AMD ZEN dieshot and structure

[Lawliet93]

1 minute ago, patrickjp93 said:

You're correct, but the fighting over the FPU still requires the cores check with each other to see which half of it is in use by the other, hence the horrible FP performance. In terms of integer instructions, for Piledriver and beyond, they are true cores. For Bulldozer, the two halves would share a scheduler too, making them far less than true cores.

 

I ran some simulations a long time ago using QEMU and an integer linear program where the constraints were the total IPC increase (40% at the time), integrality of clock cycle latencies, and a few theoretical barriers such as no instruction can take 0 cycles and some things like integer division require more cycles for more bits and there is a lower bound. In the best case scenario (used 40 different benchmarks including the Linpack Suite), Zen would end up with 3% higher IPC than Haswell. Accounting for reality, it'll be closer to Ivy Bridge.

Well, it all depends on the clockspeeds  they have chosen, if they are going to stay on their 95W TDP limit, than they will have to reduce the clockspeeds, ergo, the IPC will drop. We don´t know at what frequencies they have used. It can very well be, that a 3,5GHz Zen has the same IPC as a 3,5GHz IB cpu, therefore, thanks to the 14nm process, they can increase the frequency to 4GHz.

[patrickjp93]

1 minute ago, Lawliet93 said:

Well, it all depends on the clockspeeds  they have chosen, if they are going to stay on their 95W TDP limit, than they will have to reduce the clockspeeds, ergo, the IPC will drop. We don´t know at what frequencies they have used. It can very well be, that a 3,5GHz Zen has the same IPC as a 3,5GHz IB cpu, therefore, thanks to the 14nm process, they can increase the frequency to 4GHz.

IPC doesn't depend on clock speed unless the fetch portion and cache are unhinged from the core clock. IPC * clock speed ~= performance.

[Lawliet93]

1 minute ago, patrickjp93 said:

IPC doesn't depend on clock speed unless the fetch portion and cache are unhinged from the core clock. IPC * clock speed ~= performance.

Well that´s what I meant, I even wrote it down, but deleted it afterwards. Thanks for the information. What i meant was, that they could achieve Haswell level of performance, even with IB IPC, if they increase the clock speeds accordingly. 

[patrickjp93]

1 minute ago, Lawliet93 said:

Well that´s what I meant, I even wrote it down, but deleted it afterwards. Thanks for the information. What i meant was, that they could achieve Haswell level of performance, even with IB IPC, if they increase the clock speeds accordingly. 

If you think AMD is going to pull off significantly higher clock speeds than Broadwell-E is promising, I have a bridge to sell you. If the transistor design is for low power, clock ceilings fall. If the transistors are designed for higher power, efficiency goes away and that TDP becomes a fairytale.

[Eroda]

2 hours ago, Master Disaster said:

40% over Excavator is just about the most ambiguous term I've ever heard, which Excavator? Toronto or Carizzo? Which SKU exactly? I reckon I could prolly find a couple of benchmarks from a low end and high end SKU with a huge difference between them.

 

They might as well just say Zen will be 40% faster than the colour purple, that would be more useful.

well if they can make it as fast as Red then they will win because everyone knows red is the fastest color

[Lawliet93]

12 minutes ago, patrickjp93 said:

If you think AMD is going to pull off significantly higher clock speeds than Broadwell-E is promising, I have a bridge to sell you. If the transistor design is for low power, clock ceilings fall. If the transistors are designed for higher power, efficiency goes away and that TDP becomes a fairytale.

Of course, but they are aiming at the performance and enthusiast market, so they can clock the chip down a little bit, reduce the voltage, get that 95W TDP and advertise it as a good overclocker, but what do I know? I am just a medschool student with deep depressions and a high risk of suicide You are the genius around here, you know better.

[patrickjp93]

5 minutes ago, Lawliet93 said:

Of course, but they are aiming at the performance and enthusiast market, so they can clock the chip down a little bit, reduce the voltage, get that 95W TDP and advertise it as a good overclocker, but what do I know? I am just a medschool student with deep depressions and a high risk of suicide You are the genius around here, you know better.

Perhaps, but the GHz war ended long ago in CPUs for a reason. Silicon has a hard time getting to 5GHz at all, especially in modern, efficient FinFET designs. Even breaking 4 for many-core chips is incredibly tough without massively blowing up the TDP.

 

I just wish the hyping would stop. Zen will be a big improvement for AMD. Where it finally lands is something we should just wait and see about.

[Lawliet93]

1 minute ago, patrickjp93 said:

Perhaps, but the GHz war ended long ago in CPUs for a reason. Silicon has a hard time getting to 5GHz at all, especially in modern, efficient FinFET designs. Even breaking 4 for many-core chips is incredibly tough without massively blowing up the TDP.

 

I just wish the hyping would stop. Zen will be a big improvement for AMD. Where it finally lands is something we should just wait and see about.

That´s the spirit people think, that a 4c/4t Zen will wipe the floor with the 5960x. I have not seen so much hype since the bulldozer days. People should lover their expectations. AMD is a small company compared to intel and there are no big leaps to be made on silicon. We have hit the limit,

[Centurius]

5 hours ago, TheRandomness said:

Price something like this low enough and people everywhere will want one. I mean, an octocore for £230~ is tempting, no?

Sure, but selling something at a loss sure doesn't for AMD.

[handymanshandle]

3 hours ago, Trixanity said:

I said top SKU. The top SKU will have 8 cores/16 threads. It is two-way SMT after all - just like Intel. The hyper threading moniker is just a marketing term. But again, they will probably have to lower it to compensate for performance discrepancies. So let's just say $899 or $799 if we're lucky. If it was $499 for such a SKU, I'll assume something has gone horribly wrong.

See I wasn't sure what you meant, I thought you were saying the top tier 8 core 8 thread would be $1000...

But if it's 8 cores, 16 threads then sure it makes sense to price it closer to the price bracket of the 5960X, as long as it performs about the same or better than it.

[Energycore]

I'll take an overclockable, Hyper-Threaded quad-core for $200-250. Something that'll compete with non-OC skylake i7s but will cost less.

[othertomperson]

7 hours ago, Lawliet93 said:

AND how much money should AMD make on this 230 pound CPU? They need at least a margin of 30%, 50% would be even better. What good is it, if AMD can´t make any money on a well performing part?

£230 is £35 more than an i5 6600K, and £24 less than an i7 6700. If it is going to cost the same as Skylake (or more as you suggest) it is going to have to perform at least as well. If it doesn't, then it had better be cheap.

[Energycore]

Let's be honest, AMD's prices are going to be dictated by relative performance. They have no reason to charge less than just a bit below Intel's similar performing chip.

[Lawliet93]

50 minutes ago, othertomperson said:

£230 is £35 more than an i5 6600K, and £24 less than an i7 6700. If it is going to cost the same as Skylake (or more as you suggest) it is going to have to perform at least as well. If it doesn't, then it had better be cheap.

We were talking about the 8c/16t CPU and if it´s going to perform like a i5 6600k than they can stick it up their... you know.

[othertomperson]

2 minutes ago, Lawliet93 said:

We were talking about the 8c/16t CPU and if it´s going to perform like a i5 6600k than they can stick it up their... you know.

Fair enough.

[Dabombinable]

Those "die shots" are old and look to me like those of the FX 8*** http://www.overclockers.com/amd-fx-8350-piledriver-cpu-review/

So I wouldn't take much stock in them.

[Revinval]

So far I'm pretty sure we will see the normal price ranges $250ish for 6ish cores to compete with the I5 and a sub $150 for the I3 but I really don't think we are getting a $700+ cpu from AMD outside of server stuff for quite a while. The top end will probably be around $500-$700. Don't forget the 1k cpu is a massively inflated cost for just nonserver cpu. 

[Sharkyx1]

9 hours ago, Dabombinable said:

Those "die shots" are old and look to me like those of the FX 8*** http://www.overclockers.com/amd-fx-8350-piledriver-cpu-review/

So I wouldn't take much stock in them.

The L3 on BD is in between the 2 sets of modules facing toward each other

 

 

Zen has the L3 in between the cores facing each other