AMD Zen Initially only Coming With 8-Core Dies-8 Core and 6 Core Zen CPUs to be the First to Hit the Market

[79wjd]

10 minutes ago, Notional said:

They literally say that the i3 is at 100% load.

100% load doesn't necessarily means that's where the problem lies. Techspot doesn't show a huge difference between a 6700k and an i3 when paired with a 980ti.

[patrickjp93]

@DocSwag   @Notional  @AluminiumTech @420istoday  Sorry for the format. LTT is bugging out on Google Chrome and not always submitting and then locking down.

 

[Lauen]

Just now, patrickjp93 said:

For Opterons, AMD was accurate. For FX and Athlon, they went with consumer-typical workloads.

Fair enough. 

[Unhelpful]

1 minute ago, patrickjp93 said:

@DocSwag   @Notional  @AluminiumTech @420istoday  Sorry for the format. LTT is bugging out on Google Chrome and not always submitting and then locking down.

Calling @colonel_mortis? Or no?

[patrickjp93]

16 minutes ago, Notional said:

They literally say that the i3 is at 100% load.

Do you know how easy it is to induce 100% load on a core?

 

Spoiler

int main() {
	while(true) {}
	return 0;
}

If the CPU is just waiting for a frame to finish before submitting the data for the next and then resuming the normal pipeline and is using polling (spin locks) instead of a single core waking the others up, then you can have 100% load and not be doing any real work and have plenty of cycles left to spare.

 

For all cores, using OpenMP for brevity, obviously ignoring the complexities of working on multiple frames at once, but you get the idea.

int main() {
	bool FND = false;  //Frame Not Done
	#pragma omp parallel
    {
		while(true) { //Game loop for everyone
			//SPMD asynchronous environment
			#pragma omp master
			{
				//submit new frame data for crunching
				FND = true;
			}
			#pragma omp barrier //no threads proceed until master's task is done

			while(FND) { //Everyone waits for the frame processing to finish before proceeding with the next
				#pragma omp master
				{
					//wait until GPU responds.
					FND = false;
				}
			}
		}
	}
}

 

 

[Notional]

21 minutes ago, Starelementpoke said:

This is assuming a work load that takes advantage of using all cores.

That goes without saying, but you can only compare CPU's at their full performance. Otherwise a 4790k would beat the crap out of a 5960x due to the much higher clock speed.

[ZetZet]

36 minutes ago, Notional said:

They literally say that the i3 is at 100% load.

the point is it works for everyone else and the only thing that changes is the gpu

[Notional]

16 minutes ago, patrickjp93 said:

@DocSwag   @Notional  @AluminiumTech @420istoday  Sorry for the format. LTT is bugging out on Google Chrome and not always submitting and then locking down.

 

Yeah the forum's quote function is a bit buggy for the time being. Idk why.

 

Yeah please link those tables. I have no doubt Intel's is better though, my point is that the small difference might not be directly transferable to higher performance.

However if we compare Intel's high performance SRAM at 0.0706 µm2 to Samsung's high performance SRAM at 0.080 µm2, we are not at all at 22% difference, but rather ~12,5%.

 

As for HDL, we don't know how dense they made ZEN or how it affects clock rate on 14nm FF LPP. It might not affect all that much on the smaller node or they might not use HDL at all (although I doubt that). Carrizo has the north bridge built into the APU, so using HDL there makes perfect sense due to the lower clock speed needed.

Either way, it's too early to conclude anything about ZEN performance. As such we cannot conclude anything on price either.

 

Like I mentioned earlier, AMD doesn't need ZEN to beat Intel's KOTH products to be competitive or a success for that matter. They have to be good enough to be a viable option on both server and consumer systems.

[suicidalfranco]

if the 8 core isn't too expensive i might go with AMD and jump back on the Windows for games Virtual Machine

[Doobeedoo]

I'm looking forward to 8-core for sure. Good that they're focusing at high end. Now just for benchmarks and final specs.

[DocSwag]

3 hours ago, Notional said:

They literally say that the i3 is at 100% load.

If you have 4x Intel cores running 15% higher IPC than ZEN, then all else equal, the Intel should be 15% faster. If the ZEN has 6 cores instead of 4, ZEN would be 50% faster than a 4 core ZEN. That's a 35% performance increase on the ZEN in this example. If the ZEN has SMT and the i5 not, then that's an additional -5-30% performance increase (yes SMT can screw up things too).

You forgot about clock speeds. 

[DocSwag]

All in all, honestly we currently do not have the resources to make good assumptions about Zen. The only info we know for sure is that Zen will have greater than 40% more IPC than Excavator and will use DDR4. That's it. Everything else is rumor.

 

We can make general guesses about Zen performance, but other than that we can't make any assumptions. We can pinpoint it's IPC to somewhere between Ivy Bridge and Skylake (And if more than 40% IPC means a lot more than 40% IPC gain perhaps even equal or greater IPC than Kaby Lake), but other than that we can't make good guesses. Chances are it will go up to 8 cores, but that doesn't tell us anything about cache or other stuff, and doesn't help that much. We know it has SMT, but we don't know how AMD's implementation works and how much of a performance gain it will give. We know next to nothing about clock speeds. Clock speeds could be around 3 ghz for four core CPUs (though quite unlikely) or maybe over 4.5 ghz on six-cores.

 

Zen may be confirmed as is also one of the most hotly debated topics among CPUs today, but we don't know enough to make assumptions. Therefore all this topic about how Zen is gonna be between Ivy Bridge and Haswell or below Ivy Bridge or just below Kaby Lake or whatever is all BS. You can't argue about something that you don't know anything about. It's just stupid. So just stop arguing about how good Zen is gonna be and what Intel processor it's gonna be like. You can't make any assumptions, so you can't argue.

[Notional]

28 minutes ago, DocSwag said:

You forgot about clock speeds. 

IPC is irrelevant of clock speeds, so no I did not forget it. It's just not relevant in that comparison. It's also why I wrote "all else equal", as DRAM and all sorts of other things can influence performance.

[226477_1454181668]

6 hours ago, ZetZet said:

If the 6 core is 300+ usd rip in pieces AMD.

 

It should be 200, direct competition for the i5. (if it even has the performance of an i5)

you dont know how it will perform, so that claim is just stupid. 

[dalekphalm]

I gotta say, people saying the 6 core Zen should be $200? You're basically saying that an Intel i7-5930 should be $200. Do you see the problem with that statement? It makes no sense.

 

You're assuming AMD's 6-core Zen will be another Bulldozer, and will run quite badly for the number of cores and clock rate.

 

We can't say for certain where it should be priced, since no one has seen any verified independent benchmarks yet. But I expect the 6 core to be at least $300, if not even more then that.

 

Remember, for the Low End, AMD also has Bristol Ridge launching at the same time with their APU's. These are excavator based, so not as good as Zen, but certainly they're not bad for entry-level gaming, and should be competitive with low end i3 setups.

 

I imagine that the 4-core Zen variants will probably be coming within a few months-to-half-a-year after Zen launches. They probably just want to get some high performance hot sellers out there to gain interest first.

 

Also, we don't know what OEM's are getting. They might not be releasing any 4-core Zen units simply because the OEM channels have already earmarked all of the 4-core supply at launch.

[shdowhunt60]

It could be that they price it like Thuban? The top end Thuban chips was $400. It really depends on how Zen performs in relation to whatever Intel has, and how much AMD is willing to undercut Intel.

[ZetZet]

21 minutes ago, dalekphalm said:

I gotta say, people saying the 6 core Zen should be $200? You're basically saying that an Intel i7-5930 should be $200. Do you see the problem with that statement? It makes no sense.

yeah it's going to match an extreme intel processor with 95w tdp...

[DocSwag]

1 minute ago, ZetZet said:

yeah it's going to match an extreme intel processor with 95w tdp...

There's this thing called overclocking. It's pretty cool.

[ZetZet]

17 minutes ago, DocSwag said:

There's this thing called overclocking. It's pretty cool.

Yeah, I know, Intel has figured that one out too, that's why they call them extreme

[DocSwag]

1 minute ago, ZetZet said:

Yeah, I know, Intel has figured that one out too, that's why they call them extreme

My point was that the 95 watt TDP isn't a limiting factor. If, say, the Zen CPU can OC as far as the Intel one, the 95 watt TDP only means it's slower out of the box, but is just as fast (in clock speeds) once OCed. The TDP isn't a limiting factor at all.

[ZetZet]

27 minutes ago, DocSwag said:

My point was that the 95 watt TDP isn't a limiting factor. If, say, the Zen CPU can OC as far as the Intel one, the 95 watt TDP only means it's slower out of the box, but is just as fast (in clock speeds) once OCed. The TDP isn't a limiting factor at all.

yes, it still won't match it since you can overclock 5930k...

[Mr.Meerkat]

This is such a funny thread to read  kinda surprised how an admin/mod hasn't locked it yet for all of these butthurt people but I agree with @DocSwag, why can't AMD make a fast but efficient chip? @ZetZet just because it has a TDP of 95w =/= it's going to have a low clock speed and more importantly, TDP=/=power consumption...a CPU can have a TDP of 95w but why can't it be sucking 150w from the PSU? TDP is really only there to be used as a guide on how much heat it's going to be emitted by the CPU underload and not how much power it's going to require/consume. 

[ZetZet]

4 minutes ago, Mr.Meerkat said:

This is such a funny thread to read  kinda surprised how an admin/mod hasn't locked it yet for all of these butthurt people but I agree with @DocSwag, why can't AMD make a fast but efficient chip? @ZetZet just because it has a TDP of 95w =/= it's going to have a low clock speed and more importantly, TDP=/=power consumption...a CPU can have a TDP of 95w but why can't it be sucking 150w from the PSU? TDP is really only there to be used as a guide on how much heat it's going to be emitted by the CPU underload and not how much power it's going to require/consume. 

tdp is a measure of heat, yes. but since in processors pretty much all energy is expelled outwards because you aren't turning an engine, heat is a good measurement of power draw, as pretty much all tests throughout history have shown. 

[Mr.Meerkat]

3 minutes ago, ZetZet said:

tdp is a measure of heat, yes. but since in processors pretty much all energy is expelled outwards because you aren't turning an engine, heat is a good measurement of power draw, as pretty much all tests throughout history have shown. 

That's true but maybe AMD has broken away from the usual trend of TDP=similar to what the actual power consumption is. I don't see why their 8-core zen CPU can't be clocked at 4GHz while having a 95watt TDP, just because you're/we're not used to seeing something like that doesn't mean it's not possible. Why can't their 14nm CPUs be extremely efficient, so efficient that it's more efficient than skylake CPUs? 

[ZetZet]

1 minute ago, Mr.Meerkat said:

That's true but maybe AMD has broken away from the usual trend of TDP=similar to what the actual power consumption is.

that doesn't make any sense.

 

 

you mean AMD could defeat the laws of physics to compete with Intel? WHY HAVEN'T THEY DONE THAT BEFORE