Full Skylake reveal result? Waiting for Zen.

[Abwx]

50% in 10 years? Skylake vs Core M/P4/K8?

Most of the improvement was made in FP, for most integer based code it must be much less than this and higher perfs has been possible only due to multithreading, look at the 7zip archiver for instance and compare with Cinebench wich is FP, the two graphs below are for ST and MT.

http://www.hardware.fr/articles/921-7/vs-core-2-duo-e6600-core-2-quad-q6600.html

 

[BurnItDwn]

50% in 10 years? Skylake vs Core M/P4/K8?

Conroe was 2006, so not quite 10 years, but, that was the last real major significant leap in single threaded performance.

It was also the first time Intel really bettered AMD performance since AMD first released the Athlon 500.

P4 is not worth mentioning, netburst architecture is bad as a whole. (unless you mean the M which is based on core)

My point is, in the 90s, and early 00s, every 18 months you got HUGE booms in performance.

Things have been almost stagnant since the first core 2 duo when it comes to single threaded peformance. You are lucky to get 10%.

Intel has focused their efforts on performance per watt, and cost, rather than single threaded raw performance.

 

[ShintaiDK]

Conroe was 2006, so not quite 10 years, but, that was the last real major significant leap in single threaded performance.

It was also the first time Intel really bettered AMD performance since AMD first released the Athlon 500.

P4 is not worth mentioning, netburst architecture is bad as a whole. (unless you mean the M which is based on core)

My point is, in the 90s, and early 00s, every 18 months you got HUGE booms in performance.

Things have been almost stagnant since the first core 2 duo when it comes to single threaded peformance. You are lucky to get 10%.

Intel has focused their efforts on performance per watt, and cost, rather than single threaded raw performance.

It is still way over 100% of Core 2 in ST. Close to 200%.

 

[Ma_Deuce]

prices for the i3 didn't,and not only that you got a improved version on top...for the same price.

You missed the point. Prices on everything go up all the time. Inflation, higher production costs, whatever. 20oz sodas are twice as much as they were ten years ago and they didn't even do a node shrink . The fact that they haven't jacked up the prices on their cpus tells me that there is some kind of downward pressure. I'm guessing the stagnant PC market has more of an impact than AMD prices have on them though.

I don't think intel will need to take much notice of Zen right away either. Even if Zen is everything AMD hopes for in performance, they need to establish their rapport with all of the vendors again. It's an uphill struggle to get their products in decent machines anymore it seems.

 

[TheELF]

Most of the improvement was made in FP, for most integer based code it must be much less than this and higher perfs has been possible only due to multithreading, look at the 7zip archiver for instance and compare with Cinebench wich is FP, the two graphs below are for ST and MT.

http://www.hardware.fr/articles/921-7/vs-core-2-duo-e6600-core-2-quad-q6600.html

Sure, you don't get any improvement at all when going from running only one thread to running two threads...

Somebody screwed up.

 

[alexruiz]

I'm not so sure about this. Here is a timeline of AMD processes vs Intel processes:

AMD:
350nm: June 1996 (K5)
250nm: May 1998 (K6-2)
180nm: November 1999 (Athlon Orion)
130nm: June 2002 (Athlon XP Thoroughbred)
90nm: October 2004 (Athlon 64 Winchester)
65nm: February 2007 (Athlon 64 Lima)
45nm: February 2009 (Phenom II Deneb)
32nm: September 2011 (Llano)
28nm: January 2014 (Kaveri)
20nm: ??
"14"nm: 2016 (Zen)

Intel:
350nm: November 1995 (Pentium Pro)
250nm: January 1998 (Pentium II Deschutes)
180nm: October 1999 (Pentium III Coppermine)
130nm: July 2001 (Pentium III Tualatin)
90nm: March 2004 (Pentium IV Prescott)
65nm: January 2006 (Pentium IV Cedar Mill)
45nm: January 2008 (Core 2 Wolfdale)
32nm: January 2010 (Clarkdale)
22nm: April 2012 (Ivy Bridge)
14nm: September 2014 (Broadwell)
10nm: 2017 (Cannonlake)

So I think the scenario is more the opposite of what you're saying. In 1995 there wasn't much gap and at times they were neck and neck. Now Zen will be 2 years behind in a process that's the same in name only, in reality being in between Intel's 22nm and 14nm. So more like 3 years behind.

Great info.
I was going to point out that at 130nm, AMD was arguably in the lead, as they had 130nm SOI. Fab 30 in Dresden was also considered one of the most advanced fabs in the world... yes the memories.

On the original subject, I am also on wait and see.

And for those of you claiming that an i3 is better than a Kaveri APU, you obviously have not used BOTH in comparable hardware. The most typical use for those processors would be a browser with 50+ tabs open facebook, twitter and flash rich sites, a word document open, some music in the background and maybe even a movie streamed. See for yourself which one feels smoother, but I guess that because no tech site reviews real world usage then the synthetics numbers win...

ps. Okay, maybe the 50+ tabs is exaggerated, but the point sticks. And yes, I have used both on comparable hardware for those usages.

 

[ShintaiDK]

Great info.
I was going to point out that at 130nm, AMD was arguably in the lead, as they had 130nm SOI. Fab 30 in Dresden was also considered one of the most advanced fabs in the world... yes the memories.

On the original subject, I am also on wait and see.

And for those of you claiming that an i3 is better than a Kaveri APU, you obviously have not used BOTH in comparable hardware. The most typical use for those processors would be a browser with 50+ tabs open facebook, twitter and flash rich sites, a word document open, some music in the background and maybe even a movie streamed. See for yourself which one feels smoother, but I guess that because no tech site reviews real world usage then the synthetics numbers win...

ps. Okay, maybe the 50+ tabs is exaggerated, but the point sticks. And yes, I have used both on comparable hardware for those usages.

Why do these claims never match what people and OEMs buy? And what sites report?

 

[BigDaveX]

Great info.
I was going to point out that at 130nm, AMD was arguably in the lead, as they had 130nm SOI.

That didn't show up until Athlon 64, over a year after they moved to 130nm, and two years after Intel's first 130nm products arrived on the scene. Thoroughbred and Barton were both done on a standard 130nm process; the latter was actually meant to be the point where they moved to SOI, but it was rejigged into a Thoroughbred-B with a larger L2 cache.

If anything, the following node was the one where they really outdid Intel, whose 90nm process turned out to be a lemon.

 

[Abwx]

Sure, you don't get any improvement at all when going from running only one thread to running two threads...

Somebody screwed up.

Yeah, someone who has trouble reading what he s quoting..

Most of the improvement was made in FP, for most integer based code it must be much less than this and higher perfs has been possible only due to multithreading, look at the 7zip archiver for instance and compare with Cinebench wich is FP, the two graphs below are for ST and MT.

http://www.hardware.fr/articles/921-7/vs-core-2-duo-e6600-core-2-quad-q6600.html

Circa 2001 i did a build with a cheap Duron 700 wich did cost about 90€, a next build was for a friend circa 2009 with an Athlon II X4 2.6 wich did cost the same 90€.

In a matter of 8 years that was 20x the processing power for about the same money, 100% sure that in two years the 90€ CPU wont provide 20x the throughput of said Athlon X4..

 

[jpiniero]

For me, my next build is going to be:
-AMD Zen 8C/16HT. No brainer here.
-16nm LiquidVR-compatible GPU. LiquidVR since that's the current king for VR and Arctic Islands will be well optimized in all VR titles by that point. Thanks to the pieces already being in place today.

Good luck with that. Bonus, you are going to be very disappointed with 16 nm GPUs.

I don't think intel will need to take much notice of Zen right away either. Even if Zen is everything AMD hopes for in performance, they need to establish their rapport with all of the vendors again. It's an uphill struggle to get their products in decent machines anymore it seems.

Without an IGP AMD has 0% chance of getting any OEM deals with Zen.

 

[BurnItDwn]

It is still way over 100% of Core 2 in ST. Close to 200%.

OK yea, it seems your right ...

things slowed down after 2006 but, the stagnation really has only been since 2011 ...

2006 to 20011 we went from core 2 duo to sandy bridge. And essentially, CPU single threaded performance went up by like < 10% annually on average.

From 2011 to 2015... we have gained less than 5% performance annually on average.

 

[StrangerGuy]

Great info.
And for those of you claiming that an i3 is better than a Kaveri APU, you obviously have not used BOTH in comparable hardware. The most typical use for those processors would be a browser with 50+ tabs open facebook, twitter and flash rich sites, a word document open, some music in the background and maybe even a movie streamed. See for yourself which one feels smoother, but I guess that because no tech site reviews real world usage then the synthetics numbers win...

ps. Okay, maybe the 50+ tabs is exaggerated, but the point sticks. And yes, I have used both on comparable hardware for those usages.

Next time I'm just going buy hardware based on personal anecdotes instead of objective benchmarks.

 

[PliotronX]

You missed the point. Prices on everything go up all the time. Inflation, higher production costs, whatever. 20oz sodas are twice as much as they were ten years ago and they didn't even do a node shrink . The fact that they haven't jacked up the prices on their cpus tells me that there is some kind of downward pressure. I'm guessing the stagnant PC market has more of an impact than AMD prices have on them though.

I don't think intel will need to take much notice of Zen right away either. Even if Zen is everything AMD hopes for in performance, they need to establish their rapport with all of the vendors again. It's an uphill struggle to get their products in decent machines anymore it seems.

Methinks AMD actually has potential for rapid adoption from vendors. Most of the sub-500 dollar notebooks tend to be AMD driven already and that's the range I ever hear anyone asking about, even filthy rich executives. It seems like they won all of the consoles without even trying and it wasn't even a midrange processor design.

 

[nismotigerwvu]

OK yea, it seems your right ...

things slowed down after 2006 but, the stagnation really has only been since 2011 ...

2006 to 20011 we went from core 2 duo to sandy bridge. And essentially, CPU single threaded performance went up by like < 10% annually on average.

From 2011 to 2015... we have gained less than 5% performance annually on average.

Funny what competition will do for the market huh? A reasonably spry Phenom II kept Intel pushing, but once Bulldozer comes to market the improvements slow down. Even if you're someone that would never consider buying anything from AMD, you would still benefit from Intel being pushed. Look at what the release of the Phenom II did for i7 prices before and after launch.

 

[cytg111]

Funny what competition will do for the market huh? A reasonably spry Phenom II kept Intel pushing, but once Bulldozer comes to market the improvements slow down. Even if you're someone that would never consider buying anything from AMD, you would still benefit from Intel being pushed. Look at what the release of the Phenom II did for i7 prices before and after launch.

I think the real enemy here is/was the iPhone... and the whole slew of mobile flap flap that rode in on that wave .. (We might take some of that territory back with the upcoming dual-4K 120hz VR revolution)

 

[BurnItDwn]

Funny what competition will do for the market huh? A reasonably spry Phenom II kept Intel pushing, but once Bulldozer comes to market the improvements slow down. Even if you're someone that would never consider buying anything from AMD, you would still benefit from Intel being pushed. Look at what the release of the Phenom II did for i7 prices before and after launch.

Well, since I joined the work force when I was 16 and started building PC's a few months later after I had earned enough $$$. That was 1996 and I built my first machine on a very tight budget, and had an awful CPU (cyrix). Since then, I have run K6, K62, Athlon (both slot and socket), Duron, several generations of Athlon XP and mobile, Athlon 64, A64 x2, Phenom, Phenom II .... (I never ran a k5 or k63 though)

On the Intel front, I had a celeron 300a on my BH6 which I then later upgraded to a 533 coppermine celeron. I had a cheapo dell server which I ran as a dedicated IP Masquerade/home file server/ventrilo server/dedicated game server on and off for like 10 years, it started with a Celeron 1.8, but I replaced that with a $10 p4. I disliked netburst so skipped it as A64 was so awesome at the time. I bought a sandy bridge 2400 and run it to this day in my triple headed box.


Anyhow, I was a big big fanboy for a long time, I even bought about 400 shares of AMD stock when It was "down" to like $3 or so.

I should have sold it when it was back up around $4, but I figured when they got their console deals that they would manage to pull out of their nosedive ....

Ohh well, they look not so good now, probably should dump it and use the money at the local mexican eatery but instead I will hold it until they go defunct, or until Zen comes out and either is a success or they go defunct.

I am hopeful, but not optimistic.

 

[Tuna-Fish]

Without an IGP AMD has 0% chance of getting any OEM deals with Zen.

AMD appears to be initially targeting primarily servers. I sure hope their chips will be good enough for that...

 

[frozentundra123456]

Personally, I think Zen for consumers without any igp is because they simply did not have time/resources to develop and validate one (an igp), or they are waiting for HBM2.

But I agree, Zen better do well in servers, because despite all the salivating by AMD fans on these forums, it will have an *extremely* limited market in the consumer segment without an igp.

 

[AtenRa]

Personally, I think Zen for consumers without any igp is because they simply did not have time/resources to develop and validate one (an igp), or they are waiting for HBM2.

But I agree, Zen better do well in servers, because despite all the salivating by AMD fans on these forums, it will have an *extremely* limited market in the consumer segment without an igp.

I do agree that Servers are the main target for ZEN but in late 2016 the PC Gaming field may be completely different than today. DX-12 games will start releasing from H1 2016 and new 16nm FF Graphics Cards with HBM 2.0 in early H2 2016 could make current Quad Core 8 Threads CPUs look obsolete.
And here comes Socket 2011-v3 and ZEN multicore CPUs as a savior to Gamers.

 

[frozentundra123456]

Always something just over the horizon isnt there? And how much are they going to sell this magical cpu for? Even if it is the bee's knees for gaming, in the overall scheme of things that is a very limited market, just as I said. And I hardly think devs are going to make DX12 games that will run poorly on existing systems, over 90% of which I am sure are quad core or dual core.

 

[AtenRa]

Technology moves forward, if you want to play Ashes Of Singularity in late 2016 with 1000s of units you will need more than a quad CPU. That doesnt mean the game will not be playable with a quad Core with lower unit count.

Its like the GPUs, vast majority of Gamers have $100-200 GPUs, and yet games can make a $650 GTX980Ti crawl to its knees. I havent seen anyone implying that Game devs will not make games more GPU bound because the majority of Gamers have $100-200 GPUs.

 

[ShintaiDK]

Ohh, we still have to see on AOTS

For the benchmark a quad is more than plenty.

The octo dream will continue to be so. Just like it did since 2011. The next thing have been the excuse ever since.

 

[tential]

Also, this is assuming AOTS won't improve their game....

 

[VR Enthusiast]

People don't seem to realise what even very small (5%) gains will main for AMD in server and desktop. They're doing the right thing with Zen.

What do they need to achieve in order to make a billion in server and high-end desktop revenue? $500 for an octo-core with HT?

AMD can't fail with Zen simply because Bulldozer has already killed any market share they had. They only need to be close enough that Intel can't beat them with half the cores and if they can't do that then they might as well just pull the plug.

 

[Arachnotronic]

People don't seem to realise what even very small (5%) gains will main for AMD in server and desktop. They're doing the right thing with Zen.

What do they need to achieve in order to make a billion in server and high-end desktop revenue? $500 for an octo-core with HT?

AMD can't fail with Zen simply because Bulldozer has already killed any market share they had. They only need to be close enough that Intel can't beat them with half the cores and if they can't do that then they might as well just pull the plug.

Yeah, it's going to be that easy.