AMD Geekbenchs for engineering examples???

[ivan134]

31 minutes ago, Dabombinable said:

Look, I did mention that I'm looking forward to Zen APU (if they happen) because they'll offer far better price/performance than Intel's equivalent, and because I can't go past Windows 7 with my laptop (Phenom II N970+MR HD5650, its PowerXpress config doesn't work in newer versions of Windows correctly). However if Zen APU don't come out, if, and the only option is a Bristolridge, aka CMT, APU, then I won't be getting a laptop with one.

Who told you there isnt a Zen APU coming? There is no if they come out. Its only a matter of then, and they're most likely the same time frame as the CPUs. You've been rambling on and on about a nonexistent issue.

[Dabombinable]

56 minutes ago, ivan134 said:

Who told you there isnt a Zen APU coming? There is no if they come out. Its only a matter of then, and they're most likely the same time frame as the CPUs. You've been rambling on and on about a nonexistent issue.

You obviously are "if" blind.

[Lotte]

53 minutes ago, Dabombinable said:

You obviously are "if" blind.

It's not even an 'if', it's more like, 'when'. It's going to happen, but until then, they have to use the CMT style EX core.

[Prysin]

11 hours ago, Dabombinable said:

Still, they are clinging to CMT, and if they only use it their AM4 APU.....and don't make Zen APU....

here is four facts:

1 - I own a Athlon x4 845 (Carrizo/Excavator) CPU and a A10 7870k Kaveri APU.

The Athlon x4 845 at stock scores the same as a 7870k does,  at 4.3GHz.... That is 500MHz above the max turbo of the 845....

When you overclock the 845 to 4077MHz using baseclock, ive hit slightly lower scores in the CPU-Z benchmark then i did with my FX 8320 at 4.7 GHz....

However the X4 845 is hurt by its lack of L3 and reduced L2 cache. It may be using it's cache much much much more efficiently, but sometimes you just NEED that space, which excavator simply do not have. Cinebench illustrates this, as it favors fast and large caches. This is why a FX 6300 will scale so damn well, due to it having 8MB of L3, meaning it has 1.33MB of L3 per core, while the FX 8 and FX 9 series only has 1MB per core..

So in Cache dependent tasks, the X4 845 is dragging its ass, simply due to lack of Cache.

 

2 - The Athlon x4 845 running t 3.5-3.8GHz does NOT bottleneck a Sapphire Radeon HD 7950 in any way.

The strongest iGPU for desktop PCs (ignoring Consoles) in existence atm is the Intel Broadwell HD5000 series Iris Pro. The Broadwell "APU" is slightly weaker then a GTX 750 (stock).

The performance disparity between a GTX 750 and a Radeon HD 7950 is closer to 75% (GTX 750 is 15% slower then 750Ti). Which you can see here: http://anandtech.com/bench/product/1033?vs=1130

The likelihood of AMD or even Intel being able to make a iGPU faster then a 750Ti with todays technology is rather slim.

Thus the likelihood of a Excavator part bottlenecking anything is even slimmer.

 

3 - The Single core performance between Bulldozer and Piledriver is 10%. From piledriver to Steamroller it is +7-10%, from Steamroller to Excavator it is +11%ish.

The single core performance between Bulldozer and Intel's Sandy Bridge is 48-50%, at nearly equal clocks that is.

The single core performance between Bulldozer and Excavator, factoring in all the incremental increases in performance, is around 30-34%. assuming equal clock speeds.

Infact, Excavator is clocked slightly lower then piledriver and steamroller, indicating even higher IPC, given that it keeps beating the aforementioned versions.

 

4 - Memory speed sucks on FM2+, it doesnt really scale at all, infact, it is so atrocious that running dual channel 2400MHz on a FM2+ CPU gives about equal memory bandwidth as running 1866MHz on Intel, or 2000MHz on AM3+ CPUs... Yes it is THAT bad. If you think Kaveri APUs were bandwidth starved, they are. By the chipset/socket design, not just the memory speed.

Infact, just by going with DDR4 RAM alone, AMDs Tonga based iGPU found in the FX8800p APU should be way faster then anything Intel has, assuming memory bandwidth isnt fucked in the ass by a lousy chipset/board design.

Memory speed helps weaker CPUs, infact, if AMDs FM2+ platform could fully utilize 2400MHz dual channel DDR3, the CPUs using said platform would probably perform even better.

 

Conclusion time:

The Bristol Ridge APU/CPUs will perform around the same as a low clocked Sandy Bridge i3 CPU (think 2.8-3GHz) in Single Core, and slightly faster then say a Sandy Bridge i5 2300 in multi thread.

Basically, you are buying a locked Sandy Bridge i5 with updated instruction sets, possibly no unlocked multiplier (although they may have found a way to make Excavator OC).

 

 

Conclusion:

 

[Dabombinable]

15 minutes ago, Prysin said:

here is four facts:

1 - I own a Athlon x4 845 (Carrizo/Excavator) CPU and a A10 7870k Kaveri APU.

The Athlon x4 845 at stock scores the same as a 7870k does,  at 4.3GHz.... That is 500MHz above the max turbo of the 845....

When you overclock the 845 to 4077MHz using baseclock, ive hit slightly lower scores in the CPU-Z benchmark then i did with my FX 8320 at 4.7 GHz....

However the X4 845 is hurt by its lack of L3 and reduced L2 cache. It may be using it's cache much much much more efficiently, but sometimes you just NEED that space, which excavator simply do not have. Cinebench illustrates this, as it favors fast and large caches. This is why a FX 6300 will scale so damn well, due to it having 8MB of L3, meaning it has 1.33MB of L3 per core, while the FX 8 and FX 9 series only has 1MB per core..

So in Cache dependent tasks, the X4 845 is dragging its ass, simply due to lack of Cache.

 

2 - The Athlon x4 845 running t 3.5-3.8GHz does NOT bottleneck a Sapphire Radeon HD 7950 in any way.

The strongest iGPU for desktop PCs (ignoring Consoles) in existence atm is the Intel Broadwell HD5000 series Iris Pro. The Broadwell "APU" is slightly weaker then a GTX 750 (stock).

The performance disparity between a GTX 750 and a Radeon HD 7950 is closer to 75% (GTX 750 is 15% slower then 750Ti). Which you can see here: http://anandtech.com/bench/product/1033?vs=1130

The likelihood of AMD or even Intel being able to make a iGPU faster then a 750Ti with todays technology is rather slim.

Thus the likelihood of a Excavator part bottlenecking anything is even slimmer.

 

3 - The Single core performance between Bulldozer and Piledriver is 10%. From piledriver to Steamroller it is +7-10%, from Steamroller to Excavator it is +11%ish.

The single core performance between Bulldozer and Intel's Sandy Bridge is 48-50%, at nearly equal clocks that is.

The single core performance between Bulldozer and Excavator, factoring in all the incremental increases in performance, is around 30-34%. assuming equal clock speeds.

Infact, Excavator is clocked slightly lower then piledriver and steamroller, indicating even higher IPC, given that it keeps beating the aforementioned versions.

 

4 - Memory speed sucks on FM2+, it doesnt really scale at all, infact, it is so atrocious that running dual channel 2400MHz on a FM2+ CPU gives about equal memory bandwidth as running 1866MHz on Intel, or 2000MHz on AM3+ CPUs... Yes it is THAT bad. If you think Kaveri APUs were bandwidth starved, they are. By the chipset/socket design, not just the memory speed.

Infact, just by going with DDR4 RAM alone, AMDs Tonga based iGPU found in the FX8800p APU should be way faster then anything Intel has, assuming memory bandwidth isnt fucked in the ass by a lousy chipset/board design.

Memory speed helps weaker CPUs, infact, if AMDs FM2+ platform could fully utilize 2400MHz dual channel DDR3, the CPUs using said platform would probably perform even better.

 

Conclusion time:

The Bristol Ridge APU/CPUs will perform around the same as a low clocked Sandy Bridge i3 CPU (think 2.8-3GHz) in Single Core, and slightly faster then say a Sandy Bridge i5 2300 in multi thread.

Basically, you are buying a locked Sandy Bridge i5 with updated instruction sets, possibly no unlocked multiplier (although they may have found a way to make Excavator OC).

 

 

Conclusion:

 

I could buy a Sandybridge version of my laptop so yeah...I'd go Zen APU or no APU at all.

[ivan134]

3 hours ago, Dabombinable said:

You obviously are "if" blind.

No, there is no if. It doesnt exist. You made up the if in your head and started ranting and getting mad at AMD because of your delusions.

[Dabombinable]

Just now, ivan134 said:

No, there is no if. It doesnt exist. You made up the if in your head and started ranting and getting mad at AMD because of your delusions.

Your still not reading my posts fully are you?

[ivan134]

1 hour ago, Dabombinable said:

Your still not reading my posts fully are you?

No, I read and understood perfectly. You are insinuating that Zen will have a CMT design, or a Zen APU is not happening until 2020, or both, based on the release on this Excavator APU. Calling this insinuation idiotic is an understatement. It isn't based on reality. Unless sonething unforseen happens that causes them to cancel, there is absolutely jack shit to indicate that a Zen APU isn't releasing in 2017 or that it wont follow the design that AMD showed a while ago.

Edited April 12, 2016 by ivan134

[Prysin]

23 minutes ago, Dabombinable said:

I could buy a Sandybridge version of my laptop so yeah...I'd go Zen APU or no APU at all.

my x4 845 costs 70 USD.... That is cheaper then even used i5 2300s on ebay...

[DocSwag]

I think a small miss understanding is causing a giant flame war here...

@Dabombinable made a comment about hoping AMD doesn't stay with CMT. Everyone jumped on him.

 

Ok, first of all, his point was:

If AMD doesn't make a Zen APU, I'm out of here.

 

What everyone else though he meant:

Everyone else thinks he means: Zen is still CMT or AMD won't make a Zen APU

 

For the record, Zen is not CMT. Done.

AMD is most definitely making a Zen APU. It'd be stupid not to. Their 28nm APUs are nearly as good as Intel's 14nm iGPU with integrated DRAM. Bring them down to 14nm, and maybe even add some HBM, and I can easily see AMD having APUs two times if not more powerful than Intels offerings. As well, it's been pretty much confirmed at this point that AMD is bringing out CPUs first in Q4 2016 and APUs will follow in Q2 2017. Whether it's based on Vega or Polaris we don't know. I'd think it might be Vega, though, since the APUs will come out after Vega does.

 

There. Now shut up and stop calling each other idiots for something they didn't even say.

[Dabombinable]

27 minutes ago, Prysin said:

my x4 845 costs 70 USD.... That is cheaper then even used i5 2300s on ebay...

Talking about laptops. APU are next to pointless in desktops, and as mentioned before, I could make my laptop last until around 2020 so I have plenty of time to  wait and see what AMD comes up with. So far all we have that is realistically confirmed is Bristolridge APU. Zen APU are all speculation and rumours, which is why I will continue to use if when referring to them until such time as we have some solid evidence as to what they might be like. Now as for the possibilities in regards to them, 16nm Zen+16nm Polaris (or GCN 1.3) GPU+HBM L4 cache/vRAM will negate the current problems in regards to memory channels, and possibly the quality of the IMC. And I can say right now, Bristolridge really doesn't add enough to be a compelling buy, especially considering the IPC is around that of Sandybridge, which is from 2012 just like Vishera.

[Mr.Meerkat]

41 minutes ago, Prysin said:

my x4 845 costs 70 USD.... That is cheaper then even used i5 2300s on ebay...

(Just as a heads up, I agree with what you're saying here but...)

 

I always forget how used prices have gone up, I got my used 2300 11 months(ish) ago at $70 (£50) on ebay (UK ebay prices have gone up as well)...

 

[Dabombinable]

3 minutes ago, DocSwag said:

I think a small miss understanding is causing a giant flame war here...

@Dabombinable made a comment about hoping AMD doesn't stay with CMT. Everyone jumped on him.

 

Ok, first of all, his point was:

If AMD doesn't make a Zen APU, I'm out of here.

 

What everyone else though he meant:

Everyone else thinks he means: Zen is still CMT or AMD won't make a Zen APU

 

For the record, Zen is not CMT. Done.

AMD is most definitely making a Zen APU. It'd be stupid not to. Their 28nm APUs are nearly as good as Intel's 14nm iGPU with integrated DRAM. Bring them down to 14nm, and maybe even add some HBM, and I can easily see AMD having APUs two times if not more powerful than Intels offerings. As well, it's been pretty much confirmed at this point that AMD is bringing out CPUs first in Q4 2016 and APUs will follow in Q2 2017. Whether it's based on Vega or Polaris we don't know. I'd think it might be Vega, though, since the APUs will come out after Vega does.

 

There. Now shut up and stop calling each other idiots for something they didn't even say.

All I can say is, I really want a Zen based APU as it will:

1. Give me a mobile APU that should give similar single threaded performance to my i5 4440, which is still in my books at least quite good
2. Give me far better graphics than any current APU/iGPU solution (that's a given with the die shrink)
3. Consume far less power than the Phenom II N970+MR HD4250+MR HD5650 combined while delivering better performance (current APU already manage that, but I will not take any steps backwards when upgrading/replacing my desktops computers and laptops-Bristolridge is a step back and sideways)
4. Be inside a laptop that is better in every way to my Pavilion DV6 3010AX (which has been badly treated by the previous owner, and is working despite minor fire damage on the motherboard)

And for the record, I did try to redirect those flaming to go back and read all of my comments to get the context, but they didn't.

[DocSwag]

8 minutes ago, Dabombinable said:

Talking about laptops. APU are next to pointless in desktops, and as mentioned before, I could make my laptop last until around 2020 so I have plenty of time to  wait and see what AMD comes up with. So far all we have that is realistically confirmed is Bristolridge APU. Zen APU are all speculation and rumours, which is why I will continue to use if when referring to them until such time as we have some solid evidence as to what they might be like. Now as for the possibilities in regards to them, 16nm Zen+16nm Polaris (or GCN 1.3) GPU+HBM L4 cache/vRAM will negate the current problems in regards to memory channels, and possibly the quality of the IMC. And I can say right now, Bristolridge really doesn't add enough to be a compelling buy, especially considering the IPC is around that of Sandybridge, which is from 2012 just like Vishera.

Just saying, Zen and Polaris are on 14nm

[Dabombinable]

4 minutes ago, DocSwag said:

Just saying, Zen and Polaris are on 14nm

I thought it was a 16nm process for the GPU & APU, and 14nm for the plain CPU.

[ivan134]

17 minutes ago, DocSwag said:

I think a small miss understanding is causing a giant flame war here...

@Dabombinable made a comment about hoping AMD doesn't stay with CMT. Everyone jumped on him.

 

Ok, first of all, his point was:

If AMD doesn't make a Zen APU, I'm out of here.

 

What everyone else though he meant:

Everyone else thinks he means: Zen is still CMT or AMD won't make a Zen APU

 

For the record, Zen is not CMT. Done.

AMD is most definitely making a Zen APU. It'd be stupid not to. Their 28nm APUs are nearly as good as Intel's 14nm iGPU with integrated DRAM. Bring them down to 14nm, and maybe even add some HBM, and I can easily see AMD having APUs two times if not more powerful than Intels offerings. As well, it's been pretty much confirmed at this point that AMD is bringing out CPUs first in Q4 2016 and APUs will follow in Q2 2017. Whether it's based on Vega or Polaris we don't know. I'd think it might be Vega, though, since the APUs will come out after Vega does.

 

There. Now shut up and stop calling each other idiots for something they didn't even say.

No, i understood him clearly. What exactly is the worry that AMD isnt making a Zen APU based on? That's like worrying that Pascal, Polaris or Kaby Lake isn't  happening. It wasnt even just worrying about it. He started ranting and railing against AMD for it. So you should shut up and don't insult other people's intelligence.

[Dabombinable]

Just now, ivan134 said:

No, i understood hom clearly. What exactly is the worry that AMD isnt making a Zen APU based on? That's like worrying that Pascal, Polaris or Kaby Lake isn't  happening. It wasnt even just worrying about it. He started ranting and railing against AMD for it. So you should shut up and don't insult other people's intelligence.

I never said they weren't. Since there is no solid evidence in regards to Zen APU, its better to use if instead of speaking as if I were certain. I was not railing against AMD at all, and you as usual don't seem to be taking anything that I say in its proper context despite having said context being explained to you. Oh and I have a right to criticize a companies decisions+products just like you have a right to criticize me. Deal with it.

[ivan134]

2 minutes ago, Dabombinable said:

I never said they weren't. Since there is no solid evidence in regards to Zen APU, its better to use if instead of speaking as if I were certain. I was not railing against AMD at all, and you as usual don't seem to be taking anything that I say in its proper context despite having said context being explained to you. Oh and I have a right to criticize a companies decisions+products just like you have a right to criticize me. Deal with it.

Serioulsy? Is there solid evidence of the gtx 1070 and 1080? Kaby Lake? Vega? Navi? Zen CPUs? Volta? Are you worried about those too? Yea I definitely imagined you rambling about CMT this, CMt that, and not upgrading your laptop till 2020 if a product that is supposed to come out next year doesnt come out till then.

[DocSwag]

15 minutes ago, Dabombinable said:

I thought it was a 16nm process for the GPU & APU, and 14nm for the plain CPU.

Zen is 14nm, but Polaris is too. AMD confirmed that already.

[Dabombinable]

12 minutes ago, ivan134 said:

Serioulsy? Is there solid evidence of the gtx 1070 and 1080? Kaby Lake? Vega? Navi? Zen CPUs? Volta? Are you worried about those too? Yea I definitely imagined you rambling about CMT this, CMt that, and not upgrading your laptop till 2020 if a product that is supposed to come out next year doesnt come out till then.

Not in the slightest. I do not need to upgrade my desktop for a very long time so I don't care about "the gtx 1070 and 1080? Kaby Lake? Vega? Navi?", and APU are the best choice for a laptop that has good battery life+performance, which is why I would never go with Bristolridge and instead would be content to wait for Zen APU (if they come out). As it is Bristolridge appears to be about as good as Sandybridge (albeit with some tasks forcing 1 ALU to wait for the other to finish, effectively cutting the working threads in half), while Zen is projected to be close to Haswell in performance with the same modern feature set of Bristolridge+Skylake, while having true processing cores based around SMT, and thus none of CMT's limitations. And what's wrong with waiting until AMD releases APU that don't use CMT (if they do)? I am more worried about Zen+Polaris than any other products from Intel and Nvidia as they at the moment appear to actually be AMD taking several steps towards competing properly again.

[ivan134]

2 minutes ago, Dabombinable said:

Not in the slightest. I do not need to upgrade my desktop for a very long time, and APU are the best choice for a laptop that has good battery life+performance, which is why I would never go with Bristolridge and instead would be content to wait for Zen APU (if they come out). As it is Bristolridge appears to be about as good as Sandybridge (albeit with some tasks forcing 1 ALU to wait for the other to finish, effectively cutting the working threads in half), while Zen is projected to be close to Haswell in performance with the same modern feature set of Bristolridge+Skylake, while having true processing cores based around SMT, and thus none of CMT's limitations. And what's wrong with waiting until AMD releases APU that don't use CMT (if they do)?

No one said you should buy this. AMD has already said Zen (includung APUs, I'm still confused as to what makes you think we're only getting CPUs) is releasing in 2017 so why are you expecting it in 2016?

[DocSwag]

OMG JUST SHUT UP YOU GUYS ARE ACTING LIKE 5 YEAR OLDS.

Can we just agree to disagree? Please? This is driving me crazy but I don't want to leave because there's still some cool stuff here.

[Prysin]

2 hours ago, Mr.Meerkat said:

(Just as a heads up, I agree with what you're saying here but...)

 

I always forget how used prices have gone up, I got my used 2300 11 months(ish) ago at $70 (£50) on ebay (UK ebay prices have gone up as well)...

 

that is British Pounds.

Lol, rekt. Same price even today, but the 2300 is slightly slower due to lack of newer instructions (the X4 845 is about 5-10% faster at most) .

 

 

[Mr.Meerkat]

Just now, Prysin said:

that is British Pounds.

British Pound to US Dollar

50.01 GBP = 71.204565 USD

 

Lol, rekt. Same price, but the 2300 is slightly slower due to lack of newer instructions (the X4 845 is about 5-10% faster at most) .

Fine, it's $71.204565... to be more precise but just saying...you could get sandy bridge i5's last year for the same price of a x4 845 now  

[Prysin]

2 hours ago, ivan134 said:

No one said you should buy this. AMD has already said Zen (includung APUs, I'm still confused as to what makes you think we're only getting CPUs) is releasing in 2017 so why are you expecting it in 2016?

Actually. You are slightly wrong.

 

Bristol Ridge (Excavator APU) and the AM4 socket/platform is launching around summer

ZEN FX (or whatever they call it) is launching around end of 2016.

ZEN APUs are not destined before Q1 or Q2 2017.

 

AMD has on several occasions said that they are pushing for "high end desktop CPUs" first. probably in an effort to get the ball rolling as soon as possible.

AMD also knows that with DDR4, their Tonga based iGPU found in Bristol Ridge alongside a pretty decent 2 module 4 thread Excavator CPU is more then capable of securing the mobile and iGPU market.