I never did like the words "up to" it's very fishy.

32 cores even on 14nm is a tall order, it'll be clocked awfully low if they want to keep a 140W TDP. Not that they're afraid of a higher TDP..

If a consumer Zen 16 core releases we'll either have a low clocked 16 core or another high TDP AMD CPU, upwards of 220W I expect.

Obviously this is all speculation and depends on if we believe that AMD's 40% ipc improvement is fact.

40% sounds exactly like what you would expect if you turned the bulldozer cores into proper cores.

newtekie1I haven't seen a Displayport to HDMI2.0 adapter yet that was cheaper than $30.

Just look at the latest 980Ti Matrix benchmark here on TPU. It beats the Fury X by 17% at 4K. Sure, HBM provides more memory bandwidth than GDDR5, but when the overall card is still slower what's the point?

I get a nice discount the markup on adapters is ridiculous.

Parn40% IPC improvement over A10-7870K. Hope this is true as it will definitely bring Zen on par with Haswell at minimum.

Entirely disagree: a 40% IPC boost means that the "next-gen" Zen will be still underperforming in both single- and multi-threaded tasks when compared to an out-of-date Haswell.

In other words, and much to my disappointment because I don't fancy a world where Intel is king of the hill, AMD's best effort will still trail Haswell in IPC (by quite a lot) AND Skylake (+Skylake refresh) when Zen arrives.

This will be a very poor showing for AMD. In fact it will be the last nail in the coffin for a company that has done its utmost to drive itself into the ground.

I am not an AMD or Intel fan boy, I am pro-market, and FWIW, AMD has been run by people who - to ensure that they aren't kicked out of office - have cared only about the stock value in the short term, and failed to see what the company would be selling 1-, 3- even 5- years ahead. The only thing AMD management has been able to do so far is shore up its defences for its own demise. Even the creation of the Radeon Technologies Group is a way for management to secure some form of revenue in case the company goes tits-up. When things get worse, AMD has packaged all its worthwhile assets in a nice little parcel with a perky little bow on top, ready to be sold off to the highest bidder. I am sure this is a condition the stockholders demanded be met to ensure support.

When Zen comes out (and fails) it will be a short hop to the "we are now a fabless graphics technology semiconductor design company" announcement, "available to license our graphics technology to anyone who wishes to acquire a license", who will adopt an ARM-like strategy towards the market. AMD will be no more and RTG will be its spiritual successor. Heads will roll, golden umbrellas will be paid, Chapter 13 will be filed, AMD restructured, x86/AMD64 people laid off and only RTG will remain, which will then be sold off.

I'm just speculating here, but Zen would've been a hit... like in mid-2015. Not in early 2017.

JermelescuFor the sake of competition I hope Mr. Keller delivered magic.

He's a very talented man and I'm sure he did the best job he could but AMD just doesn't have the R&D budget that Intel has. That is why I'm skeptical that Zen will be good competition for Intel chips. We may never know what was said between Keller and Lisa Su but I suspect that once Keller went back to work for AMD he saw that they had no real focus on what would come after Zen and that they basically are just trying to survive until 2019 when one of their big debts comes due. Keller had no reason to remain with AMD if they aren't planning something for after Zen.

AMD faces difficulties even if Zen is good competition for Kaby Lake and Cannonlake. They will probably have to sell the Zen chips to computer manufacturers for cheap prices just to get them to use AMD chips and their strategy of selling cheap just puts them further in the red. Hell, they have the console market all to themselves and they posted their largest losses to date in 2015. I don't know what they are making per console for their chips but I suspect they aren't making much.

LemmingOverlordEntirely disagree: a 40% IPC boost means that the "next-gen" Zen will be still underperforming in both single- and multi-threaded tasks when compared to an out-of-date Haswell.

In other words, and much to my disappointment because I don't fancy a world where Intel is king of the hill, AMD's best effort will still trail Haswell in IPC (by quite a lot) AND Skylake (+Skylake refresh) when Zen arrives.

Underperforming compared to haswell in single threaded, yes probably still a little. However, the multi-threaded tasks should beat haswell without a problem. The 8-Core AMD parts are already within spitting distance of a 4790K in multi-threading when both are clocked the same. My multi-threading scores on a 4.6GHz FX-8350 are only about 10% behind my scores on a i7-4970k. So if they bring the single threaded performance up 40%, the multi-threaded will easily beat a 4790K and likely a 6700k as well.

AMD can’t compete with Intel when it comes to CPU reliant tasks, but on average AMD Kaveri APU’s outperform Intel’s latest Skylake processors when it comes to gaming (without dedicated graphics).

Admittedly performance isn’t ground-breaking, but from my own personal tests the A10-7800 can average around 37fps in Battlefield 4 at 1680x900 on medium settings, compared to the similarly priced Intel Core i3-6100 which gets around 24fps on the same settings. In Grid 2 the A10-7800 achieves 35fps at 1920x1080 on high settings, compared to 18fps for the i3-6100. But games which are more CPU dependant such as Far Cry 3 the gap is smaller with the A10-7800 hitting 20fps at 1680x900 on medium settings and the i3-6100 averaging at 18fps.

So depending on what type of system you are trying to build the AMD Kaveri chips can offer good value.

If AMD can improve on Kaveri IPC by 40% and also further improve on the integrated Radeon graphics, plus add in extra features such as H.265 decoding, then AMD could have the first APU that can be used for more than just casual gaming, allowing you to build a very small system that can do a bit of everything.

I think a well-balanced APU with ‘good enough’ processing and ‘good enough’ graphics at a low price point will appeal to a large market, not everyone is interested in absolute performance and that final 10% or 20% of performance for much higher cost, but instead are interested in smaller computers with a good media/gaming experience at low to medium price points.

Caring1AMD still playing catch up.
A 40% improvement in IPC will still leave them behind Intel.

Competitive mid range is all we need.
AMD was stuck on 28nm for too long.

Kurt MaverickNvidia has slightly superior but generally fair prices to their producs, it's mainly Intel the ones that are being absolutely abusive and inmobile about them.

Let's compare 960 vs 380, shall we?
amazon.de

960 is 230Euro-ish
380 is slightly below that

For 10% faster, 20% more power hungry chip. (that not even taking into account "TDP target" feature of AMD, that allows you to downscale chip a bit for not so demanding games)

Maybe high range-ish?
980 vs Fury Nano

On par power consumption, faster yet cheaper Nano with unique features.

What is fair about that?

newtekie1It isn't like AMD doesn't jack up their prices when they can. It feels like everyone forgets the $1,000 FX-57 and FX-60. When AMD has the lead, they are real quick to jack up those prices.

AMD Athlon 64 FX-57: The Fastest Single Core
www.anandtech.com/show/1722/2

PS
Remind me Intel EE costs pretty please.

newtekie1Intel's $1,000 parts are very top end parts too. The point is, as soon as AMD can, it will jack up its prices. They aren't selling cheap CPUs because they want to.

Back in 2004-5 I've upgraded to Athlon64 X2, new mobo, new mem, new CPU. All under 220$.
So, nope.

Kurt MaverickNvidia has beaten AMD in almost every DX11 game out there.

What, on earth, are you talking about?

newtekie1Underperforming compared to haswell in single threaded, yes probably still a little. However, the multi-threaded tasks should beat haswell without a problem. The 8-Core AMD parts are already within spitting distance of a 4790K in multi-threading when both are clocked the same. My multi-threading scores on a 4.6GHz FX-8350 are only about 10% behind my scores on a i7-4970k. So if they bring the single threaded performance up 40%, the multi-threaded will easily beat a 4790K and likely a 6700k as well.

I won't lie to you, I haven't benchmarked an FX-8350, much less an overclocked one, but I've benchmarked other CPUs of the same architecture and extrapolating results (to me), and checking on benchmark sites, there is no chance the 8350 comes within that 10% window - clock for clock in multithreading - unless you are quoting one very specific benchmark (that's the problem with benchmarks, right?). It does come down to what benchmark you are running and whether the system is somehow bottlenecked by something... i.e.: do you have to flush something to disk? (from your sig, you have overclocked your FX-8350 and it's got 32GB of RAM which might play a role in this). Clock for clock, single or multithreading, Haswell beats Piledriver...

ALSO: upping IPC by 40% does definitely not mean they are increasing single-threaded performance by the same ratio. In order to get IPC up 40% they will surely rein in the CPU clock. AMD is doing to its processors what Intel did when it about-faced on Netburst: it's improving execution (reducing pipeline length, branch prediction, etc... we'll see when it surfaces).

ALSO: Intel is not frozen in time, right...? It'll be a little over 2.5 years between the launch of Zen and Haswell (if Zen is on time)... Concern yourself with the fact that AMD has not been making any progress in CPUs for the past couple of years, while Intel integrated graphics have been steadily improving, generation after generation.

Don't get me wrong: I want Zen to succeed, but AMD has so far given me nothing but reasons to doubt their promises/expectations. They have systematically fallen short on delivering the gains they promise, time and again.

If things go pear-shaped, CES 2017 will be a very - believe me - VERY public place to crash and burn. Hence my "doom-and-gloom prediction".

There should be a betting pool for this... Anyone?

Here is a link to a site with full sets of benchmarks on both the i7 4790K and the FX-8350

medi01AMD Athlon 64 FX-57: The Fastest Single Core
www.anandtech.com/show/1722/2

PS
Remind me Intel EE costs pretty please.

Right at about $1,000.

medi01Back in 2004-5 I've upgraded to Athlon64 X2, new mobo, new mem, new CPU. All under 220$.
So, nope.

I'm going to have to call complete and utter bullshit on that one.

The Athlon x2 3800+, the cheapest of the Athlon x2 processors, didn't come out until Aug. 2005 and cost $350 at launch. The 4200+, 4400+, 4600+, and 4800+ came out a little earlier in May 2005, but they were priced at an insane $530, $580, $800, and $1,000 respectively. Even their single core parts weren't cheap, the Athlon64 4000+, which came out in 2004, cost $730 at launch. The 3800+? $720. The 3500+? $500.

But AMD would neeever jack up prices when the are in the lead. They didn't do that. And I'd like to point out that, in the same time frame of the Athlon X2 processor, around May 2005, Intel's Pentium D series was substantially cheaper. The Pentium D 830 came out almost the exact same time as the Athlon X2 4200+ though 4800+, and it was priced at $300 at launch. The Pentium D 820, again launched May 2005, was $240 at launch. The Pentium D 805 came out Dec 2005, and was freakin' $145! It was at the time easily the cheapest dual core you could get, and thanks to the low FSB(533 instead of 800) and the high multiplier, it overclocked like crazy. You could throw something like a Thermalright Ultra-120 on it, push the FSB to 800, and be running at 4.0GHz very easily. But the point is, when Intel knew they were behind, their mid-range products, the ones that most consumers were actually buying, were priced lower than AMD's. Their niche products, the EE line, remained overpriced, but those product lines from both companies are always overpriced.

LemmingOverlordwon't lie to you, I haven't benchmarked an FX-8350, much less an overclocked one, but I've benchmarked other CPUs of the same architecture and extrapolating results (to me), and checking on benchmark sites, there is no chance the 8350 comes within that 10% window - clock for clock in multithreading - unless you are quoting one very specific benchmark (that's the problem with benchmarks, right?). It does come down to what benchmark you are running and whether the system is somehow bottlenecked by something... i.e.: do you have to flush something to disk? (from your sig, you have overclocked your FX-8350 and it's got 32GB of RAM which might play a role in this).

That is indeed very true, but you'll note my 4790K system has 32GB of RAM as well, and faster RAM at that. Yes, the benchmark used makes a difference, and I'll admit I haven't exactly had the time to run either system through a gauntlet of tests. But in the few tests that I have run on both, in multi-threading the 8350 at the same speed as the 4790K is about 10% behind.

I actually put together the 8350 for video encoding, because I was tired of having my main rig tied up for hours encoding H254 video. The video encoding is very multi-threaded. And the 8350 is slower than the 4790k, but only by about 10%-15%. So when you really get all of those cores working, the 8350 isn't that far off.

LemmingOverlordClock for clock, single or multithreading, Haswell beats Piledriver...

Oh, most definitely. But Piledriver isn't the latest Bulldozer core. We have Steamroller after Piledriver and Excavator after Piledriver. Excavator just came out.

The real problem is we haven't seen how Piledriver and Excavator really perform in full form. All we have seen is the APU versions, which cut out the L3 to make room for the GPU. Obviously the big issue is they are only 4 cores. But the other issue is they cut the L2 from 1MB/Core to 512KB/Core on Excavator and completely cut out the L3 on all the APUs to make room for the GPU. The Bulldozer architecture really loves large cache. It really cripples the architecture when you take that away. That is why, if you just look at the CPU performance, the FX-4350 still outperforms the latest Excavator based Athlon X4 845 clock for clock even though the 4350 is technically 2 generations older than the 845. That cache just makes a massive difference.

Kevin-HTPCAMD can’t compete with Intel when it comes to CPU reliant tasks, but on average AMD Kaveri APU’s outperform Intel’s latest Skylake processors when it comes to gaming (without dedicated graphics).

This is the crux of the issue, isn't it? AMD's Zen is an architecture. This architecture will later break down into a number of different configurations - some more suited for workstations/servers, others for gaming PCs, others for HTPCs and so forth... 12-core, 8-core, 4-core, with GPU, without GPU, etc... How each one will perform is anyone's guess. Right now all we hear from AMD is that it has come up with a unicorn CPU architecture that does everything really well (if we are to believe the slideware hype).

LemmingOverlordThis is the crux of the issue, isn't it? AMD's Zen is an architecture. This architecture will later break down into a number of different configurations - some more suited for workstations/servers, others for gaming PCs, others for HTPCs and so forth... 12-core, 8-core, 4-core, with GPU, without GPU, etc... How each one will perform is anyone's guess. Right now all we hear from AMD is that it has come up with a unicorn CPU architecture that does everything really well (if we are to believe the slideware hype).

yaaay for slides!! the ones with the most colors are my favorite :)

btarunrCERN engineer Liviu Valsan, in a recent presentation on datacenter hardware trends, presented a curious looking slide that highlights some of the key features of AMD's upcoming "Zen" CPU architecture. We know from a recent story that the architecture is scalable up to 32 cores per socket, and that AMD is building these chips on the 14 nanometer FinFET process.

Among the other key features detailed on the slide are symmetric multi-threading (SMT). Implemented for over a decade by Intel as HyperThreading Technology, SMT exposes a physical core as two logical CPUs to the software, letting it make better use of the hardware resources. Another feature is talk of up to eight DDR4 memory channels. This could mean that AMD is readying a product to compete with the Xeon E7 series. Lastly, the slide mentions that "Zen" could bring about IPC improvements that are 40 percent higher than the current architecture.



Source: HotHardware

Hey.

Not sure if anyone's made this comment but I don't see any reference: It's not exactly 32 cores per socket. Apparently Mr Valsan also mentioned that it was a unique two-CPUs-one-socket using a special interconnect for the CPUs (probably the Exascale Coherent Processor they were going on about a while back). A bit like dual-GPU solutions, I guess.

LemmingOverlordEntirely disagree: a 40% IPC boost means that the "next-gen" Zen will be still underperforming in both single- and multi-threaded tasks when compared to an out-of-date Haswell.

In other words, and much to my disappointment because I don't fancy a world where Intel is king of the hill, AMD's best effort will still trail Haswell in IPC (by quite a lot) AND Skylake (+Skylake refresh) when Zen arrives.

This will be a very poor showing for AMD. In fact it will be the last nail in the coffin for a company that has done its utmost to drive itself into the ground.

I am not an AMD or Intel fan boy, I am pro-market, and FWIW, AMD has been run by people who - to ensure that they aren't kicked out of office - have cared only about the stock value in the short term, and failed to see what the company would be selling 1-, 3- even 5- years ahead. The only thing AMD management has been able to do so far is shore up its defences for its own demise. Even the creation of the Radeon Technologies Group is a way for management to secure some form of revenue in case the company goes tits-up. When things get worse, AMD has packaged all its worthwhile assets in a nice little parcel with a perky little bow on top, ready to be sold off to the highest bidder. I am sure this is a condition the stockholders demanded be met to ensure support.

When Zen comes out (and fails) it will be a short hop to the "we are now a fabless graphics technology semiconductor design company" announcement, "available to license our graphics technology to anyone who wishes to acquire a license", who will adopt an ARM-like strategy towards the market. AMD will be no more and RTG will be its spiritual successor. Heads will roll, golden umbrellas will be paid, Chapter 13 will be filed, AMD restructured, x86/AMD64 people laid off and only RTG will remain, which will then be sold off.

I'm just speculating here, but Zen would've been a hit... like in mid-2015. Not in early 2017.

So, kinda like SEGA then?

medi01Maybe high range-ish?
980 vs Fury Nano

On par power consumption, faster yet cheaper Nano with unique features.

Walking the line between misleading and flat out lies.

Power Consumption - Fury Nano is 50% higher while idle, ~400% higher during Blu-Ray playback, 20% higher average power consumption, and has a slightly higher peak\maximum draw. They are not on par, the 980 is better in literally every category.

Performance - The Fury Nano does better at 4K and slightly better at 2560x1440, but they are tied with the 980 often edging it out at 1080p. The ace in the hole for the 980 though is that it overclocks--and quite well. I imagine even a higher tier factory overclocked 980 would edge out the Nano Fury in almost all situations.

Features - Not sure what unique features the Nano has that the 980 doesn't have comparable versions of. I guess the Nano has HBM? But that doesn't affect the experience all that much other than making it better at 4K.

Price - You can get 3rd Party GTX 980's for as little as $470 (and one model that is $480 with a $30 MIR effectively making it $450) on Newegg, compared to Fury Nano's starting at $490 and most being above $500.

medi01Back in 2004-5 I've upgraded to Athlon64 X2, new mobo, new mem, new CPU. All under 220$.
So, nope.

No you didn't. The Athlon64 X2's didn't come out until mid-late 2005, and even a year after release the Athlon64 X2 3800+ was about $250. I know this because I bought 2--one for myself and one for a friends build--and both times they were at least $200 (I think the one for myself I bought in 2007).

It's just "Nano" not "Fury Nano" and not "Nano Fury". lol

The unique thing would simply be it's size, but your other points are somewhat okay, just that the Nano certainly isn't a 1080p card and therefore no one really cares (or should care) about its performance there. Also wait for the performance leap DX12 will give the Nano (and not give the GTX 980, because it's already at maximum).

Overclocking? You are funny. The Nano transforms into a Fury X if overclocked, the GTX 980 even with highest overclocks has no chance. I'd call your post Nvidia-biased, because you mention "Nvidia can overclock to reach AMD" but not mention that the Nano can be overclocked too, and runs easily away (from the GTX 980) if so.

Also the Steamroller APU has 2x2 MB Cache, not 2x 1MB www.newegg.com/Product/Product.aspx?Item=N82E16819113359
The new Excavator CPU has only 1x1 MB, but I'd call that an exception.

Lots of misinformation in this thread. Also I very much dislike those nay-sayers and negative talkers here. "AMD Zen will most likely be shit, but I don't want it to be shit". Sorry that's crazy.

We will see. My opinion is, Zen will be a LOT better than the FX processors are now, and this is a good thing, and will help AMD to get some marketshare back. I'd even bet on it. Same with their coming GPUs.

KananAlso the Steamroller APU has 2x2 MB Cache, not 2x 1MB www.newegg.com/Product/Product.aspx?Item=N82E16819113359
The new Excavator CPU has only 1x1 MB, but I'd call that an exception.

It doesn't much matter, the lack of L3 is what really cripples the CPU performance.

newtekie1It doesn't much matter, the lack of L3 is what really cripples the CPU performance.

I am really hoping for a large L3 and decently sized "L4"

newtekie1It doesn't much matter, the lack of L3 is what really cripples the CPU performance.

Thought the same, but the 2x2 MB is enough for "4" cores on the A10 APUs (someone told me or I read up in some reviews) - not sure about the 2x1 MB Excavator CPU though.

@cdawall : I'm pretty sure it will have a large L3, because AMD has it on their main CPU line since Phenom II. L4? You mean HBM for the APUs? If so, yes, that'd be nice. Really really nice, because it would erase the bandwidth problems current APUs have with their GPUs.

KananThought the same, but the 2x2 MB is enough for "4" cores on the A10 APUs (someone told me or I read up in some reviews) - not sure about the 2x1 MB Excavator CPU though.

Depends on what you consider "enough". Sure, they'll work with 1MB per core, heck they'll work with 512KB per core. And performance will be acceptable. But performance will be a lot worse than if they had that large L3.

newtekie1Depends on what you consider "enough". Sure, they'll work with 1MB per core, heck they'll work with 512KB per core. And performance will be acceptable. But performance will be a lot worse than if they had that large L3.

I meant it didn't really lose much compared to a Phenom II with 6 MB or FX with 8 MB cache. IPC was higher, bandwidth problems pretty non-existant, but I'd need to find the review for clarification. For now I'd say 2x1 MB is (a lot) more of a problem than 2x2 MB.

xenocideWalking the line between misleading and flat out lies.

Power Consumption - Fury Nano is 50% higher while idle, ~400% higher during Blu-Ray playback, 20% higher average power consumption, and has a slightly higher peak\maximum draw. They are not on par, the 980 is better in literally every category.

Performance - The Fury Nano does better at 4K and slightly better at 2560x1440, but they are tied with the 980 often edging it out at 1080p. The ace in the hole for the 980 though is that it overclocks--and quite well. I imagine even a higher tier factory overclocked 980 would edge out the Nano Fury in almost all situations.

Features - Not sure what unique features the Nano has that the 980 doesn't have comparable versions of. I guess the Nano has HBM? But that doesn't affect the experience all that much other than making it better at 4K.

Price - You can get 3rd Party GTX 980's for as little as $470 (and one model that is $480 with a $30 MIR effectively making it $450) on Newegg, compared to Fury Nano's starting at $490 and most being above $500.



No you didn't. The Athlon64 X2's didn't come out until mid-late 2005, and even a year after release the Athlon64 X2 3800+ was about $250. I know this because I bought 2--one for myself and one for a friends build--and both times they were at least $200 (I think the one for myself I bought in 2007).

what is this bullock of half information.. do you work for nvidia? haha

newtekie1I'm going to have to

Read it again and think what 2004-5 could mean.

I repeat, for 220 Euros, I've got dual core Athlon 64, new mainboard and mem. (don't remember amount though :S)

I don't remember what date it was, Wow released at least 1 year before that, roughly at that time. Wow was the reason I was upgrading.

Mobo was about 55Euro, CPU under 100, maybe as low as 80.

My gmail can't look back that far, or else I'd had exact numbers.

xenocideWalking the line between misleading and flat out lies.

Fury Nano is 50% higher while idle, ~400% higher during Blu-Ray playback, 20% higher average power consumption, and has a slightly higher peak\maximum draw.

Are you on nVidia's payroll or something?
www.anandtech.com/show/9621/the-amd-radeon-r9-nano-review/16

medi01Read it again and think what 2004-5 could mean.

I repeat, for 220 Euros, I've got dual core Athlon 64, new mainboard and mem. (don't remember amount though :S)

I don't remember what date it was, Wow released at least 1 year before that, roughly at that time. Wow was the reason I was upgrading.

Mobo was about 55Euro, CPU under 100, maybe as low as 80.

The dual cores came out in 2005. In september 2005 I bought a single core Venice 3000+ for more than €100, from one of those terrible, cheap stores. No cooler. I remember dreaming about that x2 3800+, but I just couldn't afford it. And here's an old review that puts the street price of the cheapest dual core at $354. Flash forward two years and yeah then.

FrickThe dual cores came out in 2005. In september 2005 I bought a single core Venice 3000+ for more than €100, from one of those terrible, cheap stores.

I recall it was before first WoW expansion came, so, Burning Cruzade, released in Jan 2007.

medi01I recall it was before first WoW expansion came, so, Burning Cruzade, released in Jan 2007.

First gen AM2 then, and those dual cores still was hundreds of €. €100 would still get you a single core AMD. September 2007 I bought an Intel e4300 for a bit less than €100. After TBC was released, when Core 2 had made its splash felt, I can imagine you getting an x2 3600+ for something like that if it was on sale. And now we're three years from 2004 and two from 2005. :)