AMD Zen to Bring Greater than 40% IPC Increase, New Reports Suggests

[Trixanity]

IPC has nothing to do with nodes what so ever. It's only about architecture and numbers of cores/transistors.

I'm very much aware of that. I'm trying to dispel all the falsehoods floating around in as simple terms as possible.

Many see a 40% IPC increase as being strictly a 40% performance increase and therefore seem to think that the majority of that comes from the 14nm process.

That's why I say it's independent of each other because they are not related. Of course one could go into technical details but that usually doesn't get the point across as easily.

The same way that the "40% over bulldozer, right?" still pops up despite it being refuted so many times. Excavator is already roughly 30% over bulldozer, so a 10% increase with a brand new architecture would have been very unimpressive especially considering there was a bigger jump from Steamroller to Excavator.

As has been said in this thread: in IPC alone, Zen is around or slightly ahead of Haswell. What does this mean in actual real world performance? Hard to tell. There are so many more variables than simple IPC. Zen could be inferior in a lot of other aspects than Intels Haswell and therefore be slower in many scenarios or it could be better and therefore be faster. One thing is true though and that's the IPC being in the neighborhood of Haswell and the process node being more advanced. So it has a good chance of beating Haswell which currently is still very relevant since Haswell-E is still Intels current enthusiast platform.

Of course Intels Broadwell-E and Kaby Lake is coming. The former supposedly in Q2 and the latter in Q4 I think. So by the time Zen launches it will be facing two new platforms from Intel. So it might be slightly behind already but will supposedly be a very viable alternative as well as very competitive in price-to-performance.

Edit: regarding the 4 core vs 8 core debate. 8 core Zen aka Summit Ridge is an enthusiast platform. It's competing with X99 and the likes of 5960X. It's marketed as high end desktop just like X99 is.

The current Skylake offerings are 4 cores and marketed as mainstream. An 8 core Zen is not competing against that in either price or performance. I don't think there's any chance of a Skylake 4 core SKU beating an 8 core Zen in multi threaded workloads. Single threaded though? Probably or rather it will most likely. Just like a current Skylake processor beats Intels own 999 dollar products in single threaded because of clock speed and slightly better architecture.

[AlexGoesHigh]

wrong

FX rocks

FX kicked pentuim 4 butt in the mid 2000s leading to the most successful period in AMD's history. FX was the branding used for the Athlon 64 flagship

I know that but those are almost 10+ year old chips while bulldozer, piledriver and excavator are fresh of people memories, i think i wouldn't be wrong to say that a lot more people could associate the FX name with the current chips that the old ones out of their heads, also Zen is so damn different to bulldozer they shouldn't even use FX to differentiate them more for previous CPU's, they should do what Intel did with Nehalem that they switched from Core 2 to Core i Series branding.

[Prysin]

 

 

 

 

That makes no sense...

 

Intel was leading in the benchmarks.

We want to know how much AMD has to improve to match them.

We calculate how much that is in %.

 

For example:

If we have 80 and want to get to 100, we need to add 25%.

25% of 80 is 20.

80 + 20 is 100. Therefore, we need 25% more.

That's how I counted. If Intel scores 8372 and AMD scores 3987 then we need to add 109% to AMD's score to match Intel's. That type of math is 100% correct.

 

yes, and that still makes no sense.

Why?

because that means you, LawlZ, must choose a value as the "answer". This means the choice is subjective. You can go the other way and choose AMD as the answer and get a wildly different value % wise.

This is why you need to use the average % difference.

 

https://www.mathsisfun.com/percentage-difference.html

 

read up on the explanation of "Which should we use"... it explains it perhaps a bit better then i do.

 

So, i will stick to CB R11.5 because they have easy values to work with.

To make sure IPC is the single common denominator, you must eliminate frequency differences. IPC means Instructions per clock. If you have fewer instructions per clock, you need more clock ticks per second to counter for that. This is why a 4790k scores marginally higher then a 4690k in single core performance. Clock speeds matters to determine single core performance.

CB 11.5 single thread scores:

FX 8150 3.6GHz = 0.97

FX 8350 3.5GHz = 1.05

I5 2500k 3.3GHz = 1.47

i5 3570k 3.4GHz = 1.54

i5 4690k 3.5GHz = 1.7

i5 6600k 3.5GHz = 1.92

 

 

so the differences are as following.

Difference between FX 8150 and FX 8320 is:

 

Difference: 0.97 - 1.05= 0.08

Average is: (1.05 + 0.97) / 2 = 1.01

Average % difference is: 0.08/1.01 x 100 = 7.9%

 

IPC of Piledriver is roughly 7.9% higher then that of Bulldozer. Which is slightly lower then AMDs own confirmed 10%, but that may be simply due to the benchmark run itself.

 

 

Now... Piledriver is todays product, Bulldozer is no longer produced, as as such, we will now ignore bulldozer going on.

 

Sandy vs Piledriver

Difference: 1.05 - 1.47 = 0.42

Average is: (1.05+1.47) / 2 = 1.26

Average % difference is: 0.42 / 1.26 x 100 = 33.33%

 

Sandy Bridge, core vs core (although at unequal clock speeds) are 33.33% faster. If FX and Sandy was at equal speeds, sandy would be even faster still.

 

Ivy vs Piledriver:

Difference: 1.05 - 1.54 = 0.49

Average is: (1.05 + 1.54) / 2 = 1.295

Average % difference is: 0.49 / 1.295 x 100 = 37.83%

 

Ivy Bridge, core vs core (although at unequal clock speeds) are 37.83% faster then Piledriver.

 

Haswell vs Piledriver

Difference: 1.05 - 1.7 = 0.65

Average is: (1.05 + 1.7) / 2 = 1.375

Average % difference is: 0.65 / 1.375 x 100 = 47.2727%

 

the average % difference is 47.2727%

 

Skylake vs Piledriver

Difference is: 1.05-1.92 = 0.87

Average is: (1.05 + 1.92) / 2 = 1.485

Average %  difference is: 0.87 / 1.485 x 100 = 58.58%

 

NOW... here is why average matters.

 

If we were to look at differences between Haswell and Piledriver going purely by score.

Selecting Piledriver as reference: 1.05 / 1.7 x 100 = 61% (meaning from piledriver UP to haswell)

Selecting Haswell as reference: 1.7 / 1.05 x 100 = 161.9%  (meaning from haswell DOWN to piledriver)

 

You see the issue by going purely by score now?

We know for a fact that haswell i5 4690k is NOT 161% stronger then FX 8320. Not by a long-shot. And we know that 61% is also slightly off. This is the importance of average. To remove the "subjective" part of the equation.

[LAwLz]

-snip-

That makes absolutely no sense whatsoever.

Let me ask you this, how many percent do you have to add to go from 100 to 140? Anyone with a basic understanding of math will say 40%, and that's what we should be talking about. How many percent does AMD need to improve in order to get to Intel's performance. You might call this "choosing Intel as the answer" but the reason why I "chose" that was because AMD is behind Intel, and is trying to catch up.

 

There is absolutely nothing subjective about this.

 

 

Did you even read the page you linked?

When you are talking about an old value, and a new value (such as, how much faster AMD's new processor has to be to be as good as Skylake) then you use "percentage change".

 

Here is the mathisfun page for percentage change, in case you don't know how to do that.

 

So let's take the Cinebench numbers again:

Intel (Cinebench 11.5): 1.92

AMD (Cinebench 11.5) : 1.05

 

I said Intel is leading by 83% in that benchmark. What that means is that AMD would have to improve the performance of the 8320 by 83% to match Intel. That is the truth, and the only truth.

You should NOT use the average method when talking about how much something has to change (which is what we are talking about, since we want to know how much AMD has to improve to match Intel, right?).

 

 

Let's see what happens if we use the average method of counting in order to estimate how much AMD must improve, shall we?

 

Difference: 1.92 - 1.05 = 0.87

Average: (1.92 + 1.05) / 2 = 1.485

 

0.87 as a percentage of 1.485: 0.87/1.485 * 100 = 58.59%

 

 

Now, let's see what happens when we take AMD's current score and adds 58.59% to it shall we? Do you think it will match Intel's score?

 

1.05 * 1.5859 = 1.67

 

Oh... How strange... If we use the method you suggested to count the percentage then we don't end up with how much AMD has to improve to get to Intel's performance levels. However, if we use my number...

 

1.05 * 1.83 = 1.92

 

 

Would you look at that! They match perfectly!

 

 

Now can you please stop using incorrect math to try and make AMD look better than they really are? Nobody benefits from you doing these mental gymnastics.

[Humbug]

Is IPC actually relevant to direct real world performance. Or is it just another big number we can all gawp at.

It matters quite a bit. (instructions per clock)

 

Hypothetically if a 4Ghz FX-8350 is modified and given a 40% IPC boost then that means it would only need a 2.4Ghz clock-speed to achieve the same performance. So then you have a CPU which is running much quieter and producing less heat. Or alternatively if you clock it back upto 4Ghz and get 40% faster performance for the same TDP. And AMD isn't even talking about 40% IPC boost above piledriver, they are talking about 40% over excavator.

 

However there are other factors which come into play. Particularly how well does Zen clock. Hypothetically if it cannot exceed 3Ghz then it will be well faster than AMD's current line-up but nothing earth shattering. On the other hand if it can do 4Ghz + more overclock headroom combined with the IPC gain then AMD is back in the game... But we just don't know right now.

[looncraz]

Good point, although I am pretty skeptical of that benchmark. Nehalem to Sandy Bridge was a huge leap and in that graph it only shows it being like a 10% increase in IPC.

Also, do we have good numbers for how Excavator performs, and are we sure AMD means 40% increase over Excavator and not over Bulldozer?

 

Actually, Sandy Bridge was not as large of jump for IPC as it may first appear, though it was a big jump for clock rates.  While many popular benchmarks (such as Cinebench) show a near 20% improvement, most real world applications (which I use in my multibench) show about a 9% improvement on average.  Intel only claimed 11%, and many have found that to be generally accurate.

 

https://en.wikipedia.org/wiki/Sandy_Bridge#Performance

 

The difference with my numbers is that I exclude instruction set improvements and try to account for other variables (RAM performance, mainly), so my numbers are sometimes lower.

 

However, I actually witnessed larger improvements than claimed for Ivy Bridge and Haswell.  By the time we get to Haswell, Intel claims a 39% total improvement over Penryn, and I see 37%.  Skylake has a larger spread, though, as Skylake has very targeted improvements (and those can be quite large).

[RexinOridle]

AMD CPU side has been all talk for the last few years. Their GPU is far superior when it comes to raw power but no application optimized for them. Makes me real sad.

[Djole123]

Thank God!!!

[Prysin]

Actually, Sandy Bridge was not as large of jump for IPC as it may first appear, though it was a big jump for clock rates.  While many popular benchmarks (such as Cinebench) show a near 20% improvement, most real world applications (which I use in my multibench) show about a 9% improvement on average.  Intel only claimed 11%, and many have found that to be generally accurate.

 

https://en.wikipedia.org/wiki/Sandy_Bridge#Performance

 

The difference with my numbers is that I exclude instruction set improvements and try to account for other variables (RAM performance, mainly), so my numbers are sometimes lower.

 

However, I actually witnessed larger improvements than claimed for Ivy Bridge and Haswell.  By the time we get to Haswell, Intel claims a 39% total improvement over Penryn, and I see 37%.  Skylake has a larger spread, though, as Skylake has very targeted improvements (and those can be quite large).

yeah, DDR4 also makes it hard to get a good apples to apples comparison, as the internal latencies (bus timings) are nowhere near DDR3 latencies...

 

although, @MageTank would probably be thrilled to spend some time dissecting how to match everything for a proper apples to apples comparison

[Prysin]

You might call this "choosing Intel as the answer" but the reason why I "chose" that was because AMD is behind Intel, and is trying to catch up.

 

There is absolutely nothing subjective about this.

uhm, any choice made by a human being, in this case you, is 100% subjective.

 

to explain to you, since you seem to be incapable of understandign the meaning of this:

 

sub·jec·tive (səb-jĕk′tĭv)

adj.

1.

a. Dependent on or taking place in a person's mind rather than the external world: "The sensation of pain is a highly subjective experience that varies by culture as well as by individual temperament and situation" (John Hoberman).

b. Based on a given person's experience, understanding, and feelings; personal or individual: admitted he was making a highly subjective judgment.

2. Psychology Not caused by external stimuli.

3. Medicine Of, relating to, or designating a symptom or complaint perceived by a patient.

4. Expressing or bringing into prominence the individuality of the artist or author.

5. Grammar Relating to or being the nominative case.

6. Relating to the real nature of something; essential.

source: http://www.thefreedictionary.com/subjective

 

 

 

ob·jec·tive (əb-jĕk′tĭv)

adj.

1.

a. Existing independent of or external to the mind; actual or real: objective reality.

b. Based on observable phenomena; empirical: objective facts.

2. Uninfluenced by emotions or personal prejudices: an objective critic. See Synonyms at fair¹.

3. Medicine Relating to or being an indicator of disease, such as a physical sign, laboratory test, or x-ray that can be observed or verified by someone other than the person being evaluated.

4. Grammar

a. Of, relating to, or being the case of a noun or pronoun that serves as the object of a verb.

b. Of or relating to a noun or pronoun used in this case.

n.

1. Something worked toward or striven for; a goal. See Synonyms at intention.

2. A thing or group of things existing independent of the mind.

3. Grammar

a. The objective case.

b. A noun or pronoun in the objective case.

4. The primary optical element, such as a lens or mirror, in a microscope, camera, telescope, or other optical instrument, that first receives light rays from the object and forms the image. Also called object glass, objective lens, object lens.

Source: http://www.thefreedictionary.com/objective

 

Hopefully you will now realize the fallacy of your statement.

 

Also, you fail to understand my math, quite horribly so. But that is fine, ill leave it up to you to try understand what it is your are not understanding, rather then spend time having you twist my words back and forth to suit your own opinion.

 

cheers.

[FellTheSky]

the question is 40% from what? and at what frequency?

 

if zen scores stock sandy bridge single thread, but running at 5 ghz i dont see it being competitive with intel

[Dabombinable]

Or, they do what Intel does, and advertise TDP as "Typical use". 4790k has an 88w TDP under typical use circumstances, but touch AVX2, and you can add 50w to that easily. Under normal use, 95w TDP with 8 cores does not sound too impossible, but I do not see them having 5ghz chips. I am expecting 3.6-4ghz base out of the 8 core models. Smaller core configurations might start with higher base clock speeds, but it's all speculation at this point.

 

My only hope is that we see a more refined memory controller this time. AMD's lackluster memory support on previous FX series was a major turnoff. Seeing as they wont be using a dual IMC like Skylake, I hope they focus on a really strong IMC for DDR4. 

Considering with the Mobile Phenom II lineup AMD had the quad cores with the lowest stock clock speeds, and the dual cores with the highest (and the tri cores being in the middle), they probably will have the 8 core versions (if they do come out) with the lower clock speeds, and the dual and quad core variants with a higher clock speed.

 

 

 

 

If AMD actually comes through with Zen and if they start making CPU that actually have 8 cores (The FX series should have been marketed as being CMT not SMT-which is why people think they have 4, 6 and 8 cores) and the IPC is around that of Haswell or better, then I'll be doing the following:

• Selling my 4790K+Z97 Sabertooth MKII+Seidon 120V
• Selling my i5 4440+H87M Pro+Arctic Freezer Pro 7 Rev 2.0
• Selling my GTX 650 ti OC 2GB
• Selling my GTX 970 G1 Gaming (really got screwed over by fucking Gigabyte with it-its not binned at all+its cooler is audible over the case fans at idle)
• Selling all 4x4GB sticks of DDR3 (might bundle them with the motherboards, 8GB each)
• Buying the best Zen+mobo+DDR4+390X Sapphire/XFX combo that I can afford
• Using my old and good Coolermaster Eclipse (which came out before quad cores and can handle well over 120W with this cheap and-far-from-best thermal compound(couldn't go wrong, $25 for a 150g tube that doesn't require burn in))

[Glenwing]

On 1/22/2016 at 10:00 AM, Prysin said:

yes, and that still makes no sense.

Why?

because that means you, LawlZ, must choose a value as the "answer". This means the choice is subjective. You can go the other way and choose AMD as the answer and get a wildly different value % wise.

This is why you need to use the average % difference.

 

https://www.mathsisfun.com/percentage-difference.html

 

read up on the explanation of "Which should we use"... it explains it perhaps a bit better then i do.

 

So, i will stick to CB R11.5 because they have easy values to work with.

To make sure IPC is the single common denominator, you must eliminate frequency differences. IPC means Instructions per clock. If you have fewer instructions per clock, you need more clock ticks per second to counter for that. This is why a 4790k scores marginally higher then a 4690k in single core performance. Clock speeds matters to determine single core performance.

CB 11.5 single thread scores:

FX 8150 3.6GHz = 0.97

FX 8350 3.5GHz = 1.05

I5 2500k 3.3GHz = 1.47

i5 3570k 3.4GHz = 1.54

i5 4690k 3.5GHz = 1.7

i5 6600k 3.5GHz = 1.92

 

 

so the differences are as following.

Difference between FX 8150 and FX 8320 is:

 

Difference: 0.97 - 1.05= 0.08

Average is: (1.05 + 0.97) / 2 = 1.01

Average % difference is: 0.08/1.01 x 100 = 7.9%

You have absolutely no idea what you are doing. Have you tried checking your answer to even see if your math works?

 

 

On 1/22/2016 at 10:00 AM, Prysin said:

Sandy vs Piledriver

Difference: 1.05 - 1.47 = 0.42

Average is: (1.05+1.47) / 2 = 1.26

Average % difference is: 0.42 / 1.26 x 100 = 33.33%

Sandy Bridge, core vs core (although at unequal clock speeds) are 33.33% faster. If FX and Sandy was at equal speeds, sandy would be even faster still.

If that's the case, then if Piledrivers's score is increased to 133.33% of its normal value, it should be the same as Sandy Bridge. That's what "33.33% higher" means.

Let's check: 1.05 × 1.3333 = 1.3999. Whoops, that's not quite 1.47.

The actual percentages are quite simple:

1.47 / 1.05 = 1.40. Sandy Bridge's score is 140% of Piledriver's score, which is in fact 40% higher than Piledriver (Piledriver naturally is 100% of Piledriver's score, and 140% is 40% more than 100%).

1.05 / 1.47 = 0.71. Piledriver's score is 71% of Sandy Bridge's score, which is in fact 29% less than Sandy Bridge (Sandy Bridge naturally is 100% of Sandy Bridge's score, and 71% is 29% less than 100%).

So, if you want to make statements:
Sandy Bridge's score is 40% higher than Piledriver's. (But Piledriver's score is NOT 40% lower than Sandy Bridge)

Piledriver's score is 29% lower than Sandy Bridge's. (But Sandy Bridge's score is NOT 29% higher than Piledriver)

So yes, wording matters. But if you understand what means what, the math is very clear cut, you just need to understand which direction you're going. But at no point does any of this averaging nonsense come into play.

Let's take some more workable numbers as an example using your "method":

CPU A Score: 100

CPU B Score: 200

Difference: 200 - 100 = 100

Average: (200 + 100) / 2 = 150

Average % Difference: 100/150 × 100 = 66.6%.
According to you, 200 is 66% higher than 100.

Yes, as you've discovered, percentages invert when you switch the direction of comparison. This does not make them arbitrary numbers. Every pair of numbers has two percentages associated with that pair.

Statement: 50% of 200 is 100

Validation in both directions:
200 × 50% = 100
100 / 50% = 200

Statement: 200% of 100 is 200

Validation in both directions:

100 × 200% = 200

200 / 200% = 100

If you have a little training in algebra you'll notice that all of these are the same equation rearranged (which is what we'd expect if they're all true and accurate). Take that last equation for example — 200 / 200% = 100. Dividing by 200% is the same as multiplying by 1 / 200%, which is to say, dividing by 2 is the same as multiplying by 1/2.

So, we have 200 × (1 / 200%) = 100, or 200 × (1 / 2) = 100. But 1/200% (1/2) is the same as 50%. So really this equation is the same thing as 200 × 50% = 100, which is to say, 50% of 200 is 100, which is the original statement from the upper pair. So, all four equations are the same; 100 is 50% of 200, and 200 is 200% of 100. They're both forms of saying the exact same equality, even though the percentages are different depending on how you phrase it.

But that does not mean you can just do some tricks with your wording and make it into any percentage you want. It isn't subjective. You may notice there is a relationship between the two percentages, 50% and 200%, namely that they are reciprocals; 50% is (1 / 200%), and 200% is (1 / 50%). There are exactly two percentages associated with any comparison, depending on the direction of comparison that you use when phrasing the statement, but they are always reciprocals of each other, not just any arbitrary percentage you want.
 

If AMD says "40% increased IPC over excavator" which IS what they said, then it is as simple as this:

(Excavator IPC) + (another 40% of Excavator's IPC) = New CPU IPC

Or if you want to get fancy, (Excavator IPC × 1.4) = New CPU IPC
 

You can review the scores of Intel CPUs compared with Excavator (or whatever) to see what they'd have to be multiplied by to match them.

Taking your numbers, 1.05 for Piledriver and 1.92 for Skylake.

1.05 × (???) = 1.92

To solve for the number, just isolate it by dividing both sides by 1.05.

(???) = 1.92 / 1.05

Just do the division operation, and the number is:

1.83. You can verify this by substituting it back into the original problem:

1.05 × 1.83, the answer is indeed 1.92

This means that Piledriver's score (1.05) must be multiplied by 1.83 to match Skylake.

You could actually split that up and distribute if you want to validate that "an 83% increase" translates to multiplying by 1.83:

As we just established above, 1.05 × 1.83 = 1.92, so in other words:
Piledriver score × 1.83 = Skylake score
1.83 is the same as (1 + 0.83), and "Piledriver score" is equal to 1.05 in this case, so substitute those into the equation:
1.05 × (1 + 0.83) = Skylake score

Distribute the 1.05:
(1 × 1.05) + (0.83 × 1.05) = Skylake score
1.05 + (0.83 × 1.05) = Skylake score

If you remember some basic algebra word problem training, remember that "of" means "multiply" and vice versa, so "0.83 × 1.05" translates to "83% of 1.05", and since 1.05 is Piledriver's score, this whole statement translates directly to:

Piledriver score + (83% of Piledriver score) = Skylake score

"+" means "increased by", so Piledrivers score, increased by 83% of Piledriver's score, is the same as Skylake's score. Therefore, Piledriver's score must be increased by 83% of itself to equal Skylake's score. We got this statement from simply rearranging the original statement algebraically, so these two statements are therefore equivalent, and thus, "increase by 83%" is equivalent to multiplying by 1.83.

Piledriver score × 1.83 = Skylake score
So, Piledriver's score must be increased by 83% to match Skylake.

It's amazing what you can do with some basic knowledge of algebra.
 

(EDIT: Before you post any more links, just a word on "percentage difference" which is the averaging stuff you are talking about and is not applicable at all here; "Percentage difference" is used for multiple measurement trials of the same thing to quantify the spread of the margin of error as a percentage, because in science you run tests more than once to ensure your measurement process is consistent and your results are repeatable. "Percentage Difference" is used to quantify the consistency of your multiple measurement trials of the same thing. For example if you ran Cinebench 5 times in a row with no hardware changes and got slightly different scores each time, you would use your averaging method to find the "percentage difference" between the results from the same CPU. It is not used to compare the difference between the results of different experiments as a percentage.

[marldorthegreat]

the question is 40% from what? and at what frequency?

 

if zen scores stock sandy bridge single thread, but running at 5 ghz i dont see it being competitive with intel

I think the bigger one is, is this per core, or per cpu. And 5ghz can be competitive if power consumption is low enough and heat is lower

[That Norwegian Guy]

No, if AMD deliver's on their promise, expect to pay $400+ for a Zen 8-core.

 

 

Nope, if AMD exceeds Intel's enthusiast (read: more money than sense consumers) pricing they will be in for a world of hurt.

 

Zen are consumer CPUs with consumer-level die size regardless of nearly reaching/ matching the levels of old Intel enthusiast chips.

 

They might price Zen at the level of Intel consumer 4-cores, but that would be it. They are far away from the brand recognition to be able to pull Intel's crap.

[Jahramika]

There are other factors as well such as M.2 PCIE 3.0 vs SSD or Sata at PCIE 2.0 Gen speeds that add to performance. Wont know to we see new Zen operating on the same variables as Intel. Until then t many factors to say this or that. How much of the 11 percent gain from X97 to Z170 was due to M.2 speed increase. On a X97 Samsung 951 was like 800 ish on a Z170 well t goes to 1500 read speeds. I guess my point is a lot of factors effect performance. From my eyes M.2 increases system performance more than going from x97 to Z170 than the CPU upgrade.

[Lawliet93]

Hi guys, does anybod know, if Zen is going to offer more than one thread per core? (it was said, that Zen is going to have 2 threads per core): http://techreport.com/review/28228/amd-zen-chips-headed-to-desktops-servers-in-2016

[Dabombinable]

Hi guys, does anybod know, if Zen is going to offer more than one thread per core? (it was said, that Zen is going to have 2 threads per core): http://techreport.com/review/28228/amd-zen-chips-headed-to-desktops-servers-in-2016

AMD is supposedly including their own version of hyperthreading, so it should have 2 threads per core. It remains to be seen however if it boosts performance by 20% per core (like Pentium 4) or 40% core (like i3, i7 and mobile i3/i5/i7.

[TechGod]

Hi guys, does anybod know, if Zen is going to offer more than one thread per core? (it was said, that Zen is going to have 2 threads per core): http://techreport.com/review/28228/amd-zen-chips-headed-to-desktops-servers-in-2016

Yes it will...

[Trixanity]

Hi guys, does anybod know, if Zen is going to offer more than one thread per core? (it was said, that Zen is going to have 2 threads per core): http://techreport.com/review/28228/amd-zen-chips-headed-to-desktops-servers-in-2016

Yes. Some products will at least. I don't think it will be enabled on all products.

[TechGod]

the question is 40% from what? and at what frequency?

if zen scores stock sandy bridge single thread, but running at 5 ghz i dont see it being competitive with intel

Are you insane? It's 40% IPC over Excavator. Plus the 14nm LPP process will allow it to clock pretty high so expect better than Haswell level performance.

[Lawliet93]

Yes it will...

 

NICE!!! If it´s going to be like 15% slower than intels offering and about 15-20% cheaper...I will buy a 4c/8t CPU

[patrickjp93]

Nope, if AMD exceeds Intel's enthusiast (read: more money than sense consumers) pricing they will be in for a world of hurt.

Zen are consumer CPUs with consumer-level die size regardless of nearly reaching/ matching the levels of old Intel enthusiast chips.

They might price Zen at the level of Intel consumer 4-cores, but that would be it. They are far away from the brand recognition to be able to pull Intel's crap.

Intel isn't pulling any crap. Like every company good at business, they address different market needs with market segmentation and price accordingly. AMD asked for $1000 for the FX 9590 at launch, so you can't complain about Intel's pricing. If they were overpriced their profits would tank because no one would buy. And it's not like software is actually becoming that much more CPU-intense outside the HPC and professional spaces. What exactly is Intel doing that's so criminal to you?

[patrickjp93]

Are you insane? It's 40% IPC over Excavator. Plus the 14nm LPP process will allow it to clock pretty high so expect better than Haswell level performance.

FinFET ate into Intel's clockspeed ceiling after Sandy Bridge. Don't expect super high clocks, especially with 8 cores that each have dedicated AVX 256 units.

[FireFox]

40% from bulldozer?   :lol:

 

40% from Piledriver can't be much better than Sandy or Ivy ehh?