AMD
Ryzen 7 3700X $326.99 8 core, 3.6GHz base/4.4GHz boost
Ryzen 7 3800X $369.89 8 core, 3.9GHz base/4.5GHz boost
Ryzen 9 3900X $499.99 12 core, 3.8GHz base/4.6GHz boost
Ryzen 9 3950X $749.00 16 core, 3.5GHz base/4.7GHz boost

Intel
i9-9900 $439.00 8 core, 3.1GHz base/5GHz boost, Intel UHD Graphics 630
i9-9900KF $469.89 8 core, 3.6GHz base/5GHz boost
i9-9700K $484.99 8 core, 3.6GHz base/5GHz boost, Intel UHD Graphics 630
i9-9900KS $513.00 8 core, 4.0GHz base/5GHz boost
(i9-9900KS Recommended Customer Price $513.00 - $524.00 )

All prices in U.S. dollars as of October 30th, 2019

For normal work loads, not gaming, the AMD CPUs are on par or surpass their Intel equivalent giving more value per dollar.

The new i9-9900KS only slightly improves over the i9-9900K by 1 to 2% across multiple tests & reviews
& is basically looked at as another "refresh" of the Intel 14nm product.

When the Ryzen 9 3950X is released next month, we will have a clearer option of pricing & choice.

 

Last edited: Nov 7, 2019

The CPU market is getting interesting lately.
AMD last gen CPUs are using 7nm while Intel is clearly having a lot of problems with 10nm. And the clock speed is already at its limit. That makes them squeeze the 14nm CPUs (like @quadcore64 well said with the poor improvements expected of i9-9900KS). Some even say they'll directly skip 10nm and go to 7nm.

That being said, Intel CPUs are faster than AMD with the same technology (14nm vs 14nm). A good rule of thumb is to compare AMD 10nm with Intel 14nm, and so on. But AMD is already in 7nm and Intel still in 14nm...

And don't trust so-called experts YTubbers/bloggers saying "Intel is dead. AMD is the new king and the next is ARM". There is some truth but are very exaggerated claims (Patreon me pleaseeee!). The only clear thing is that nowadays AMD can compete with Intel high-end CPUs and that's good news for us consumers.

 

Last edited: Nov 1, 2019

Errr... The 7nm 3800x vs the 14nm 9900k is almost the same, sometimes faster-fewer times slower, both by a small margin, than the Intel. And cheaper please. Except gaming where Intel wins clearly. But in multi-threaded workloads AMD wins.
Clock for clock difference has been practically eliminated in favor of AMD. Intel wins on turbo speed and this gives them still a small margin ahead in some areas.
So Intel cpus are not currently faster than AMD.
Especially when even after Intel's recent price cut, AMD's Ryzen 3xxx (and Ryzen 2xxx) still represent best value for money, regardless in which area of the market you want to move, from low end budget pc or high end workstation. But it's simple to understand, in one gen they increased their IPC by 15% and moved from 10nm to 7. They kinda leaped about 20-25% ahead of their previous gen which was already fast enough, now performance wise they 're playing the major league.
But still, its AMD 7nm we 're talking about vs 14nm Intel. Which means Intel will answer "better" eventually so it isn't going anywhere, business-wise they just reported their best ever quarter. But they have been known to lie occasionally lol, they 've done it in the past.
Anyway, when Intel finally moves to higher yields the competition will be fierce. Atm i see Intel's only competition is to tag their new HEDT line at half the previous gen's price, which is actually what it should be in the first place. So all this for me, is a clear win by AMD atm, and TR3 isn't even released yet.

Agreed. But the clearest of all if you look a little deeper, should be that for the time being, Intel cannot compete with AMD's high end cpus. 28cores vs 64 man. For less money and much less power consumption per core. Me wants it hahahah.
Cheers mate.

PS: There is enough room for AMD to improve, their x570 and server chipsets (not the cpus) are still 14nm and actively cooled because of the pci-e gen 4 marked as consuming.

 

Last edited: Oct 31, 2019

The real good thing about the competition is the price drop. Also AMD is a fair company, i remember the past year, they sent me at their expenses one CPU to update a brand new motherboard BIOS for new Ryzen compatibility. Many of YT reviewers are aimed at videogame players and everything is translated into frame rate, but for audio production AMD is great and this is a good time for hardware upgrade.

 

Always a pleasure to be complemented and corrected by an expert like you @taskforce .

I didn't meant that so I edited and explained better what I said. I meant that AMD 10nm can be roughly compared in performance with Intel 14nm. But of course, AMD is now at 7nm. Intel clocks more and AMD fits more cores but if things keep up AMD will be faster than Intel, right now is on equal foot.

It could be even possible, at the end of the day Intel's main source of income are high/mid-end laptop CPUs. But they clearly have a problem with 10nm. How much of course is a big secret. And I agree, they eventually counter-attack. But it seems that the 10nm problem is real, and that means a lot of time lost because there're real chances that they'll drop (for desktop high-end) 10nm to go for 7nm in 2021-2022. This could represent AMD taking over. But in the mid run we'll see, Intel still has way more money than AMD.

Me too (salivating...). It seems one of the main problems for Intel to lower to 10 and 7 nm is the SIMD instructions (SSE, AVX,...) that are one of the key factors for Intel CPUs high-performance are not energy-friendly. That wasn't a problem years ago, but now with <14nm and a shitload of cores it is.

 

Thanks for your kind words, i really am no expert, i am just building workstation comps constantly and this has led me to personally evaluate each platform's advantages and misgivings. You are not wrong that Intel rule the laptop cpu market. For the very same reason you can always add that they even rule the laptop GPU market, without ever even releasing a single discrete gpu lol (that is not yet, 2020 will see their first dedicated GPU). But make no mistake, their biggest source of income is the server segment. You and i, and all the rest of us as consumers are small fish. Whatever Intel say or announce at times, they really don't give sht about the desktop or even the laptop market if you like.
The typical scenario is, when one of their "better" client companies decides to upgrade their 3-5 years old servers, that will be a margin of a few thousand(!) Xeons, making it practically a purchase of millions of dlrs. It can even be as high as tens of thousands of Xeons. And this is only one client, take also into account they have been ruling the server market with no practical competition since the Opteron days (IBM is niche, Oracle uses Intel and only now they also do AMD), a good 10 yrs now. Needless to say their desktop products, are derivatives of their server lines.
At this point, people need to understand that it is not just the cpus, but a whole inter-depended ecosystem and very specific workloads that may prevent a company from directly switching to AMD's offerings, even if they want to.
Intel is making waves right now with their non-volatile Optane DDR ram, which connects to a typical ddr4 slot but can be up to a whopping 512gb (atm) and speed-wise sits between DDR4 and the fastest SSDs. This new paradigm, adds a new, unprecedented storage layer between a system's ram and pci-e ssds, taking storage and execution speeds to all new -better- standards. The fact that admins can leverage this new type of memory either as auxiliary volatile ddr ram or non volatile ssd with the "flip of a button" and some soft config, will take the enterprise world and why not, our workstation machines too eventually, to new speed heights.
I am referring to this to indicate that this proprietary technology is well guarded by Intel. (Micron/Crucial has it too since they developed it together but since now they are split, we all know who has the most clients to implement this asap). It is one of those things that can make Intel 's ecosystem seem viable even with their cpu lithography disadvantage. We can be sure that they will mimic AMD's chiplet design to the letter lol. They do have a 56-core Xeon which is practically 2 of their 28 core cpus "glued" together to a monstrous 400+ watts TDP cpu. A friend of mine would call this, a "wishy washy" job lol. Nevertheless, select B2B Intel clients like Oracle & Dell have samples of these cpus already and sell servers with them on demand. Press did sample it too. So on paper they are trying hard to stay relevant with 56core vs AMD's 64. But, the estimated price of the 56core being anywhere from 15 to 20k dlrs versus the much more potent 64core EPYC already selling volumes at 7k dlrs, may put off new clients and make older clients reconsider even changing their entire ecosystem.

Yep and not only. They cannot get a yield of sufficient enough quality at 10nm to withstand production costs, hence they will currently only output laptop cpus at 10nm.

Ffff, i said a lot again, thanks for baring with me
All the best

 

Last edited: Nov 1, 2019

This and also just the time I've spent reading you here makes you an expert. I am too, tough outdated basically because I'm almost broke lol. Of course, if you were a professional in the CPU industry, you wouldn't be. everything has its context ;)

Very true. I was wrong thinking only in user products and I forgot this. Not that the laptop part isn't important, but with the cloud/AI etc era data centers/server farms are the biggest income source there. Besides it's a pretty inelastic market.

So cool! That would be a game changer. I didn't know this. Now we are talking, because the current Optane SSDs are no better than a regular fast SSD used as cache.
It's a thing yet or only claims at this moment?

 

Yep it's a thing alright. Suggested prices from Intel are 577$ for the 128gb dimm stick, 2125$ for the 256gb and 6750$ for the 512gb. But expect this at about 20% higher prices for the time being. Visually you can't distinguish it from a normal DDR stick unless you take its heat spreader apart. Here is how the current models look like:

Also, about the SSDs. The Intel Optane 905P SSD, is the fastest SSD you can buy mate. I 've had only one cross my fingers, personally i can't afford such a drive (700euros/480gb) atm. But setting up that particular workstation pc back in spring, made me hate my already ultra fast Samsung 970 Evo. And i honestly mean it. This thing has 1/10 the latency of every other NAND ssd and this is about the same as their DDR4 product. Given you have no cpu or ram bottlenecks, practically everything you do with an Intel 905P as a system drive is godam instant.
Among the soft installed was Magix Vegas and the fkn thing loaded a 4k /16ch video montage project with a myriad of plugs both audio and video in 1.5 secs. Disk Latency is the real value as i 've said recently. Max speed plays its role of course, and in this for the moment Samsung wins, especially with their upcoming PCI-e 4 drives reportedly hitting a 6.5/7gb /sec. But Intel knows this and they will answer soon enough albeit i presume this will be after everyone else.

Cheers mate

 

So cool. So the golden question is, what's the latency(es) of this new RAM?

Because every new generation of DDR RAM lowers the speed/latency ratio, to the point high performance programming lies more on taking profit of the CPU caches than anything.

It's kind of similar to the decrease in HDDs speed/latency ratio. But we have SSDs now.

 

Last edited: Nov 2, 2019

Errr, first of all, in order to install any of the (Intel) Persistent Memory Modules or PMMs, they have to be supported on the cpu level. Just a bunch of Xeons support these atm. One cpu can support up to six of these PMMs.
There two different modes implemented by the user to focus on for this post.
One is, Memory mode, in which the PMMs behave as typical volatile DDR ram. This will typically be used for systems that need large sums of memory but can't afford the cost of DDR4 in large quantities.
The other one, is App Direct Mode where this employs the non volatile ability of this tech. Using this, lets the OS and apps be explicitly aware that there 2 different types of memory installed in the system. This way they can direct workloads that require low latency and don't need permanent storage to be executed straight in DRAM and those which require perm storage to be lead to the Intel PMMs.
The way i understand it is, by implementing these, the system will magnify it's max read/write bandwidth from 2 to 3 times. Apps who favor sequential read (eg. Large sample libraries like Kontakt,VSL etc., falls in this category) will have the most benefit from this, achieving max bandwidth with the lowest latency possible. Apps with very random workloads will see less benefit at a little higher latency. Best case read intensive scenarios will see a bandwidth of about 40-45gb/sec per cpu. Totally random read/write scenarios may fall to about 15-20gb/sec. Real world mixed read/write loads will exhibit a 25-30gb/sec.
Already Cisco,Lenovo,Dell,Supermicro and Tyan sell servers with these.
By now, you will be wondering, why hasn't this guy answered my latency question, hehehe. Because anything that fits the DDR4 socket, is "greeted" with data hitting the slot from 100 nanoseconds and below and has to cope with it of course, so there's your answer alright.
A typical high end ssd like the Samsung 970 Pro 1tb, exhibits latency from ~350 μs (3500 ns or 0.35 ms if you like) up to 5-6 ms depending the workload.
Typical DDR4 is about 14ns to 20ns (0.0014-0.0020 ms). Note here, that for the most part (but not entirely) DDR4 is read/write agnostic, meaning it doesn't favor or dislike neither read nor write operations.
Intel DC Persistent Memory should be around 100ns or 0.01 ms.
As i see it from a professional music producer's view who makes heavy use of sample libs, if the tech is implemented correctly, it will give our workstations blazing read speed. Already larger ssds help soft samplers with 5-10 times better loading times than hhds. Imagine this x10, given the platform has enough bandwidth to hit those numbers and sustain them.
But as always there are still many factors that need to be sorted out. The tech is sold by a select number of vendors only atm at +20% premium over Intel's announced prices. Also, Intel announced that PMMs will be available at least for half the price of DDR4, but given the constantly falling prices of DDR4 recently, Intel will have to radically adjust their pricing if they wanna stay relevant with their tech and not just some niche product being only high end datacenter friendly. Also OSes and file systems in particular will have to adjust/upgrade for this tech to be viable.
Intel DC PMMs are supported only by a handful of Xeon Platinum Scalable cpus atm. And no one knows if and when this will be available to the workstation/enthusiast end users like you and me, as i don't see us buying a 12k Xeon anywhere near this millennia hahaha.
All the best my friend

 

Last edited: Nov 3, 2019

I'm on hangover so I can't understand this yet but I'll edit this post later when I read it ... thanks in advance for the info