isn't compatible (unless you sit there and re-code the whole thing), thus making the whole system null.

Some thoughts:

Spoiler

I would say for myself, the high core count CPU's come in handy with ArmA / DayZ, which is my main game, and benefits even more if each processor core is stronger. Thus is why I'm getting a 3930k, since the base core power is much higher than most AMD's. In the cfg you can set how many cores it recognizes, set the affinity, and in the Properties or DayZ Commander, you can set instructions to how many threads it recognizes. The thing that I think is cool though, is that if you set it to more than the actual threads it has, it goes back to thread 0, and basically "using it again". So for example, it already recognizes 1 thread, which is thread 0. The start instructions for extra threads, I believe, is " ex_Thread= X " (something like that) X being the extra thread count.

So with my current 4 core / thread CPU, I would set it to "ex_thread = 3", so then it uses all 4 threads. However, if I set the number to 4, it'll use all the threads, but then use thread 0, twice, and so on with however many numbers you put. For that CPU, it barely helps, and make's it a little hotter, but it's possible.

That's why with the 3930k, I can use the 12 threads, and then double over them to get better performance. I'll test it, but it shouldn't be a problem if it's overclocked to 4.5 Ghz.

"however the compatibility for it is really for those scientists and custom coders" What compatibility? What are you saying?

Games? Besides traditional linux games, Valve is porting Steam to Linux and Alienware already sell gaming PC with linux pre-installed. The next SteamBox console will use Linux. There are rumours Google plans a linux based console. The PS4 OS is a kind of brother of linux. Besides that you can play many windows games on linux using Wine.

Rendering? The 95% of animation and visual effects companies use Linux: Dreamworks Animation, Pixar, Weta Digital, and Industrial Light & Magic... all them use linux

@juanrga

you missed the point, gaming will be slower on the APU, period.

aside from that, what about Dgpu performance with openCL?

even your chart shows how fast it is with a 7770 vs APU. AMD is pushing OpenCL, but its not striclty limited to IGP of the APU.

As stated before, 4-cores Kaveri would have the performance of a Sandy-Bridge i5. It seems rather good for gaming. Did you remember the rumours about a six-core version? It is not in the 2014 roadmap, but maybe will be in a future roadmap. A six-core Kaveri would be at the i7 level.

I offered one OpenCL benchmark only as example of improvement when the iGPU is used. The main point was HSA and HSA != OpenCL. HSA will provide advantages beyond a traditional CPU + dGPU(OpenCL) arch.

As I said above performance depends on IPC, number of cores, frequency... If "the CPU is getting its work done", as you say, then adding more cores, or increasing the IPC, or overclocking, or any combination of those will not increase performance. Before, we discussed benchmarks showing how an OC 4770k was obtaining the same FPS than at stock.

Ahh, but here's the point you miss:

Likely GPU bottleneck somewhere around 45 FPS or so. So you wouldn't expect CPU gains to lead to any farther increase in performance past that point.

You can continue negating benchmarks, you can continue negating that the FX-8350 @ 4Ghz is much faster than the FX-4300 @ 3.8 GHz and that the difference in FPS cannot be explained by freq. You can also negate that the i7-3770k @ 3.5GHz is much faster than the i5-3570k @ 3.4GHz and that the difference in FPS cannot be explained by freq.

Remember the Crysis 3 CPU usage for the i5 CPU? The one that showed the cores loaded to about 85% each? Now subtract 15% IPC from those numbers to simulate FX. Oh right, that would get us somewhere around 100% CPU usage, otherwise known as a CPU bottleneck.

In short: The FX-4300, at stock clocks, is CPU bottlenecked in Crysis 3. Hence the FX-8350's performance advantage. Heck, based on the results, it looks like a FX-4300 @ 4.7 is ALSO CPU bottlencecked, though I'd want to look at CPU usage statistics to confirm.

Also note: Intel Quads at stock offer significantly higher performance at a lower clock, so they AREN'T CPU bottlenecked.

Also note 2: The i3 at stock outperforms the FX-4300 at stock. Again: IPC edge of the duo outperforming the quad (and performing about the same as the FX-6300). Remember, 9 worker threads, but two do about 80% of the work. As a result, the higher IPC of the i3 is more important then the extra cores of the FX-4300 or FX-6300.

Hence again, why 15% IPC gains for SR would be HUGE for AMD.

--------------------------------------------------

And note: I am in no way advocating that duo's are a good way to build a gaming system around. I'm simply pointing out that with enough clock/IPC, duo's can still outperform quads in gaming, even as games scale to beyond two processing cores. Same logic applies to quads. Simply "adding more cores" will not improve performance if per-core performance is lacking.

Once again. Look at the FPS for the OC FX-4300 and the FX-8350, both are 30 FPS. There is no bottleneck, because OC the 8350 gives an improvement on the FPS that coincides with the OC.

The same about the i5 and i7. Both obtain the same FPS and OC both obtain more FPS.

This is a clear symptom that "Roof" is not threaded enough. That is why the i3 perform so well. Whereas "Welcome" is heavily threaded, and the i7 is much faster than the i5 when HT is activated and the FX-8350 is much faster than the FX-4300.

In "Welcome", the FX-4300 @ stock is loaded at 95%. It is the dual cores, which are loaded at 100%.

I am going to explain this again. Performance depends on IPC, freq. number of cores.... yes if you OC enough a hypothetical k-series dual core i3 it can outperform a quad i5. In poorly threaded apps, you will probably need a modest OC, in well-threaded apps you would need extreme OC of about 8 GHz to match the quad performance. And don't forget that quads can be also OC.

The same about IPC. Imagine a hypothetical dual core Haswell i3 with an IPC that is 200% that of Ivi-Bridge. That hypothetical dual core would beat an i7. The problem is that Haswell provides only a 5% gain for reasons mentioned before.

The future is not dual core but more cores. Did I already mention that Intel announced its first octo-core for desktop? Did I already mention you that Intel released Haswell i5/i7, but not i3 for desktops? The i3 would appears latter this year, but are not a priority, all gamers that I know are buying i5 and i7.

 

[The Q6660 Inside]

AMD News: AMD VP of channel sales (Roy Taylor): PhysX and CUDA are doomed http://www.tomshardware.com/news/roy-taylor-apu-opencl-cuda-physx,23797.html

 

[GOM3RPLY3R]

AMD News: AMD VP of channel sales (Roy Taylor): PhysX and CUDA are doomed http://www.tomshardware.com/news/roy-taylor-apu-opencl-cuda-physx,23797.html

This is hilarious. -_-

I don't see what talking about Intel processors and people adopting OpenCL have to do with CUDA and Phyx being doomed. Sure AMD runs a little better with OpenCL, however, the APU isn't meant for high end gaming like you see with the 680, 780, Titan, and 690. Even with a claim that an AMD employee has, it doesn't mean that that top sellers are going to adopt it. It's all about popularity.

For example, everyone here can agree, the Xbox One is a piece of doodoo. However, since Xbox has "hypnotized" their comsumers into thinking everything about Xbox is better, they have the lead in sales despite the amount of players going to XboX 360 to PS4.

Thus, even so more people will be adopting OpenCL, it doesn't directly mean that like DICE, Polyphony Digital, Activision, Bohemia Interactive, and so on, will indefinitely he going for it. Honeslty, OpenCL isn't that great. It performs much better obviously, but in the end, the visuals aren't better. You can run a high graphics benchmark (Unigine Valley for example) on OpenCL, and it'll do great with the AMD card, however, on Direct X or PhysX, it does run a few to i'd say 10 frames slower, but the end graphics output looks so much cleaner and nicer.

It's really a hit or miss, so we'll have to see. But unless someone can really manipulate the OpenCL code, or the code becomes better, I don't really see that happening in the least.

 

[The Q6660 Inside]

AMD News: AMD VP of channel sales (Roy Taylor): PhysX and CUDA are doomed http://www.tomshardware.com/news/roy-taylor-apu-opencl-cuda-physx,23797.html

This is hilarious. -_-

I don't see what talking about Intel processors and people adopting OpenCL have to do with CUDA and Phyx being doomed. Sure AMD runs a little better with OpenCL, however, the APU isn't meant for high end gaming like you see with the 680, 780, Titan, and 690. Even with a claim that an AMD employee has, it doesn't mean that that top sellers are going to adopt it. It's all about popularity.

For example, everyone here can agree, the Xbox One is a piece of doodoo. However, since Xbox has "hypnotized" their comsumers into thinking everything about Xbox is better, they have the lead in sales despite the amount of players going to XboX 360 to PS4.

Thus, even so more people will be adopting OpenCL, it doesn't directly mean that like DICE, Polyphony Digital, Activision, Bohemia Interactive, and so on, will indefinitely he going for it. Honeslty, OpenCL isn't that great. It performs much better obviously, but in the end, the visuals aren't better. You can run a high graphics benchmark (Unigine Valley for example) on OpenCL, and it'll do great with the AMD card, however, on Direct X or PhysX, it does run a few to i'd say 10 frames slower, but the end graphics output looks so much cleaner and nicer.

It's really a hit or miss, so we'll have to see. But unless someone can really manipulate the OpenCL code, or the code becomes better, I don't really see that happening in the least.

I agree with you, however, he was trying to note how people generally prefer universal figures\tecnolgies and not proprietary technology. AMD is honestly in a bad position to go out smearing others (Must I bring out those Radeon commercials?). It really all depends on how it goes, we will see in the long run if OpenCL lives up to what its worth. https://www.youtube.com/watch?feature=player_detailpage&v=KbWgUO-Rqcw :33

 

[noob2222]

@juanrga

games have ALWAYS favored cpus with l3 cache. athlon > duron ; pentium > celeron ; phenom > athlon ; fx > apu.

granted 16mb is probably a bit much, intel's normal cpus only have ~6mb and only the extremes have over 10mb. no l3 cache will have lower performance unless by some miracle the cpu core is strong enough to compensate the lack of. even if kaveri is faster than the I5, fx SR will be that much faster.

 

[mayankleoboy1]

Look to what happens when the iGPU is used

For Kaveri, the improvement will be larger with HSA: about 5x the performance of the CPU alone.

Because some game dev says that he sees 5X improvement in games, doesnt mean WinZip compression will get the same benefit. Stop using the same figure for every workload.

 

[juanrga]

'Interestingly the document also made it clears the days of the big core are numbered, leaving the FX-9590 as the final legacy of the FX line.'via VR-Zone

http://wccftech.com/amd-possibly-abandoning-entire-fx-processors-lineup/#ixzz2b9pY9tkQ

AMD only have 4 core apus lined up it seems. The fx9590 was amd last fx series for now at least. AMD are going to the intel route it seems as less cores that are better is the way to go.

eit:amd will still make piledriver server cores though.

The vr-zone / wccftech analysis coincides with what I said in this thread one page ago: no FX steamroller.

No. AMD is not going the "intel route".

In the first place, Intel is following the AMD route and announcing its first octo-core.

In the second place, as mentioned in my previous post, AMD will be releasing the new Warsaw chips with 12/16 core, whereas eliminating the 4/6/8 core variants.

In the third place, AMD is merely replacing a kind of cores by others. The Opteron 8 cores CPUs will be replaced by Berlin 4+8 core APUs: 4 SR cores + 8 CU.

In the fourth place, AMD will continue releasing custom multi-core chips. E.g. the 8 jaguar cores used in the Xbox One and PS4.

 

[esrever]

AMD News: AMD VP of channel sales (Roy Taylor): PhysX and CUDA are doomed http://www.tomshardware.com/news/roy-taylor-apu-opencl-cuda-physx,23797.html

This is hilarious. -_-

I don't see what talking about Intel processors and people adopting OpenCL have to do with CUDA and Phyx being doomed. Sure AMD runs a little better with OpenCL, however, the APU isn't meant for high end gaming like you see with the 680, 780, Titan, and 690. Even with a claim that an AMD employee has, it doesn't mean that that top sellers are going to adopt it. It's all about popularity.

For example, everyone here can agree, the Xbox One is a piece of doodoo. However, since Xbox has "hypnotized" their comsumers into thinking everything about Xbox is better, they have the lead in sales despite the amount of players going to XboX 360 to PS4.

Thus, even so more people will be adopting OpenCL, it doesn't directly mean that like DICE, Polyphony Digital, Activision, Bohemia Interactive, and so on, will indefinitely he going for it. Honeslty, OpenCL isn't that great. It performs much better obviously, but in the end, the visuals aren't better. You can run a high graphics benchmark (Unigine Valley for example) on OpenCL, and it'll do great with the AMD card, however, on Direct X or PhysX, it does run a few to i'd say 10 frames slower, but the end graphics output looks so much cleaner and nicer.

It's really a hit or miss, so we'll have to see. But unless someone can really manipulate the OpenCL code, or the code becomes better, I don't really see that happening in the least.

CUDA may be still around for a while but physx has been pretty much dead for years. CUDA will be replaced since people don't need to use it any more, as nvidia loses more and more shares in HPC space to intel and AMD, CUDA will go with it since OpenCL will work with all platforms.

OpenCL isn't gaming oriented, it only leverages GPU resource for general computations. OpenCL isn't even meant to be use for gaming. You seem to be confusing openCL with openGL. MS has already adopted directCompute inside DX11 for the next xbox, there is no need for any physx in any future games unless nvidia wants to shell out money to devs for the PC releases. Sony will likely use openGL implementations of the same functions since they are both using GCN. GPU physx will not be in any next generation consoles. DirectX has nothing to do with physx. You can run either without the other. The only thing about physx is nobody has use of it now that GPU compute can be driven directly from DX11, unlike when games had to stick with dx9 so gpu compute acceleration has to be done mostly through the physics engine, some of which uses phsyx.

OpenCL code runs on any AMD, Nvidia and Intel gpu. CUDA only runs on nvidia. OpenCL can do pretty much everything CUDA can do. Nobody will use CUDA if they want to reach out to bigger audiences. That is why adobe is supporting openCL now and is likely going to slowly drop support for CUDA, especially since Mac pros will be shipping with AMD GPUs.

And all that babbling about the xbox one doesn't have anything to do with this. Both console will use AMD hardware. Both will contribute to the end of GPU physx. Both will likely support openCL applications.

 

[griptwister]

'Interestingly the document also made it clears the days of the big core are numbered, leaving the FX-9590 as the final legacy of the FX line.'via VR-Zone

http://wccftech.com/amd-possibly-abandoning-entire-fx-processors-lineup/#ixzz2b9pY9tkQ

AMD only have 4 core apus lined up it seems. The fx9590 was amd last fx series for now at least. AMD are going to the intel route it seems as less cores that are better is the way to go.

eit:amd will still make piledriver server cores though.

AMD isn't going the route of Intel so much as "Intel" is going the route of AMD. They added friggin graphics to their CPUs. AMD is just refining their architecture. Also, that's only the APU line up mentioned, the guy said and knows nothing about their AM3+ platform.

 

[sarinaide]

My Kaveri/SR predictions just so I may throw it out there well in advance before any definite leaks come out;

x86 - Tweaks to Cache, IMC and subsystems along with evolutionay tweaks to the architecture will see performance per watt and performance per clock improvements. Over the A10 6800K at same clocks should see around 20-40% improvements depending on the application, around 15% general performance increase along with power improvements.

iGPU - Spectre should be around 2x faster than the HD8670D on Richalnd

I wait a 20% over Piledriver and probably a clock reduction from Richland: 4.1Ghz --> 4 GHz

Based exclusively in GPU cores, I wait about a 30% over Richland/Trinity plus a clock boost up to 900 MHz. This could add to about a 40% increase. How do you got the 2x estimation?

It seems Kaveri is clocked the same as Richland from hearsay so it will offer performance/clock increases over the PD based arch.

I believe the Spectre iGPU is around and between a HD7730 and HD7750 performance which is roughly double where Trinity is at, The GCN cores are far better at tessellation and compute than the VLIW parts so its GPGPU and General purpose(gaming) will improve significantly. At the same time OBR stated that FX9000 would be the last ever AMD traditional DT CPU which is transpiring to be true the same rumours at that time were that Kaveri would be at entry level gaming ability to be at that point the APU's needed roughly twice the performance.

 

[noob2222]

I see the bigger problem with all the lack of confirmed rumors/news is the lack of GF 28nm working parts.

look at from that point of view and you will see where the problem is.

say for example gf tells amd that 28nm yield is as follows:

2b transistors ~ 5% fully functional
1.75b - 10%
1.5b - 17%
1.25b - 25%
1b - 35%

amd cpus

[strike] sr fx 1.75b[/strike]
[strike] kaveri 6c 1.5b[/strike]
kaveri 4c 1.25b

lowering transistor count also massively increases the # of possible chips. say at 1.75b you could get ~300 chips from a full wafer at $5000. lower that to 1.25b and its ~ 500 possible chips.

which route do you take, 300 at 10% fully functional or 500 at 25% fully functional.

also as for the chip itself, which one has more sensitive area that if bad cannot be binned at all?

apus are very forgiving. fuse off gpu cores and you get a10, a8, a6 and a4.
500 possible turns into ~50% at least partially usable.

FX has too much cache area that causes it to be unusable by design so your only binned on core area flaws. 300 possibles turns into ~ 25% partially usable chips.

these figures are not accurate, just to show a possibility why certain things have changes/disappeared from the roadmap.


the lack of news from both AMD and GF to me points to GF being unable to produce the chip as well as the kaveri delay. There is a reason kabini and tamesh are at TSMC.

 

[Cazalan]

AMD isn't going the route of Intel so much as "Intel" is going the route of AMD. They added friggin graphics to their CPUs. AMD is just refining their architecture. Also, that's only the APU line up mentioned, the guy said and knows nothing about their AM3+ platform.

Intel added graphics to their CPUs about a year before AMD did. They just sucked.

 

[Cazalan]

the lack of news from both AMD and GF to me points to GF being unable to produce the chip as well as the kaveri delay. There is a reason kabini and tamesh are at TSMC.

It's likely a mix of reasons. With AMD paying to sever their contracts with GF, it hasn't put them on the best working terms. AMD doesn't have the money to throw around to grease the wheels per say. Any new process has a yield ramp and AMD probably is waiting for yields to get better while Richland stock is being depleted. AMD mentioned around 700mil in inventory last month, and that wasn't all PS4/Xbone chips.

 

[palladin9479]

@juanrga

games have ALWAYS favored cpus with l3 cache. athlon > duron ; pentium > celeron ; phenom > athlon ; fx > apu.

granted 16mb is probably a bit much, intel's normal cpus only have ~6mb and only the extremes have over 10mb. no l3 cache will have lower performance unless by some miracle the cpu core is strong enough to compensate the lack of. even if kaveri is faster than the I5, fx SR will be that much faster.

It's not the presence of L3 but total cache size respective to each thread and the speed of that cache. Intel's caching design is radically different then AMD's, this results in some interesting results. Intel use's an ultra fast L2 cache that is fairly small by today's standards (256KB) vs AMD's larger slower 1MB (2MB per core). This is why Intel use's a large L3 cache, their 256KB isn't large but they can always fall back to the L3 prior to hitting system memory. Cache L2 cache miss's aren't a big deal for the Intel design, their absolutely crippling to the AMD one. This is also why you must be EXTREMELY careful with synthetic benchmarks. They tend to use incredibly small working data sets that are easy to predict and fit into cache memory, this in turn can mask real world performance issues.

 

[noob2222]

AMD isn't going the route of Intel so much as "Intel" is going the route of AMD. They added friggin graphics to their CPUs. AMD is just refining their architecture. Also, that's only the APU line up mentioned, the guy said and knows nothing about their AM3+ platform.

Intel added graphics to their CPUs about a year before AMD did. They just sucked.

while technically true, clarkdale wasn't a true igp but a dual chip package consisting of a 32nm cpu and a 45nm gpu.

brazos apu and sandy bridge same out roughly the same time frame.

design wise, amd started the whole APU project back in 2006, intel just had the money to beat them to the "punch".

www.tomshardware.com/reviews/fusion-hsa-opencl-history,3262-4.html

 

[sarinaide]

^Go back to Skyrim