Let me introduce you RyZen - newest AMD desktop CPUs

[astilo]

View attachment 19204

Well, pretty much all the rumors were indeed true. Hat off AMD
Enjoy the videos
[video=youtube;1v44wWAOHn8]https://www.youtube.com/watch?v=1v44wWAOHn8&feature=youtu.be[/video]
[video=youtube;3rUndzpdo1I]https://www.youtube.com/watch?v=3rUndzpdo1I[/video]

I guess Intel cannot just say we still have the best CPU and expect people to simply buy that.
View attachment 19205

View attachment 19206

 

[mustache143]

Agreed, competition is back in the high end x86 space! The question is can it last? Anandtech.com reported the following a couple weeks ago. Anyone know if there is any truth to it?

As the company (Global Foundries) is preparing to start high-volume manufacturing (HMV) of chips using its 7 nm FinFET technology in the second quarter of next year (so, several months ahead of the plan), the actual output of the Fab 8 remains to be seen. Initially, GlobalFoundries plans to use deep ultraviolet (DUV) lithography with quadruple patterning to produce chips using its 7 nm process, but sometime in 2019 it intends to start using extreme ultraviolet (EUV) lithography for a new wave of 7 nm designs.
(source: GlobalFoundries to Expand Capacities, Build a Fab in China)

If that is indeed true, and given what AMD has already said about it's die size advantage over Intel, despite the Ryzen silicon pack(ing) 4.8 billion transistors which includes two million engineering hours and four years of hard workmanship. In comparison, Intel’s Broadwell-E chips pack a total of 3.4 billion transistors. (source: AMD Ryzen CPUs Officially Launched - Massive IPC Increase, Pre-order Today) a quick move to 7nm would further widen the gap.

Everyone agrees that GF 14nm is not as advanced as Intel's 14nm yet AMD was able to offer a chip with 50% less power consumption, pack 1.4 billion more transistors into smaller die size and output better performance than Intel. Give AMD a cutting edge process next year and we will see a tremendous amount of innovation.

 

[astilo]

Well, Intel 14nm is indeed a bit more dense than GF´s one, but then, process technology and chip design are different beasts. AMD is packing xtors better it seems, also due to larger sram sizes to be fair.
As far as we know, 10nm from Intel should come first and that could give again an additional edge to them. It should not last that long though, if 7nm from GF is really seeing the light as per plan. 7nm from Intel, is coming much later then GF 7nm. Yielding faster and delivering is everything.
That said, unless Intel comes out soon with a revolutionary process technology or design, they have to fight Mhz per Mhz, Watt per Watt and dollar per dollar for many years with AMD in the CPU market. Bye bye massive margins.

 

[lefty]

I'm surprised Intel is still selling their enthusiast CPUs at $1000+. Well, they'll pay in marketshare. Ryzen is already the top selling CPU on Amazon. Digitimes claims they will sell 1 million units at launch.

 

[astilo]

I'm surprised Intel is still selling their enthusiast CPUs at $1000+. Well, they'll pay in marketshare. Ryzen is already the top selling CPU on Amazon. Digitimes claims they will sell 1 million units at launch.

BROADWELL-E family is simply dead.
SKYLAKE-X is coming though, most likely in September. Of course, Intel cannot even think to price those enthusiast CPUs at the same level of the BROADWELL-E ones, no matters if they can manage to increase the performances (these CPUs are supposed to have higher frequencies).
The only CPU that can probably still sell well is the i7-7700K, because in terms of single-threaded performance, due to its very high clock, it is still the fastest CPU out there, but since it lose badly in terms of multi-threaded workload, Intel should really consider to drop the price quite a bit.

And that´s just about the desktop segment. Troubles are coming for Intel also for laptop and server markets under the names of Ravenridge and Naples. Good luck Intel, welcome back AMD, customers should be happy about that.

 

[benb]

Summarizing Ryzen:
52% improvement in IPC
Exceeds Kaby Lake IPC on a same-clock, single core basis (by a small margin, in preliminary leaks)
First-of-a-kind: 8 core desktop processor using high volume commercial desktop motherboards
In the $400-600 price range, offering 8 cores, compared to Intel 6, lower TDP, and same clock speed.
In the $300-400 price range, offering 8 cores, compared to Intel 4, and comparable clock speeds to Intel offerings.
In the $200-300 price range, offering 6 cores, compared to Intel 2-4, and comparable clock speeds to Intel offerings.
Launch focus is on providing the Ryzen chip and motherboards to enthusiasts (meaning people who can build their own systems from components).

---------------
52% IPC increase is simply unprecedented, and a tremendous achievement. It will be very interesting to unpack what they did to achieve this, as I hope the reviews on March 2 will do.
The launch focus is somewhat puzzling however. It tends to make me think they have enough chips to seed the reviewers and enthusiasts but not the volume system builders. To supply in volume, and provide supply assurance, two fabs are required (at least at Intel), and AMD currently only has 1 (Fab 8 in NY). So perhaps the impact of AMDs 52% IPC achievement will really only be felt next year, or the year after, when a second fab is underway, most likely with 7 or 10nm chips.

 

[astilo]

Summarizing Ryzen:
52% improvement in IPC
Exceeds Kaby Lake IPC on a same-clock, single core basis (by a small margin, in preliminary leaks)
First-of-a-kind: 8 core desktop processor using high volume commercial desktop motherboards
In the $400-600 price range, offering 8 cores, compared to Intel 6, lower TDP, and same clock speed.
In the $300-400 price range, offering 8 cores, compared to Intel 4, and comparable clock speeds to Intel offerings.
In the $200-300 price range, offering 6 cores, compared to Intel 2-4, and comparable clock speeds to Intel offerings.
Launch focus is on providing the Ryzen chip and motherboards to enthusiasts (meaning people who can build their own systems from components).

I agree, but to be fair, the 4 cores/8 threads i7s (like the 7700K) have higher stock clocks than Ryzen. That means, that while AMD will smoke Intel in multi-thread workloads in this 300$ segment, will be still a bit behind in the single-threaded applications. I´m not considering of course overclock here, just vanilla stock chips. As you said, IPC is very very close and that´s really an extraordinary achievement for AMD. If there are no surprises, AMD has the better performance per dollar at the moment (and also per watt).

It tends to make me think they have enough chips to seed the reviewers and enthusiasts but not the volume system builders. To supply in volume, and provide supply assurance, two fabs are required (at least at Intel), and AMD currently only has 1 (Fab 8 in NY). So perhaps the impact of AMDs 52% IPC achievement will really only be felt next year, or the year after, when a second fab is underway, most likely with 7 or 10nm chips.

They played the game very carefully this time. They had to wait until the very last second to be sure that everything was optimized.
There was no reason to wait for bigger stocks, considering the current advantage versus the competition. As soon as they had enough chips for the launch they did it (I guess a couple of millions, no more). R7 are anyway for enthusiast customers and they are the ones who buy things as soon as they can. The demand for this segment will go down to a well sustainable amount very soon.
The real volume market is in the low-mid end instead, and the R5 and R3 are yet to launch, most likely because they want to have enough parts to satisfy the demand. R3-R5 for what we know are in fact just R7 chips with some disabled/not working cache and cores.

 

[smeyer0028]

Interesting AMD Ryzen versus Intel benchmarking wars

March 2 was the first day AMD's NDA for its new Ryzen processors (1800x is fastest) ended. Original reviews were quite negative on the AMD Ryzen 1800x versus the Intel Kaby Lake I7 7700K. The reviews used speeds on PC games because I think there is now a huge market of PC gamers who buy the latest fast microprocessors and overclock them sometimes with water or liquid Nitrogen cooling. The AMD part lists for half the price of most of the Intel I7s but is about the same as the I7 7700K. The Ryzen 1800x has 8 cores and 16 threads while the 7700K has 4 cores and 8 threads (possibly an advantage in single core IPC rates).

The Intel 7700K clock speed is 4.2 Ghz with 4.5 Ghz boost while the 1800X clock speed is 3.6 Ghz with 4.0 Ghz boost. The 12% higher clock rate on the 7700K explains the game advantage because most PC games are seemingly single threaded (not the display GPU(s) but the bench marks were for CPUs).

Since the original reviews, most of the reviewers have changed so say it is too early to tell which part is really faster.

One interesting review describing the slower 1800x came from Ars Technica. It is interesting because in a Jan. 3, 2017 review of the 7700K, the Ars Technia reviewer said the 7700K was not much faster than the 6900k (also the 6700K). Here is most relevant section of the Ars Technica criticism of the 7700K from January.
URL:
URL: Intel Core i7-7700K Kaby Lake review: Is the desktop CPU dead? | Ars Technica

====
The quad-core i7-7700K, which sits at the very top of the Kaby Lake lineup, is less interesting. Compared to the sixth-gen i7-6700K Skylake processor that preceded it, the i7-7700K gains a small bump in base and boost clocks to 4.2GHz and 4.5GHz respectively, as well as the deceptively named Intel HD Graphics 630. The latter, while new in name, is largely identical to the Intel HD Graphics 530 found in the i7-6700K. There are no major architectural changes, and it runs at the same 1150MHz clock speed. What you do get is support for 4K media decoding inside Windows 10's PlayReady 3.0 DRM, which makes 4K Netflix possible on PC.

 

[astilo]

Interesting AMD Ryzen versus Intel benchmarking wars

Since the original reviews, most of the reviewers have changed so say it is too early to tell which part is really faster.

It is very easy to get only a biased point of view if you pick just one review at random in the web. And that's the reason why I have analyzed a bunch of them.

Let's start from the obvious facts. Ryzen is a very nice CPU, and if we look strictly at the productivity workloads, performance per dollar, Intel has nothing to compete with (closest match is the i7-6900K, but it costs twice as much).
So, why so many disappointing reviews? Well, because of the gaming performance. Is Ryzen really so bad? Of course not.
Let me tell you that the controversies happened mostly because of the testing methodology.
To me the correct approach would have been to test the new CPUs at 1080p, 1440p and 4K to get the full picture.
Gamer sites, in order to focus mainly on the CPU side, tested games at 1080p resolution and below, where the GPU cannot be never a bottleneck for the system (unless it is an entry level card).
Here is where the Ryzen is behind. Early stage of optimization and lower clocks are the main issues there. At 1440p and above though, Ryzen is doing absolutely well.
What does it mean? If you buy a top CPU (>300$ and above) and a top GPU, there is no reason to bother with the 1080p resolution, since people will game at least at quad HD and above.
Should we then discard completely the 1080p results. Of course not. If you are on a tight budget and you cannot afford a decent GPU, then this kind of tests could still make sense. Anyway, if you are on a limited budget, then the R7 family is not really suitable for you. You'd better go with an i3/i5 or wait for the R5/R3 Ryzen family. Because of that, I found many reviews about the gaming performance really misleading (they are not representative of real cases).

Here is my opinion about Ryzen. If you need a powerful CPU to do multitasking, video and audio editing, professional workloads, virtualization, then Ryzen sounds like an easy choice. No, assuming you pair it with a mid-high end GPU, you do not have to give up on games, rest assured. Actually, on 4K, is even doing better than Intel. More cores and threads is the future, since high frequency is already hitting a wall.
Anyway, if you only have to build a gaming machine, I guess that the i7-7700K is still king for the money you spend (although the R7-1700 is in the same price range).
If you are on a lower budget, then you should definitely wait for the R5 and R3 or consider some i5 or i3 Intel cheaper options. It is better to put few dollars more on the GPU in this case, since you get way more FPS than an high end CPU would provide with a modest GPU, since the GPU is going to be the bottleneck.

Below is one of the gaming reviews that I consider a very good overall average, definitely one of the least biased (link to summary and conclusions):
An In-Depth Look at Ryzen's Gaming Performance: 16 Games Played at 1080p & 1440p > Taking Averages, Making Conclusions - TechSpot

 

[lefty]

From the reviews that I read, the 1800X was very competitive with Broadwell-E. However, for gamers neither of these two is a good buy, because of the price and the fact that most games only use 4 cores. Besides that most gamers are running a setup that's GPU bound.
I'm quite interested in what the R3 will be though. The R5 is obviously just the R7 with two cores fused off. That makes it more difficult for AMD to make good margins, because of die costs.
I'm guessing that the R3 will be a desktop version of Raven Ridge (i.e. the laptop chip). The other options are to use the R7 die and fuse off 4 cores (bad idea because of die costs), or use a complete different die with only 4 cores and no iGPU (unlikely, because AMD don't have resources to create so many dies)

 

[astilo]

I'm quite interested in what the R3 will be though. The R5 is obviously just the R7 with two cores fused off. That makes it more difficult for AMD to make good margins, because of die costs.
I'm guessing that the R3 will be a desktop version of Raven Ridge (i.e. the laptop chip). The other options are to use the R7 die and fuse off 4 cores (bad idea because of die costs), or use a complete different die with only 4 cores and no iGPU (unlikely, because AMD don't have resources to create so many dies)

Both R5 and R3 are basically just not fully working R7 dice. So your point about losing money because of the increased die size is not really valid. If one core is not working or a portion of the cache is damaged, they can just ship the R7 as an R5 or R3. It would have been wasted silicon anyway, so actually it is a very smart idea (Intel does of course the same). My guess is that the R7 prime yield is already quite high, so they had not enough downgraded parts to make a proper R5 and the R3 launch since day 1 (and very likely they will sell far more of them, since the mid and low end market is much bigger than the enthusiast one). The binning is instead the way they differentiate the different frequency offerings.
When the Raven Ridge APU chip is ready, there might be another option to get more R3s. Yes, all the defective GPU dice could also make working R3 CPUs.
Anyway, your idea of disabling working portion of the chips is not so uncommon either. But in this case it is going to happen only in case of stellar yields and a very low market for the R7.

 

[lefty]

Yeah, I know the concept of binning, but I reckon that there simply aren't enough defective R7 dies to support both the R5 and R3 product lines.
The R3 is a budget part and budget buyers are going to want a iGPU to save on costs of buying discreet GPU. It makes more sense to me that they use a desktop version of Raven Ridge. Notice also R3 and Raven Ridge are both in 2nd half - that's why they're not releasing it at same time as R5.

 

[astilo]

Notice also R3 and Raven Ridge are both in 2nd half - that's why they're not releasing it at same time as R5.

I do not disagree in principle, but it could be simply be that they get more R5 than R3 from a defective R7, since at the end it is just a matter of chances. So they just need more time to build a proper stock level.
Second point (about the lack of iGPU). Raven Ridge aims at the laptop segment, very true, but nothing stops AMD to use the exact same chip also for desktops.
Depending on the binning, an R7 can be a 95W+ 1800X or a less power hungry 65W 1700, but in principle is exactly the same chip with the same process flow. So a 15W/35W laptop APU, could be used as a desktop APU by just increasing the clock and making it a 65W part.

Zen is a single design which can scale across chips for everything from thin fanless notebooks to high-powered workstations and servers. AMD is particularly proud of this, saying that performance scales in an almost perfectly linear fashion.

 

[kristinalva]

It should not last that long though, if 7nm from GF is really seeing the light as per plan. 7nm from Intel, is coming much later then GF 7nm. Yielding faster and delivering is everything.