Discussion: AMD Ryzen

[8350rocks]

http://www.anandtech.com/bench/CPU/1028

Even if you compare the 6300 and the i3-4360, the 6300 is 3.6 stock, and the i3 is 3.7 stock

FX6350 is 96 in ST, and i3-4360 is 149 in ST.

40% is 134

50% is 144

55% is 148

Hmm...148 vs 149, with intel having a negligible clock advantage, or just about haswell. That is going from PD as well...not Excavator.

So, I would say, haswell has to be *really* close to what the ST performance will be.

EDIT: Just above the i3 is an i7-6850K (BWE) with a ST score of 152 at same clocks as FX6350, which is right in the ballpark. I think they are really close on this one.

 

[sarinaide]

http://www.anandtech.com/bench/CPU/1028

Even if you compare the 6350 and the i3-4360, the 6350 is 3.6 stock, and the i3 is 3.7 stock

FX6350 is 96 in ST, and i3-4360 is 149 in ST.

40% is 134

50% is 144

55% is 148

Hmm...148 vs 149, with intel having a negligible clock advantage, or just about haswell. That is going from PD as well...not Excavator.

So, I would say, haswell has to be *really* close to what the ST performance will be.

EDIT: Just above the i3 is an i7-6850K (BWE) with a ST score of 152 at same clocks as FX6350, which is right in the ballpark. I think they are really close on this one.

The issue is that no benchmark is universal and not even cinebench R15 in the disclaimer say that they measure true IPC but give only a CPU design throughput. IPC and throughput is not the same.

 

[8350rocks]

http://www.anandtech.com/bench/CPU/1028

Even if you compare the 6350 and the i3-4360, the 6350 is 3.6 stock, and the i3 is 3.7 stock

FX6350 is 96 in ST, and i3-4360 is 149 in ST.

40% is 134

50% is 144

55% is 148

Hmm...148 vs 149, with intel having a negligible clock advantage, or just about haswell. That is going from PD as well...not Excavator.

So, I would say, haswell has to be *really* close to what the ST performance will be.

EDIT: Just above the i3 is an i7-6850K (BWE) with a ST score of 152 at same clocks as FX6350, which is right in the ballpark. I think they are really close on this one.

The issue is that no benchmark is universal and not even cinebench R15 in the disclaimer say that they measure true IPC but give only a CPU design throughput. IPC and throughput is not the same.

They measure ST throughput though with that aspect of the benchmark. Those are the single thread benchmarks, multi-thread would clearly have the 6850K significantly higher than both the i3 and the FX 6350.

 

[sarinaide]

http://www.anandtech.com/bench/CPU/1028

Even if you compare the 6350 and the i3-4360, the 6350 is 3.6 stock, and the i3 is 3.7 stock

FX6350 is 96 in ST, and i3-4360 is 149 in ST.

40% is 134

50% is 144

55% is 148

Hmm...148 vs 149, with intel having a negligible clock advantage, or just about haswell. That is going from PD as well...not Excavator.

So, I would say, haswell has to be *really* close to what the ST performance will be.

EDIT: Just above the i3 is an i7-6850K (BWE) with a ST score of 152 at same clocks as FX6350, which is right in the ballpark. I think they are really close on this one.

The issue is that no benchmark is universal and not even cinebench R15 in the disclaimer say that they measure true IPC but give only a CPU design throughput. IPC and throughput is not the same.

They measure ST throughput though with that aspect of the benchmark. Those are the single thread benchmarks, multi-thread would clearly have the 6850K significantly higher than both the i3 and the FX 6350.

You didn't factor in turbo, the FX6350 runs a turbo around 4.2 so to score at 3.7 it gets 86. The Bulldozer uarch was fundamentally bad for IPC, even a Deneb and Thuban ran circles around a FX8370 clock for clock.

The point of contention is, if AMD means that their CPU can universally at least increase instructions per cycle by 40, that could alone equate to 50% in one bench while 80-90% in another benchmark. You cannot blanket a generic 40% across every bench.

COMMENT=Results Disclaimer - CINEBENCH results are indicative of overall system performance when using CINEMA 4D,
COMMENT=and do not necessarily reflect the performance of the tested hardware with other applications.
COMMENT=Performance of each component (processor, graphics card) does rely somewhat on other components in the system.
COMMENT=Results provided are typical, although not derived from specific testing procedures.
COMMENT=

 

[8350rocks]

http://www.anandtech.com/bench/CPU/1028

Even if you compare the 6350 and the i3-4360, the 6350 is 3.6 stock, and the i3 is 3.7 stock

FX6350 is 96 in ST, and i3-4360 is 149 in ST.

40% is 134

50% is 144

55% is 148

Hmm...148 vs 149, with intel having a negligible clock advantage, or just about haswell. That is going from PD as well...not Excavator.

So, I would say, haswell has to be *really* close to what the ST performance will be.

EDIT: Just above the i3 is an i7-6850K (BWE) with a ST score of 152 at same clocks as FX6350, which is right in the ballpark. I think they are really close on this one.

The issue is that no benchmark is universal and not even cinebench R15 in the disclaimer say that they measure true IPC but give only a CPU design throughput. IPC and throughput is not the same.

They measure ST throughput though with that aspect of the benchmark. Those are the single thread benchmarks, multi-thread would clearly have the 6850K significantly higher than both the i3 and the FX 6350.

You didn't factor in turbo, the FX6350 runs a turbo around 4.2 so to score at 3.7 it gets 86. The Bulldozer uarch was fundamentally bad for IPC, even a Deneb and Thuban ran circles around a FX8370 clock for clock.

The point of contention is, if AMD means that their CPU can universally at least increase instructions per cycle by 40, that could alone equate to 50% in one bench while 80-90% in another benchmark. You cannot blanket a generic 40% across every bench.

COMMENT=Results Disclaimer - CINEBENCH results are indicative of overall system performance when using CINEMA 4D,
COMMENT=and do not necessarily reflect the performance of the tested hardware with other applications.
COMMENT=Performance of each component (processor, graphics card) does rely somewhat on other components in the system.
COMMENT=Results provided are typical, although not derived from specific testing procedures.
COMMENT=

Then you can deduce that you could also say the same for intel, since their turbo core is superior to the FX line turbo core as well.

It gives a method of comparison, at least.

 

[juanrga]

The main goal of clustering those structures is not to save power .... [but higher clocks]

In this case the throughput is not linear to the size of the window of instructions being scheduled. Distributed schedulers (like in Zen) have better perf/watt and perf/mm, but at cost of generating a less optimal schedule for a single sequence of instructions, which in the end means a lower IPC....

the fact that you can reach higher clock is true but only for multiplication and sqroots. more importantly, the CPU clock will still be limited by the AVX unit, so a gain from this is not very relevant. That is why intel chose to implement negative differential AVX2 clock in BIOS to allow KL to hit 5Ghz.

Not only for multiplication and square roots... Also there is no AVX unit on Zen. Zen has 4 ALUs plus 2x128bit FMAC units implemented via four pipes. The reason why Intel has lower clocks for AVX-mode is because it has very power hungry SIMD units. Haswell/Broadwell have 2x256bit SIMD units or about twice the max. thoughput of a Zen core (32 FLOP vs 16 FLOP). Moreover, this high throughput requires a lot of bandwith from the cache, in fact Haswell doubled the BW compared to Sandy/Ivy

Also as AMD mentioned their goal is power.

AMD also mentioned high-frequency desing. And clustering is a standard technique to achieve higher clocks.

sorry but can any claim that you made in the last para regarding lower IPC be backed by maths? unfortunately there is nothing that can confirm your assertion. why? because it has no logic and there is no scientific paper illustrating your claims, not on sciencedirect anyway

... That is the reason why he thinks that Zen will have less IPC than Haswell/Broadwell and why he thinks that AMD will have higher gains with SMT due to the more distributed nature of the muarch.

even if what he said was right, which I would dispute him directly if he were here, you are doing a cross uarch/vendor comparison, which cannot be done. what I mean is, if there were two versions of zen one with a static another with dynamic scheduler, you can prove/disprove that theory, but not in this case I am afraid. too many other more important things are not the same.

i.e. you cannot make that claim unfortunately.

The principles behind clustering and distributed microarchitectures are well-known, and there are plenty of commercial chips using those. Our opinions about Zen microarchitecture are based in our understanding of such principles and design techniques. Moreover, I finished my post with an important remark:

I agree with him. Of course, only a full review and detailed analysis of the Zen muarch will prove/disprove this.

 

[juanrga]

However, I am getting rumors that Zen could be beyond $700.

3. Ryzen available in first half of March. Only 8C/16T model will be available at launch.

4. The 4C/8T model available one month latter.

5. "Prices are not as cheap as you might think".

I would not consider this rumours all that seriously because
3/4. AMD claimed all will be available at once on several occasion. a bios bug will not prevent them from releasing 4c8t CPU on date with 8c16t.

If I recall correctly AMD only said that different SKUs will be available on launch. No one confirmed that both 8C and 4C will be available at launch. Therefore both claims are compatible.

No one said that a BIOS bug is the reason for the delay of 4C part. What is being said is that a BIOS bug is the reason for this last delay of the launch of Zen, because the 8C Zen CPUs are ready, but mobos aren't.

I don't know why the 4C CPUs would be launched latter, but three different sources are saying the same now. BitsAndChips and CPCHardware also reported the rumor of delay for the 4C CPUs.

I actually think the reason for no 4C cpu at launch is simple- it's a business / marketing decision. I can't remember where I read it now but I'm certain last year there was a quote from either Raja or Lisa Su regarding the fact that AMD were planning to maintain a series of staggered launches throughout the year- I'm guessing in a bid to help bolster the stock price and keep AMD in the news more. It's an approach that appears to be working as I think it's at least one reason why AMD's stock is looking so strong- remember that stock market analysts and many investors have little in depth knowledge of the products. AMD being mentioned on a regular basis is likely to give a better perception "oh look, ANOTHER new AMD product, wow they are innovative"... If there is one thing I've learned keeping an eye on stock prices- the price of a stock is 99% about *perception* of a company. I think this is AMD's way of playing the game a bit more on stocks and the stock prices are looking better for it (obviously the fact they had a good launch with Polaris and the fact Zen looks promising aren't hurting either but I do think in the past they have missed a trick in terms of how the presented / launched their products).

Good point! Often I focus on the technical stuff and forget the economic side of the business.

 

[juanrga]

http://www.anandtech.com/bench/CPU/1028

Even if you compare the 6350 and the i3-4360, the 6350 is 3.6 stock, and the i3 is 3.7 stock

FX6350 is 96 in ST, and i3-4360 is 149 in ST.

40% is 134

50% is 144

55% is 148

Hmm...148 vs 149, with intel having a negligible clock advantage, or just about haswell. That is going from PD as well...not Excavator.

The FX-6350 has a base clock of 3.9GHz, not 3.6. Moreover, this is ST and one has to use turbo clocks instead. The FX-6350 has a single-core turbo of 4.2GHz. This is the reason why it gets the same ST score than the FX-8350 (also 4.2GHz turbo) which appears in the list with 96 points as well. The i3-4360 doesn't have turbo and it runs all the benchmark at 3.7GHz. Correcting for the clocks we obtain

(149 / 96) * (4.2 / 3.7) = 1.76

I.e. Haswell IPC is about 76% above Piledriver, not 55%. The same conclusion can be obtained from the 3GHz data I provided above

(121 / 68.10) = 1.75

One can also see that the presence or absence of L3 doesn't alter the result in any significant way: 76% (using a PD CPU with L3) versus 75% (using a PD APU without L3). This puts to a rest the argument that I was using an incorrect baseline for my early estimations of Zen IPC, when I was using a real Excavator-based Athlon CPU (no L3) as baseline, instead using a non-existent Excavator-based FX CPU with L3.

 

[sarwar_r87]

Not only for multiplication and square roots... Also there is no AVX unit on Zen. Zen has 4 ALUs plus 2x128bit FMAC units implemented via four pipes. The reason why Intel has lower clocks for AVX-mode is because it has very power hungry SIMD units. Haswell/Broadwell have 2x256bit SIMD units or about twice the max. thoughput of a Zen core (32 FLOP vs 16 FLOP). Moreover, this high throughput requires a lot of bandwith from the cache, in fact Haswell doubled the BW compared to Sandy/Ivy

Spoiler

What other instructions other than multi/div/sqroot?

you seemed to have changed the topic there.

Also as AMD mentioned their goal is power.

AMD also mentioned high-frequency desing. And clustering is a standard technique to achieve higher clocks.

"AMD says the much-touted "40 percent higher IPC" over the Excavator core came from three design goals: core, cache, and power. For the core, AMD made everything bigger and wider, introducing a micro-op cache (something Intel has been using for some time), as well as a larger dispatch, larger retire, larger schedulers, and better branch prediction. On the cache side there's a faster prefetch, while L1 and L2 bandwidth has been doubled, and L3 more then quadrupled. Full details on the improvements are in the slides below."
http://arstechnica.com/gadgets/2016/12/amd-zen-performance-details-release-date/

even if what he said was right, which I would dispute him directly if he were here, you are doing a cross uarch/vendor comparison, which cannot be done. what I mean is, if there were two versions of zen one with a static another with dynamic scheduler, you can prove/disprove that theory, but not in this case I am afraid. too many other more important things are not the same.

i.e. you cannot make that claim unfortunately.

The principles behind clustering and distributed microarchitectures are well-known, and there are plenty of commercial chips using those. Our opinions about Zen microarchitecture are based in our understanding of such principles and design techniques. Moreover, I finished my post with an important remark:

I agree with him. Of course, only a full review and detailed analysis of the Zen muarch will prove/disprove this.

sorry? But my point still stands

I.e. Haswell IPC is about 76% above Piledriver, not 55%. The same conclusion can be obtained from the 3GHz data I provided above

yes. but you forgot to add the 10-12% improvement EV has over piledriver.

1.55*1.12=1.74 ~ Haswell

 

[con635]

Hey guys was thinking about an old poster here, hafijur? (banned?) and remembered he used to use the super pi benchmark in examples, I know its obsolete but was wondering does anybody know how zen might perform in it?

 

[de5_Roy]

The FX-6350 has a base clock of 3.9GHz, not 3.6. Moreover, this is ST and one has to use turbo clocks instead. The FX-6350 has a single-core turbo of 4.2GHz. This is the reason why it gets the same ST score than the FX-8350 (also 4.2GHz turbo) which appears in the list with 96 points as well. The i3-4360 doesn't have turbo and it runs all the benchmark at 3.7GHz. Correcting for the clocks we obtain

(149 / 96) * (4.2 / 3.7) = 1.76

I.e. Haswell IPC is about 76% above Piledriver, not 55%. The same conclusion can be obtained from the 3GHz data I provided above

(121 / 68.10) = 1.75

One can also see that the presence or absence of L3 doesn't alter the result in any significant way: 76% (using a PD CPU with L3) versus 75% (using a PD APU without L3). This puts to a rest the argument that I was using an incorrect baseline for my early estimations of Zen IPC, when I was using a real Excavator-based Athlon CPU (no L3) as baseline, instead using a non-existent Excavator-based FX CPU with L3.

zomg my time machine works!

 

[juanrga]

Also as AMD mentioned their goal is power.

AMD also mentioned high-frequency desing. And clustering is a standard technique to achieve higher clocks.

"AMD says the much-touted "40 percent higher IPC" over the Excavator core came from three design goals: core, cache, and power. For the core, AMD made everything bigger and wider, introducing a micro-op cache (something Intel has been using for some time), as well as a larger dispatch, larger retire, larger schedulers, and better branch prediction. On the cache side there's a faster prefetch, while L1 and L2 bandwidth has been doubled, and L3 more then quadrupled. Full details on the improvements are in the slides below."
http://arstechnica.com/gadgets/2016/12/amd-zen-performance-details-release-date/

I know. It doesn't change a bit what we are saying about how distributed schedulers can affect IPC and SMT.

I.e. Haswell IPC is about 76% above Piledriver, not 55%. The same conclusion can be obtained from the 3GHz data I provided above

yes. but you forgot to add the 10-12% improvement EV has over piledriver.

1.55*1.12=1.74 ~ Haswell

No. He was comparing Haswell vs Piledriver, not Haswell vs Excavator. He divided the scores of the FX-6350 and the i3-4360 without considering the clocks and got a 55%

(149 / 96) = 1.55

When we correct for the clocks (the FX-6350 runs that ST bench at 4.2GHz) we obtain 76% instead

(149 / 96) * (4.2 / 3.7) = 1.76

 

[sarwar_r87]

No. He was comparing Haswell vs Piledriver, not Haswell vs Excavator. He divided the scores of the FX-6350 and the i3-4360 without considering the clocks and got a 55%

(149 / 96) = 1.55

When we correct for the clocks (the FX-6350 runs that ST bench at 4.2GHz) we obtain 76% instead

(149 / 96) * (4.2 / 3.7) = 1.76

he mixed up the frequencies. that is besides the point. it does not negate the fact that you have to take PD IPC and calculate the EV IPC for a CPU.

 

[010TheMaster010]

AMD says the much-touted "40 percent higher IPC" over the Excavator core came from three design goals: core, cache, and power. For the core, AMD made everything bigger and wider, introducing a micro-op cache (something Intel has been using for some time), as well as a larger dispatch, larger retire, larger schedulers, and better branch prediction. On the cache side there's a faster prefetch, while L1 and L2 bandwidth has been doubled, and L3 more then quadrupled. Full details on the improvements are in the slides below.

If this was said by AMD then is it to be interpreted as they measured over a never released XV with L3 or as a mere slip of the tongue?
Thanks @Sar for the link
http://arstechnica.com/gadgets/2016/12/amd-zen-performance-details-release-date/

 

[truegenius]

sarinaide said
Ivy/sandy performance would be disastrous to market,. 5 year old and 6 generations behind type performance on 14nm and almost 4x the on die resources would essentially lock AMD into a slow death spiral,

Somehow i doubt that, i mean if they launch a 4c/4t cpu with base clock around i5-3570 under 65w having overclocking capability and at cheaper price than KL i3 or atleast around chepest KL i3 then it will be great, there lots of amd owners looking for upgrade and lots of people who can only afford i3 thus settle for it.
And if they price 4c/8t at below highest KL i3 or below cheapest kl i5 then and clock it around i7-3770 then it will be like having faster cpu than 3770 at price between I3-I5 , and not to forget, if they clock it around 3770 then it will be around fx9570 in multithreading and that too at 65w which is a huge leap.
And a 6c/6t version priced below highest KL i5 or below lowest kl i7 and clocked between 3.5-4 ghz then it will be great upgrade for i5 sb/ib owner because games are using more than 4 cores now
To tackle i7 they need 6c/12t or 8c/8t cpu priced around top mainstream i7 and clocked between 3.5-4
I mean to say market is wide open, intel left so much gap and amd needs to price their products right to score win in consumer market.
Afterall real thing of amd is not raw performance its aggressive pricing.
Btw zen looks promising for server market and imo will give them huge gains from server.

jdwii said
Seems late to still be in that stage we are only a month away(i mean for tweaks). Can't wait to see actual numbers on these things as well as price i still think the 4 core part will be priced between a I3-I5 and i do still think we will see 2 8 core Ryzens one at 380$ or so and maybe one with higher clocks at 500$. If we see a 6 core part that would be nice and hopefully its around a 280-340$ price.

4c priced between sl/kl i3 and i5 will be disastrous
They are surely behing kl ipc and pricing them between I3-I5 will mean make intel a better option over amd.
Doing equal to competition and pricing around competition will result in disaster, that will be overconfidence.

Imo pricing should be like this for amd to regain consumer confidence and consumer market share ( all intel cpu are kl )
Slowest 4c/4t cpu - highest pentium with htt
Highest 4c/4(8)t apu - highest i3 or lowest i5
Highest 4c/4t cpu - below slowest i3
Highest 4c/4t cpu - highest i3 or slowest i5
Highest 6c/6t cpu - highest i5 or slowest i7
Highest 6c/12t or 8c/8t - highest mainstream i7
Highest 8c/16t - above highest mainstream i7 and around 50% of comparable 8c/16t i7 extreme

 

[sarwar_r87]

AMD says the much-touted "40 percent higher IPC" over the Excavator core came from three design goals: core, cache, and power. For the core, AMD made everything bigger and wider, introducing a micro-op cache (something Intel has been using for some time), as well as a larger dispatch, larger retire, larger schedulers, and better branch prediction. On the cache side there's a faster prefetch, while L1 and L2 bandwidth has been doubled, and L3 more then quadrupled. Full details on the improvements are in the slides below.

If this was said by AMD then is it to be interpreted as they measured over a never released XV with L3 or as a mere slip of the tongue?
Thanks @Sar for the link
http://arstechnica.com/gadgets/2016/12/amd-zen-performance-details-release-date/

o. nice catch. I overlooked that part.

As I said before, it is counter productive for AMD' s PR to compare their product with subpar products from themselves. it will be a disaster once the benchmarks indicates they were being oversmart/misleading.

also if you consider the fact that they compared with a XV with L3 and do all math, a 40% increase like they claim meets the expectation of an Zen 8c16t @ 3.4 Ghz to match 6800k without having to play with SMT scaling better with no a shred of evidence. Finally, this is in line with the latencies and throughput comparison between zen, KL, and BR that CPC published

but I am sure there will be some who will still present alternate facts. hehe

 

[sarwar_r87]

Somehow i doubt that, i mean if they launch a 4c/4t cpu with base clock around i5-3570 under 65w having overclocking capability and at cheaper price than KL i3 or atleast around chepest KL i3 then it will be great, there lots of amd owners looking for upgrade and lots of people who can only afford i3 thus settle for it.

while I agree that performance/price will be a big deciding factor, we have to remember that vast majority of consumer does not really read reviews but goes on what he said and she said. if the consensus is that intel is better than AMD, it will hard for AMD to get back the market share, at least in desktop segment.

also it will have to perform too. because right now, you can get a 8 core for the same price as an i3. but you will hardly see anyone recommending a FX 8 core over i3. an IPC of SB will foot the bill, but will not help much as they will still be half a decade behind intel.

it all comes down to perception. amd should not look to replace IPC with more core. They have tried it and failed. they must have comparable IPC and make up any deficit using higher clocks or even higher TDP (I really do not understand why ppl complain about a 125W CPU when a hairdryer consumes 1000W)

 

[truegenius]

8350rocks said
How do you think that excavator is 40% behind Sandy Bridge when piledriver was only 25% behind Sandy Bridge in single thread?

Here you are looking at different approach mathwise
That bench shows i7 as 100% and fx8350 as75% but we are taking fx8350 as 100%
So we convert that bench to fx@100% then i7 will be
(100/75)x(4.2/3.8) ≈ 150 means i7 is 50% better

Eg. A = 10, B=20, c=30
If we take B as base 100% then A will be 50% and C will be 150%
Then we can look at it in 2 way
A is 50% of B ( 50% slower than B)
Or B is 2x or 200% of A ( 100% faster than A)

This shows a 2600K @ 4.6 and the 8350 @ 4.6.

The 2600K is a 9.05 and the 8350 is a 7.56

That is a 17% gap...which is actually even smaller than the other link claims.

You are comparing whole cpu means throughout, which is not the whole game, fx8350 have 50% less single thread ipc thus it looses in single thread tasks and games
Just recently techspot re-reviewed fx9570 and it lost to i3-6100 in gta5 with a huge margin which makes it clear that no matter how multicore friendly our game is, single thread performance still matters.

 

[sarwar_r87]

AMD says the much-touted "40 percent higher IPC" over the Excavator core came from three design goals: core, cache, and power. For the core, AMD made everything bigger and wider, introducing a micro-op cache (something Intel has been using for some time), as well as a larger dispatch, larger retire, larger schedulers, and better branch prediction. On the cache side there's a faster prefetch, while L1 and L2 bandwidth has been doubled, and L3 more then quadrupled. Full details on the improvements are in the slides below.

If this was said by AMD then is it to be interpreted as they measured over a never released XV with L3 or as a mere slip of the tongue?
Thanks @Sar for the link
http://arstechnica.com/gadgets/2016/12/amd-zen-performance-details-release-date/

o. nice catch. I overlooked that part.

As I said before, it is counter productive for AMD' s PR to compare their product with subpar products from themselves. it will be a disaster once the benchmarks indicates they were being oversmart/misleading.

also if you consider the fact that they compared with a XV with L3 and do all math, a 40% increase like they claim meets the expectation of an Zen 8c16t @ 3.4 Ghz to match 6800k without having to play with SMT scaling better with no a shred of evidence. Finally, this is in line with the latencies and throughput comparison between zen, KL, and BR that CPC published

but I am sure there will be some who will still present alternate facts. hehe

UPDATE:
https://www.youtube.com/watch?v=sT1fEohOOQ0

The AMD Zen Technical Briefing @ Hot Chips 2016

7:48--- Our L3 bandwidth has actually gone 5x compared to previous generation.

so AMD Senior Fellow and Design Engineer, Michael Clark did say it.

Slide also states "Faster L3 cache". XV has no L3, so obviously they measured over a never released XV with L3. its on the slide, so it cannot just be a slip of tongue. and one simply cannot make L3 faster than previous gen, if it did not have any.

so how do they know its 40% faster. I think it can be pointed out from theoretical study which I am sure they have or from engineering samples of XV with L3 which they may have a few for testing purposes. companies do make a lot of prototypes of their system.

Thanks 010TheMaster010 for pointing it out.

 

[truegenius]

while I agree that performance/price will be a big deciding factor, we have to remember that vast majority of consumer does not really read reviews but goes on what he said and she said. if the consensus is that intel is better than AMD, it will hard for AMD to get back the market share, at least in desktop segment.

And thats the reason i think for launching 8c/16t first followed by other sku
Because reviewers will have only 8c/16t cpu to pick up and that means only competition from intel will be 8c i7 and they will compare them and will conclude "amd matched intel's i7 extreme at lower power and price" thus causing good first impression in consumer's mind.

If they launch 4c/4(8)t along with 8 then we know that reviewers will compare it to kl i5/7 and amd will loose without any doubt and reviewers will say " amd released another major disappointment, comparable to only several gen old intel cpu" thus causing bad first impression ( while we know intel haven't improved much over several generations but reviewers will not mis any opportunity for catchy heading)

 

[sarwar_r87]

And thats the reason i think for launching 8c/16t first followed by other sku
Because reviewers will have only 8c/16t cpu to pick up and that means only competition from intel will be 8c i7 and they will compare them and will conclude "amd matched intel's i7 extreme at lower power and price" thus causing good first impression in consumer's mind.

If they launch 4c/4(8)t along with 8 then we know that reviewers will compare it to kl i5/7 and amd will loose without any doubt and reviewers will say " amd released another major disappointment, comparable to only several gen old intel cpu" thus causing bad first impression ( while we know intel haven't improved much over several generations but reviewers will not mis any opportunity for catchy heading)

yes. but I am sure most reviewers, if not all, will disable 4 cores to get to 4c8t and benchmark it with i7 4c8t CPUs. so I do not think it will work. It might be more benifitial to go all out to show that they have high confidence in their new products.

 

[sarinaide]

Performing in games like i5, consuming about same power than 140W chips, and overclocking below 4.5GHz is the new definition of "looks better"...

It is just a response to Juan's post on ExtremeTech and on his blog.

The i7 5960X which consumes even more power cannot overcome an i5 6500 or 6600 either which is mostly down to clockspeed. Per Dresdenboy calculating the performance gained running the Ryzen at 3.4Ghz it would more or less match the 6900K 105%/107%

The i7 5960X would score exactly the same gaming result per CPC Hardware around 98-100% at it's given clockspeed.

Similarly Intel's $1800 6950X gets pounded in gaming by a $399 part, does that make its 200W+ usage less impressive?

You have reiterated it yourself on many occassions, the SR7 is not really targeted towards gamers, it is targeted to professionals not in need of X99 or dual socket Xeons to get top end performance.

So yes it is a massive step up given that Bulldozer was worse than the uArch it replaced.

 

[juanrga]

Performing in games like i5, consuming about same power than 140W chips, and overclocking below 4.5GHz is the new definition of "looks better"...

It is just a response to Juan's post on ExtremeTech and on his blog.

On my blog? I didn't write anything about Zen.

http://www.juanrga.com/p/archive.html

Note: Don't read so much on the article about heat. There are a pair of mistakes and I am going to update it soon.

 

[sarinaide]

Another issue is the selection of benchmarks for validation. Cinebench is no performance dependent as Blender or Handbrake and to claim that Blender and Handbrake are GPU affected is incorrect as the numbers would have been universally different for every person running the blender demo. It was set to CPU level performance and Blender is no less threaded than Cinebench.

IPC is not universal, it varies from application to application. IPC is architectural and is the product of single thread and multi threaded performance as both are not mutually separable.

Let me demonstrate this using a i7 5960X (Haswell) and i7 4770K (Haswell)

i7 4770K

Turbo clock - 156/791 ratio 5.07
@2.9Ghz - 116/588.178 ratio 5.07
@3.0Ghz - 120/608.461 ratio 5.07
@3.1Ghz - 124/628.743 ratio 5.07

Single = 4 per 100mhz
Multi = (4x5.07) = 20.28 per 100mhz

The degree of SMT or put otherwise the extent at which threading helps the i7 4770K is 23.5943%

i7 5960X
@Turbo - 140/1337 ratio 9.55
@2.9 - 116/1107.80 ratio 9.55
@3.0 - 120/1146 ratio 9.55
@3.1 - 124/1184.20 ratio 9.55

ST = 4/100mhz
MT= 38.2/100mhz

SMT performance is only 17.6638% at highly threaded applications.

Despite the greater output in multithreaded situations the 4770K is still more efficient as the more threads you add the lower the efficiency. This is why I dispute any claim that AMD's SMT is 40% efficient as the cost in Single thread would be severe.

Zen ES ~ 5960X

 

[salgado18]

also it will have to perform too. because right now, you can get a 8 core for the same price as an i3. but you will hardly see anyone recommending a FX 8 core over i3. an IPC of SB will foot the bill, but will not help much as they will still be half a decade behind intel.

I do, sort of. Went from an FX 8120 to an i3-6100, and while the i3 is a lot faster in most stuff, it can't handle multitasking, to the point where there's stuttering with the mouse and sound on slightly heavy tasks. Anyone asking me today, I'd say: if you can't afford an i5, get an FX-6xxx. The system will be a lot smoother, games will run good, and the general experience will be better.

However, you are right in saying the issue is perception. Consumers prefer an i3 over any AMD. The job now is to change this perception, by pleasing enthusiasts first (they recommend hardware to people around them), by clever marketing (people must see AMD as a brand they can trust), and by good partnerships (like notebook builders).

I believe they can do it with IPC around Ivy Bridge-Haswell, because that's good enough, and they can play the value game again, but with a lot more margin this time. And if enthusiasts perceive value, they will sell value to their friends (who don't care about top performance, but value).