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

[DocSwag]

4 minutes ago, patrickjp93 said:

Having dedicated FPUs and 256-bit vector units per core is not going to come cheap electrically or thermally, and AMD has said the TDP will be 95W.

 

Anandtech proved AMD wasn't going by torture tests, but no one actually knows the metric. It does line up rather well with heavy gaming workloads though.

I realize that, but I'd imagine Jim Keller knows what he's saying and so AMD has probably made optimization elsewhere. Plus the move to 14nm FinFET is gonna help a lot.

[Notional]

7 minutes ago, patrickjp93 said:

 

The difference is still 15% on the whole, 22% if just counting SRAM. At 10nm Samsung gets a 5% density boost over Intel's 14nm.

In what SRAM config? Their high density, low voltage or high performance? Because they are all different in size. TSMC and Samsung also has different SRAM sizes on their 16nm FF/16nm FF+ and 14nm FF LPE and LPP.

 

Remember size is less interesting if performance is degredated too much. So for Intel to use their high density 0.0499 µm2 SRAm makes little sense, if their High Performance 0.0706 µm2 SRAM are much better in performance. Both are 14nm FF based btw.

 

That being said, we know AMD are much better at compressing architecture than Intel. So does it matter that Intel's process is 15% more dense, if AMD's architecture is e.g.30% denser?

4 minutes ago, ZetZet said:

i hope intel finally fixes their naming scheme and makes i3 a 4 core i5 a 6 core and i7 an 8 core.

No gamer should ever buy an i3 as is, as most games will be quad core min spec. So yeah, Intel needs to get their act together. That is also what Zen should get us.

[ZetZet]

2 minutes ago, Notional said:

 

i3 is a quad.

[Lauen]

Just now, ZetZet said:

i3 is a quad.

in one way, yes. they all have hyperthreaded cores. 2 of them. http://www.intel.com/content/dam/support/us/en/documents/processors/corei3/sb/IntelCorei3ComparisonChart10.pdf

[ZetZet]

1 minute ago, Lauen said:

in one way, yes. they all have hyperthreaded cores. 2 of them. http://www.intel.com/content/dam/support/us/en/documents/processors/corei3/sb/IntelCorei3ComparisonChart10.pdf

yes, but for games it's a quad

[Notional]

4 minutes ago, ZetZet said:

i3 is a quad.

Intel disagrees with you:http://www.intel.com/content/www/us/en/processors/core/core-i3-processor.html

 

Hyperthreads MIGHT fool a game to run, but it will run very badly. The 2 extra threads can only run on parts of the pipeline not occupied by the 2 primary threads. So if all 4 threads needs the same ressources at the same time, the game will crash and burn. So no an i3 is not at all useful for quad core min spec games.

[79wjd]

13 minutes ago, Notional said:

Intel disagrees with you:http://www.intel.com/content/www/us/en/processors/core/core-i3-processor.html

 

Hyperthreads MIGHT fool a game to run, but it will run very badly. The 2 extra threads can only run on parts of the pipeline not occupied by the 2 primary threads. So if all 4 threads needs the same ressources at the same time, the game will crash and burn. So no an i3 is not at all useful for quad core min spec games.

Other than Arma 3 (which is a mess even on an i5), what games don't run well on an i3?

[Castdeath97]

5 minutes ago, djdwosk97 said:

Other than Arma 3 (which is a mess even on an i5), what games don't run well on an i3?

Rise of the Tomb Raider 

[ZetZet]

22 minutes ago, Notional said:

Hyperthreads MIGHT fool a game to run, but it will run very badly.

uh, based on what?

[ZetZet]

5 minutes ago, Castdeath97 said:

Rise of the Tomb Raider 

https://www.youtube.com/watch?v=Vu0MWxQtAfU

[Notional]

6 minutes ago, djdwosk97 said:

Other than Arma 3 (which is a mess even on an i5), what games don't run well on an i3?

I don't keep up with i3 game benchmarks, so I wouldn't know. The issue is that games becomes more multithreaded as times go (some can use up to 8 threads now). I3's will have issues fast, especially if all 4 threads starts to clash.

 

Then again, if you have an i3 you probably don't have a great GPU either, so who knows. I just still feel sorry for all those poor bastards who bought Pentium anniversary editions (and later models), that are pure dual core. They should have gotten AMD CPU's instead. Because more and more games outright refuses to run on those chips.

[Castdeath97]

3 minutes ago, ZetZet said:

https://www.youtube.com/watch?v=Vu0MWxQtAfU

I remember digital foundry encounter a issue with it.

[79wjd]

8 minutes ago, Castdeath97 said:

Rise of the Tomb Raider 

http://www.techspot.com/review/1128-rise-of-the-tomb-raider-benchmarks/page5.html

4 minutes ago, Notional said:

I don't keep up with i3 game benchmarks, so I wouldn't know. The issue is that games becomes more multithreaded as times go (some can use up to 8 threads now). I3's will have issues fast, especially if all 4 threads starts to clash.

In the future, no doubt. But for right now i3s are still perfectly good. There have been games that are able to run eight cores for a couple years now.

[AlTech]

1 hour ago, DocSwag said:

I see what you are saying, but you are saying AMD will price Zen over 100 dollars below Intel's offering, or about 30% lower. I said just below Intel. I'm pretty sure just below doesn't mean 30% lower. I'm saying that AMD may price Zen, say, 40 dollars below Intel.

 

True, that is a good point. I am making the assumption Zen is good because AMD they are screwed if otherwise. However, it is true I might be wrong.

If it isn't cheaper than why will anyone give a shit?

If I had to buy a new CPU I would pick AMD If Zen were released but as it stands I would not buy an AMD CPU for $40 less than the equivalent Intel. Intel's quad cores are way overpriced. IMO. They should be around $120-160 not $180-250).

 

And then companies wonder why customers don't upgrade PC's frequently. Part of it is to do with cost and the other part is performance benefit they will get.

 

[GoZone]

We need better APUs for cheaper builds so I want to see how this will go for AMD.

[420istoday]

1 hour ago, ZetZet said:

yes, but then he said 6 core should cost 350-400 that doesnt make any sense. it should be 200.

Nope, by my math a 6 core zen will handily outperform a similar 4 core Intel CPU. Simply because the IPC is only expected to be about 10-15% below Skylake, and it has 2 more physical cores and 4 more threads. So yeah, a 6 core zen cpu would be worth about $350-$400.

[ZetZet]

30 minutes ago, Castdeath97 said:

I remember digital foundry encounter a issue with it.

it's 750ti fucking up not the i3.

[Starelementpoke]

2 minutes ago, 420istoday said:

Nope, by my math a 6 core zen will handily outperform a similar 4 core Intel CPU. Simply because the IPC is only expected to be about 10-15% below Skylake, and it has 2 more physical cores and 4 more threads. So yeah, a 6 core zen cpu would be worth about $350-$400.

Dude, you can´t say,¨by my math,¨ and not show your math.

[Agent181]

well shit this was interesting to read lol

[patrickjp93]

1 hour ago, Lauen said:

the toms hardware slides posted in this thread earlier showed that power draw in torture tests were like.. one or two watts under rated TDP. 

 

EDIT: in this document, Intel have gotten details from AMD about it.. apparently. 

screencap: 

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

[Demonking]

2 hours ago, DioOmicida said:

Ii doesn't surprise me that they launch with high core count CPU's.
They do like their cores. Lets see what they can do.

well in FX case modules. 

[Notional]

9 minutes ago, ZetZet said:

it's 750ti fucking up not the i3.

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

8 minutes ago, Starelementpoke said:

Dude, you can´t say,¨by my math,¨ and not show your math.

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).

[Starelementpoke]

1 minute 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).

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

[420istoday]

3 minutes ago, Starelementpoke said:

Dude, you can´t say,¨by my math,¨ and not show your math.

I mean, I gave you all of the information you needed to do it yourself.

 

Say you have a synthetic benchmark and zen performed about 15% worse than skylake in a single threaded workload.

Your scores would look like this: Zen - 85, Skylake - 100.

 

But when you consider multi-threaded workloads, a 6-core zen cpu will have 12 threads, where a 4-core skylake i7(~$330) will have only 8.

Thus your scores become: Zen - 1020, Skylake - 800.

 

This is a direct IPC calculation though, and doesn't consider that hyper-threads don't scale perfectly, clock-speeds, system bottlenecks, silicon lottery, or even hard evidence that zen performs that well.

With the information we have now, this is the best estimate.

[Starelementpoke]

3 minutes ago, 420istoday said:

I mean, I gave you all of the information you needed to do it yourself.

 

Say you have a synthetic benchmark and zen performed about 15% worse than skylake in a single threaded workload.

Your scores would look like this: Zen - 85, Skylake - 100.

 

But when you consider multi-threaded workloads, a 6-core zen cpu will have 12 threads, where a 4-core skylake i7(~$330) will have only 8.

Thus your scores become: Zen - 1020, Skylake - 800.

 

This is a direct IPC calculation though, and doesn't consider that hyper-threads don't scale perfectly, clock-speeds, system bottlenecks, silicon lottery, or even hard evidence that zen performs that well.

With the information we have now, this is the best estimate.

Your the one saying it, not me. On you to show the math.