Ryzen 7 CPU – Your Best Option?

AMD have released their most competitive CPU design in 15 years; eight core monsters than can stand toe to toe with Intels best chips — but should you buy one?  The answer is maybe.

This article will help you make a purchasing decision based on public data.  I will explain where AMDs new design will shine through as well as its limits and how the Ryzen 7 compares with Intel processors.

What’s with this Multicore Technology?

The advantage of Ryzen 7s eight core over Intels comparably priced four and six core CPUs is the most interesting feature of the Ryzen 7.  It is important to understand how multi core CPUs help your programs work faster in order to determine whether the Ryzen is right for you, so I will start with a high level outline of multicore technology.

For many years, a CPU was built with processing elements working together as a single processing engine, with each new version incrementally improving clock speed (how many times a second the CPU can perform a cycle of work) as well as doing more per clock cycle.

The single core design strategy hit its limits in the days of the Intel Pentium 4 – a high clocked, hot and noisy processor that offered only modest benefits over its predecessor and some disadvantages.  At about the same time AMD hit their peak with the Athlon 64, the choice of many in-the-know gamers and others chasing the fastest possible CPU performance.

Intel’s next release was a radically redesigned CPU with a focus on power efficiency and improving performance with multiple CPU cores, rather than just trying to make a single core faster.  These “cores” are discrete processing engines sharing the same physical CPU package that can handle a computers workload in parallel.  The individual cores were not necessarily as fast as prior single core CPUs, but their advantage would show when they could efficiently work on a problem in parallel (not to be confused with hyper threading, that’s sort of a cheat for a single core to more efficiently handle multiple tasks).

Consider a program that needs to add a column of four numbers before it can continue on to the next step.  We assume a CPU core takes 1 cycle to add two numbers.  Say we want to add 3, 6, 7, and 4.  A single core CPU can add 3 + 6 = 9 and a cycle is consumed.  It may then add 9 + 7 = 16 and finally 16 + 4 = 20 over subsequent cycles.  Only then can the program use the result and move on to its next step, so 3 cycles needed.  If a dual core is doing the same job, it is now possible for one core to add 3 + 6 = 9 as before, but on the same cycle the second core might add 7 + 4 = 11.  On the second cycle one of the cores adds 9 + 11 = 20 and the second core is waiting around for more work.  Problem solved with two cycles instead of three.  This is a gross simplification but it is a reasonable demonstration of the possible advantage of multi core.

The prior example can also demonstrate potential inefficiency related to multi core CPUs.  Note that if an eight core CPU was used to solve the same problem, it will still need two cycles since the problem cannot be split up into more than two “threads”, essentially wasting the extra cores.

Lets assume we can build a single core chip that runs at twice the clock speed of a dual core where each core can do the same work per cycle.  You might naturally assume that two identically performing cores at half the speed of a otherwise identical single core would get work done in the same time, but considering the addition example you can see that’s true only in an ideal situation.  In our example, one of the cores in the dual core is waiting around with nothing to do for one cycle where a single core chip is always busy, and as a result the single core would get the job done significantly faster.

On average faster cores are better than adding the same theoretical capacity across additional cores.  This suggests that if AMD have sacrificed performance per core to build an inexpensive eight core Ryzen, the inherent inefficiency introduced will lower the real world effectiveness of these chips more than a modest core performance gap may suggest. A four core chip need not be twice as fast as an eight core to solve multi-threaded problems faster.

To further understand benefits of multi core, and to avoid a common incorrect assumption, it is important to understand how a modern CPU appears to seamlessly run many processes on your computer at once.  Fire up a windows PC and take a look at the Task Manager and the list of processes running.  You might note perhaps 100 different processes, essentially programs all running at the same time.  Even with many cores they can’t all be executing instructions on the CPU simultaneously.  The operating systems job is to share around the CPUs resources so all those programs get a fair and necessary share of the CPUs time.

Consider that modern CPUs run at over 3GHz, or 3,000,000,000 time per second so even with 100 processes of one thread each you are looking at ballpark 30,000,000 cycles for each process per second!  Not all processes are created equal.  Many processes/programs don’t need a lot of CPU time so just because you are running many processes that does not mean many cores will be of benefit.  Again, take a look at your task manager and note that overall CPU use is probably sitting at low single digits.  That shows you have far more CPU processing capacity than you need at that moment.  This idea applies to major applications like word processors, web browsers and so on.  Just because you might have many applications open at once that does not necessarily mean the CPU is being pushed to its limits or that a multicore solution is necessarily better.

Multicore works best when you have a number of threads of execution (one program might have many threads) that are limited by how much grunt the individual cores can offer such that all cores are kept busy.

Specification and Price Comparison

The below table lists comparable Intel processors alongside the new Ryzen 7 CPUs at current retail prices.  Number of cores and base clock speeds are included to provide a broad comparison, however the table does not tell us how much work these cores can do per cycle.  Given their clocks speeds are comparable, the work done per cycle will be a major determinant of the new Ryzen CPU competitiveness.

 

  Cores Clock Speed Price (at March 7 2017)
i7 7700 4 3.6GHz $425
Ryzen 7 1700 8 3.0GHz $469
i7 7700K 4 4.2GHz $469 (no HS)
Ryzen 7 1700X 8 3.4GHz $569 (no HS)
i7 6800K 6 3.4GHz $625 (no HS)
Ryzen 7 1800X 8 3.6GHz $699 (no HS)
i7 6850K 6 3.6GHz $885 (no HS)
i7 6900K 8 3.2GHz $1495 (no HS)
i7 6950X 10 3.0GHz $2359 (no HS)

 

Single Core Performance

With varying core designs and different ways to build an application to run on them, CPU designs will see relatively different performance across differing applications.  For that reason it is inadvisable to look at a single benchmark to draw conclusions across many applications.

There a large number of reviews and benchmarks of Ryzen already available on the Internet, so I will draw from those rather than repeating the same tests.  I strongly advise that when buying a new system for use with specific applications that you check benchmarks for that specific application, or if not possible one that is comparable.

The following is borrowed from a review at PC Perspective and is an interesting benchmark of single threaded rendering performance.  Single threaded means it is using just one core of the CPU, so it gives us an idea of how much work each core in isolation can handle per clock cycle.

Source: https://www.pcper.com/reviews/Processors/AMD-Ryzen-7-1800X-Review-Now-and-Zen/Media-Encoding-and-Rendering

The 1800X has a base clock of 3.6GHz where for comparison the 6800K has a base clock at 3.4GHz.  Both chips can burst faster for a time, and sustainably when other cores are idling, and you would expect that behaviour to be a little different so the average clock speed while running this test is uncertain.  We can reasonably assume the Intel chip is running no more than 10% slower than the 1800X.  With benchmarks of 160 for the AMD vrs 147 for Intel at a little slower speed, it appears from these numbers that the per clock performance of each core is ballpark comparable (147 x 3.6/3.4 = 156), at least for this application.

Taking a look around the Internet for similar benchmarks across other applications, you will find comparable results, however there is quite a large variance between some outliers.  For example this review http://semiaccurate.com/2017/03/02/amds-ryzen-7-1800x-ryzen-review/ shows the 1800X slightly ahead of a 6950X for the same cinebench benchmark, but it shows winrar, common a compression program, running much slower on the Ryzen, about half for single and multithreaded performance.  Interestingly, the winrar benchmark comes up much closer to the other benchmarks when compared to a 6700K, a same gen core, suggesting memory speed/cache might have a lot to do with the performance difference for that particular application.

There are also select applications where single threaded performance is significantly higher on the 1800X.  The lesson to take away from this again is to review performance data for the particular application you have in mind but for the most part the Ryzen 7 is competitive with respect to work done per core.

 

Multi Core Performance

Given the basic specifications of these processors and that we have seen single core performance to be competitive, it is in multicore performance we would expect to see Ryzen shine.

Lets start with the multi threaded version of Cinebench as tested alongside the single threaded performance above.

As expected, the Ryzen suddenly blasts away the 4 core 7700, even though the 7700 has substantially higher throughput per core.  8 vrs 4 core does not lead to a doubling in performance, but a 67% boost is very nice.

In terms of raw performance for this particular benchmark we see the 6950X beat the 1800X by a small margin, but a 8% benefit to a 10 core $2359 processor over a 8 core $699 chip, well, I don’t know many people who could justify the higher price chip for this particular application.

Again, looking around the Internet at numerous benchmarks we see comparable results for heavily multithreaded applications, most commonly media creation applications, where the Ryzen in absolute terms is competitive and with respect to performance per cost is the clear winner.

…but I just want to Surf the Net, use Office, and chat on Skype

And that sums up many of our customers, many home users and typical light business use.  If you are one of the perhaps 80% of computers users with modest requirements, then you are probably not running any heavily multithreaded applications that can take advantage of that extra grunt and probably not running very demanding applications at all.  In that case your beast of a CPU will just be sitting there, twiddling its transistors, waiting for something to do.  As I mentioned earlier but I would like to reinforce, just because you have a lot of programs open at once, a faster CPU will not offer a significant advantage unless at least some of those applications are CPU hungry.

While the Ryzen 7 is currently well priced against Intel’s high end chips, they are still not cheap and for general use you will still find a lower to mid range Intel CPU will give you practically the same responsiveness and leave more in the bank.  If you want to boost responsiveness for modest use, my biggest tip is to put some extra $ into a SSD and perhaps better monitor before worrying about a higher performance CPU.

I need grunt, I play a lot of high end games

But hang on you say, I like to play cutting edge 3D games, now that needs some grunt, sounds a good match for Ryzen?

The nature of many games means their workload is split between CPU based processes such as AI, and display processes such as rendering the 3D world.  Your graphics card is better suited than a CPU for 3D rendering and CPU requirements can vary widely by game.  Most games tend not to generate a large number of threads that need a lot of CPU time and don’t scale well with additional CPU cores.  Very fast one or two cores might contribute more to performance than a large number of slower cores.  Some games tend to be limited by how fast your GPU can draw the game world, so past a point adding a faster CPU doesn’t help.

One of my favourite games series over the years is Civilisation.  I still remember playing the original on an Intel 386SX back at Uni and I can tell you waiting minutes between turns for the computer to do its thinking was painful toward the end game.  It was very much CPU limited.  The more recent versions of Civ are known to take advantage of extra cores more so than most games, so it might be instructive to take a look at Civ VI benchmarks.

Another one from https://www.pcper.com/reviews/Processors/AMD-Ryzen-7-1800X-Review-Now-and-Zen/Gaming-Performance gives a Civ VI benchmark:

Hmmmm, well, that’s a bit surprising.  The high end 1800X is coming out with frame rates under even the mid range four core 6700K Intel CPU.  Looking around the Internet, a number of other sites all show comparable results across a range of game settings and test rigs.

It is not unusual for particular games to do better under one design of CPU or another, so to look further you might expect to find that Civ is a outlier and other games do better.

…I have been hunting through published gaming benchmarks.  Its not looking good for the Ryzen 7 and gaming.  Very different games are coming in with not terrible, but certainly not great numbers, comparable to what we see above for Civ VI.  Given the price points of these chips, and single core performance, the fastest four core Intel chip (currently 7700K) still look to be the best value and when compared to an 1800X is faster and cheaper.  If buying a Ryzen 7 for media work and you also want to play a few games, it will get the job done but it is not a chip your should buy specifically for gaming at this time.

I have a theory which might point to future improvements.  Game developers tweak their code to take advantage of various CPU designs, and given the dominance of Intel CPUs in recent times it may simply be than AMD chips are underperforming through lack of attention.  Its possible OS updates or game patches may help in the future, and more so if developers put increasing effort into optimising their games for Ryzen, but all of that takes time and doesn’t help for now!

When should I consider a Ryzen 7?

For high end media creation where your applications can make use of many cores then a Ryzen 7 based machine may be significantly faster and/or less costly than an Intel CPU based system.  Check benchmarks to verify before buying!

When are Intel CPUs the best choice?

If your computing needs are modest and restricted to typical Office applications, browsing the internet and similar, you won’t see much benefit out of a high end CPU and given the price points of the Ryzen 7 and its limits for gaming, stick with mid to low end Intel CPUs.

If you need a grunt machine for high end gaming, do not buy a Ryzen 7 at this stage.  The top end of Intels four core range is your best bet.  There may be rare exceptions to this rule, specific games that may take advantage of those 8 cores effectively but at best those games are few and far between.

The Future for AMD and Ryzen CPUs

Someone needs to keep Intel honest and it will be bad for performance and cost over time if Intel lose a strong competitor.  The Ryzen 7 has kicked Intel right in the sweet spot of high margin CPUs and as a result I suspect you will see some more aggressive pricing in the high end and more urgency to keep that technical lead.

Later in the year AMD are due to release lower specification Ryzen CPUs targeted to compete against Intel i5 and i3 CPUs.  The Ryzen 7 looks to be very competitive in a niche area right now and if the new chips are well priced AMD should build significant market share across the range.  Just fix the gaming performance!

Related Products

Ryzen @ Computer Alliance

 

Leave a Reply

Your email address will not be published. Required fields are marked *