AMD RyZen CPU Architecture for 2017

I'm starting to feel bad for intel
It's a new for me

 

I feel that the problem is a bit different, when you look the core/turbo boost, you can see that 6cores have slighty lower clock than 8cores ( 3.7ghz Turbo ), and that the 4 cores is even less....
So even if they are just disabled 8cores, the only reason i see ( at least on the first row ) is for a market positionement on clock speed .. if this is the only reason, i really think it is not the best strategy vs the quadcore Intel ( as 6700-7700K) ( maybe they just want to selll more 8cores. )..
I have just to read comment here and there on IPC ( vs 7700K ) for see on marketing side, they was maybe something to pull there differently. But well, lets wait that the full performances figures are there ( We could have some surprise as it seems the ipc is higher than expected on Zen that the +40% figure initially given and clock for clock, efficiency seems way higher on zen than their actual Intel counterparts ) ..

This said, thats the first row, untill the end of the year, different sku with different clocks could appear when production will stabilize.

 

Some humor is good from time to time.

 

Figured it wouldn't be long before that look.

It's OK we can still do 1 cycle AVX

 

Here's another one. The guys at AMD might have a good chuckle spree despite whatever was the original intent of the video to mock the Ryzen hype it may actually backfire.

 

A whole New World of price/performance/watt

 

I think it's because AMD wants the higher priced CPUs to consistently win benchmarks (among their own CPUs), which wouldn't be the case with same or higher clocked 4 cores.

 

Which is sad as they are holding back a bit of performance. For now

 

It might not be a good idea for AMD to release harvested chips (failed cores, lower clock stable) as 4-core models. AMD 4-core chips will be compared against 4-core Kaby Lake, while AMD 6-core and 8-core chips will be compared against 6-core and 8-core Broadwell (Kaby Lake and Skylake only have 2/4 core models). AMD has similar IPC as Broadwell and clock advantage over it (and TDP advantage). This means that they are going to win these battles -> good press.

I fear that 4-core comparisions against the higher IPC + higher clock Kaby Lake are going to be lopsided. Intel brings their latest generation at highest clock (112W new models) to the battle. And AMD has only harvested low clock 4-core chips. Press might draw the conclusion that AMD single threaded perf is bad and conclude that AMD can't compete on high end desktops. If I was AMD, I would release all the harvested chips as 6-cores (how many have more than 2 cores broken?), and wait for 4-core dies (or even Zen v2) to fight the 4-core enthusiast desktop battle against Intel. Their 6/8-core chips will be awesome (assuming leaks hold true). AMD can definitely sell every one they can produce.

 

There's a good chance the quad core enthusiast desktop market will be gone by then. Intel is coming out with 6 core SKUs and AMD is going to have 6 and 8 core SKUs.

A Ryzen quad core can be either a single CCX with all cores working, or a dual CCX with two cores disabled in each. The latter will have a higher TDP but performance will probably be similar; CCX to CCX transfers add latency but the aggregate LLC is twice as big which will compensate somewhat.

Low TDP, single CCX quad cores would be high margin, going into the laptop market. High TDP, single CCX and salvaged dual CCX quad cores, would be going to the low end desktop market and thus be low margin.

Cheers

 

The 112W models are Kaby Lake X parts, on a different soket. I hence don't exactly see them as direct competion. Being a quad-channel DDR4 soket, motherboards will also increase the cost. People going for socket 2066 (I'm seriously considering it as an upgrade out of my FX8350) will be rather interested in the higher core counts of Skylake X for workstation duties.

That's why I believe greater TDP and higher - clocked Kaby Lake parts were the plan all along, irrespective of Zen.

The prices however should not be that "Zen-independent" this time

 

I wonder how far this XFR tech will actually push performance.
Frankly if it does a decent job I like the idea of that rather than me messing around manually.
I haven't been able to find the time/effort to mess with my i5 6600k other than setting the bios 4.4Ghz overclock profile.

Regarding the AI neural net tech: does anyone know if this is a small scale thing over a relatively low number of CPU cycles or are they talking much bigger scale like how GPU drivers stored shader caches?

 

It's very likely that AMD will phase out the quad core salvaged Ryzen chips for Raven Ridge once the yields pick up.

On that note 25W Raven Ridge mobile quads, please!

 

Perhaps all Raven Ridge chips will have configurable TDPs.

 

But the damage would be already done. Reviewers would compare the salvaged low-clocked 4-core Zens to 4-core Kaby Lakes and that would make Zen look bad, hurting the Zen brand. AMD needs a good first impression to restore their CPU brand value. I am just trying to say that it's way better from marketing and brand perspective for AMD to attack Intel's Broadwell based $617 six-cores and 1089$ eight-cores and win the fight in all three categories (performance, price and TDP). That is what AMD needs. Fantastic press reviews.

I am just talking about the first impression. Release 6-core and 8-core lineup first. Get fantastic reviews vs Broadwell. Release the salvaged 4-cores one quater later to fill the lower price points. Don't even try to market them against Kaby Lake. Ensure that everybody knows that the fast dedicated 4-cores are coming later. That would be my plan

 

I don't know that quad-core Ryzens will be compared to quad-core Kaby Lakes, when they will most likely be much, much cheaper. Reviewers rarely ignore prices when making comparisons.

 

http://wccftech.com/amd-ryzen-7-1700x-processor-tested/

• The Core i7 5960X scores around 1318 cb
• The Core i7 6900k scores around 1565 cb
• The Ryzen 7 1700X scores around 1527 cb
  
  FireStrike Physics score
  
  

• The Core i7 5960X scores around 14,640 points
• The Core i7 6900k scores around 17,100 points
• The Ryzen 7 1700X scores around 17,916 points

Is this too good to be real? that is a 400 bucks CPU scoring equal or better than 1k ones...

 

Zen 1700X has 200 MHz higher clocks than Intel 6900K. Intel 8-cores are based on Haswell and Broadwell (2 and 3 gen older architectures than Kaby Lake). 1700X results seem believable.

Also the top of the line Zen 1800X (489$) has 200 MHz higher clock than 1700X. If we assume linear scaling over 1700X: Cinebench = 1617 cb. FireStrike = 18,970 points. Awesome results against a 1089$ product.

 

Decent reviewers will simply compare CPUs between similar price points. The FX processors were rarely compared to Core i7, they were compared with their equivalent priced models from Intel, and that's from Pentium against FX4000 to the cheapest Core i5 against FX8000. Power consumption will be something to consider in the "final score", but not what determines the competition as

Quad-core HT-less Ryzen 3 will be compared to dual-core i3, quad/six-core + HT Ryzen 5 will be compared to quad-core HT-less i5 and 8-core + HT Ryzen 7 will be compared to quad-core i7.
I'm guessing the Core i3 will start to die shortly because most games are now fully using 4 cores and HT just doesn't cut it, as recent benchmarks put the 3GHz i5 7400 on par with an i3 7340K overclocked to 4.8GHz (which also consumes over 50% more power). That or the i3 will go down to Pentium's price bracket, eventually leaving core-based Celeron models to die.

What was too good to be real was Intel's record profits year after year after year. Those $1000 CPUs are probably selling with margins around 1000%.

 

Small error on the WCCF article, 1700X ( on their screenshoot ) have 1537 CB not 1527.. ( not much important, but well. )