LightBlog

jeudi 19 avril 2018

AMD 2nd Gen Ryzen Review: Improvements In All The Right Places

Last year AMD finally pulled off something that many had, even until the release, believed wasn’t possible: They brought back real competition to the CPU marketplace. AMD’s Ryzen didn’t accomplish that feat by raw performance alone though. It did it by offering a very competitive product in comparison to Intel’s CPU lineup. But then Ryzen went even further by allowing AMD to challenge Intel’s status as the go-to for many PCs across the varied segments. It added cores and a very real competition in multi-threading that didn’t previously exist, and it offered all of this at prices that almost demanded a re-evaluation of the loyalty to Intel that many consumers and business decision makers had in place.

As we noted in our previous reviews of Ryzen, the blend of performance at a selling point to make higher processing accessible to even those on modest budgets has been its selling point. It’s safe to assume that if Ryzen had not succeeded as well as it did last year that we wouldn’t have seen Intel respond with the end of a workhorse four core, eight thread mainstay flagship. Intel responded not only with two more cores and four threads in the desktop arena, but its brands were entirely re-done as a detachment between its enterprise Xeon lineup and the consumer Core lineup was completed. And very recently, Intel introduced a higher core and thread count to the mobile segment.

The competition was great in 2017, resulting in consumers winning overall. Now we start our look at 2018’s new products with the second generation of Ryzen and its socket AM4 motherboards. And at the onset of the review, we had one main request for AMD, one that I personally asked its CEO at SIGGRAPH last year: please keep doing what you’re doing right. The second generation of Ryzen seems to show that AMD has very carefully listened to its consumers and read the reviews. While this is a process improvement of the original Zen, many things beyond the technology seem to have been done here that will help keep AMD’s competitive streak in place.

2nd Generation Ryzen Changes

One of the biggest issues with the first generation lineup was that despite the different speeds, many of our samples in the first Ryzen review ended up in virtually the same performance level when manually overclocked. This really meant that there were three price tiers for essentially the same performance. After all, if a person is going to manually overclock it—why pay extra for the 1800X instead of the 1700 when both typically could achieve the same (or near same) results?

AMD seems to have taken this particular feedback to heart—in this case by using the KISS (“Keep it simple, stupid”) approach. Rather than three product levels at the six and eight core counts, we now see two: a version with and without XFR support. AMD sent us the 2600X and 2700X for review. We did perform some basic, multiplier only, overclocking tests with both. Based on what we have seen with both the first and second generation, we suspect that little has changed in manual overclocking. This means that overclockers, while factoring in the ever-present silicon lottery, should theoretically be able to get near similar results overclocking a non-XFR. AMD’s own testing suggested that the sweet spot here would be 4.2 GHz. Since last year’s estimates were very different to our actual findings, we’ll pay a bit of attention to this to see if it holds true this year. The other benefit is that this leaves room for a new 2800X to enter later in the year, should AMD desire to do so.

Overclocking aside, the 2nd generation Ryzen benefits from a 12nm LP (leading performance) manufacturing process which should offer us improvement by itself. At the same speeds, AMD’s own testing indicated about an 11% decrease in power draw, part of an overall 10-15% range of improvement for this new generation over the previous one. The additional power can then be used to increase clock speeds, which now easily is possible both for single and all cores over 4 GHz. The XFR for the 2700X, for example, can achieve a single core speed of 4.3 GHz out of the box.

Separately a SSD caching feature that AMD is calling StoreMi (in a similar naming to its SenseMi features) was also introduced. Much like Intel’s Optane, these seem to rely upon Windows usage and so for now we have not tested these new features. But the benefit to those who wish to use it should be fairly similar, except that the limitations in SSD cache sizes are only what the consumer wishes those to be. This could actually be a good selling point for Intel Optane SSDs as they have been confirmed to work in this role.

Unboxing, Set Up – And Put Aside?

The review package we received from AMD was a bit unusual in that we received two X470 motherboards. Both seem to be the very high end offerings – in this case, from MSI and ASUS. The Republic of Gamers Crosshair series is well known for its very deep options in configuring and tweaking, so we decided to use this for the review. As for the other motherboard and included memory—well, I did mention something in a recent video about a build. But rather than link to that video let’s add the short unboxing video that was made.

Testing Configuration and Methodology

In order to minimize variables during testing, we kept the changes to processors and motherboards only. All other components were shared across the testing platforms and all used the same build and install of Ubuntu 17.10.1 with all updates through April 10th installed. As usual, any components were sourced by us unless noted that it was provided by others. In that case, we will inform who provided us with the particular component. And, as requested by other reviewers, we are including the BIOS version on each.

Motherboards

Processors

Test Bench (Shared Across All)

Operating System & Software

Testing Methodology

As requested, we’ve been adding the “monitor=all” flag for Phoronix Test Suite, but I’ve yet to see details coming back from the sensors. All 6 processors were tested at stock speeds and then given a multiplier only overclock at the highest possible speeds where the Phoronix Test Suite completed. As we have continued to do now for some time, our testing script is available on Google Drive, with a few missing packages now added.

Non-Build Benchmarks

Benchmark Notes: Phoronix Test Suite’s CPU suite offers a plethora of tests and not all are included in this review. The full list of tests and results are available here, with the exception of our LineageOS build times. Those will be included later in the article. Color scheme for benchmarks continues to follow XDA’s traditional color scheme.

FFTW

FFTW is a single-threaded benchmark of fast Fourier transform. No surprises here—we’re still bound by clock speeds. 2nd generation Ryzen gets a boost here as a result, especially with Precision Boost 2 in effect at stock speeds. Noteworthy here is that the gap between Intel and AMD is shrinking—which may indicate the IPC is indeed improving with this new generation.

GZip Compression

GZip is a common compression method and so it makes sense to check out the performance here. Similar story to what we saw with FFTW, though that gap shrink is a lot more visible here. So if your tasks are heavy on single core performance, these results have been worth your while. A jump from first to second generation Ryzen may not be a bad idea if this helps improve your performance.

SciMark 2 (Java) v1.3.1

The SciMark 2 benchmark utilizes Java for arithmetic operations and then provides scoring based on those results. Intel continues its domination here, and it’s odd that we see little improvement in this benchmark between first and second generations of Ryzen.

John The Ripper

On the cryptography front, John The Ripper offers similar results as before. The additional cores get put to work quite well here, but even at the same core/thread count, the 2600X at a lower clock speed stays within a margin of error of the 8700K at 4.9GHz. That’s quite impressive, though we see that 8 cores continues to be the best performer.

C-Ray v1.1

C-Ray demonstrates a similar result, with the 2600X now properly competing against its 6 core counterpart i7-8700K instead of deferring to the higher core counts of the 1800X and 2700X. The i7-8700K still beats the 2600X when given a massive boost thanks to high overclock speeds, but at stock speeds, it’s a lot tighter than it used to be. That’s great news for consumers and for AMD.

Benchmarks: Build Performance

Build Test: ImageMagick

Although we saw one outlier PTS build test (LLVM) most of the build tests this time showed a significant improvement with Intel over AMD. But even then we’re talking differences of 5% between the 2700X and 8700K, which sticks to AMD’s original strategy of Ryzen. Ryzen won’t win every benchmark against Intel… but it will stay close enough that this performance combined with good pricing will still make it worth consideration.

Build Test: LineageOS cm-14.1 Pixel XL

Since the Pixel XL is no longer sold by Google this will likely be our last time using it as our device of choice. It’s also the last time we’ll be building using cm-14.1 (Android 7.1) given its age. Thankfully we’re now in a place where more cores and higher speeds both seem to play a role in build times—and so while the i7-8700K beats out the 1800X at an even higher overclock than before, the 2700X takes back the title of building LineageOS from source in the shortest time possible.

But let’s look more at the 2600X against the i7-8700K. Yes, the i7-8700K outperforms the 2600X. But that difference, before over 10 minutes, has now seen that gap reduced to 7 minutes at stock speeds and 5 minutes at a 4.2 GHz overclock. That’s 15% slower at stock and 11% at overclock speeds. But at MSRP the 2600X is 40% less expensive than the i7-8700K, and for about 13% less expensive a CPU is available that can outperform the i7-8700K in builds here.

One thing that did finally rear its ugly head in these benchmarks was the issues building with the first generation of Ryzen. I was able to accomplish builds for all but the Ryzen 5 1600X using ccache, but it took several attempts in order to complete these. I did not run into these issues with the second generation samples.

Final Thoughts

Ryzen’s second generation took what it needed to from the first generation, improved on it, and bring those improvements to the masses. But this begs the question—who should make the jump to this? And if you are an owner of first generation Ryzen and an X370 motherboard, does it make sense to upgrade?

The motherboard question is almost assuredly yes. We’re going to revisit this here soon because I was able to achieve a better overclock with the Ryzen 7 1800X on the ASUS X470 Crosshair VII Hero than I was ever able to on any 300-series motherboard. This is likely due to improvements that came with the 400-series chipset, knowing how to squeeze even a little more from those processors as possible. But then this leaves another question unanswered: How much do overclockers gain in performance from a first generation CPU like the 1800X over the 2700X?

There’s no doubt that Ryzen, for the second year in a row, continues to offer the best performance at its price point. It’s not always the fastest out there and, for those who want that, they’ll pay a premium to go with Intel. But AMD’s strategy is to get consumers to even raise the question and think if they should look at AMD. And right now the answer for many is undoubtedly yes. As for which generation to go with, that will largely depend on your circumstances at the time you buy.

Right now there’s likely to be some good sales going on with the first generation lineup—enough that it may be possible to get a first generation Ryzen 7 for the price of a second generation Ryzen 5. Based on the benchmarks and results, almost everyone would be better off going this route. Those sales will eventually end, and as it does the clear choice for most will be the second generation of Ryzen. AMD has also done very well by removing some of the fragmentation in the various offerings, consolidating them into very effective options.

We’re still working on the idea of awards or badges, but in the absence of those, we would continue to say this: If you’re in the market for a new desktop, you really should be looking at AMD and what Ryzen can offer your use case at your budget. In almost every case right now a consumer is better off choosing Ryzen and investing that difference into other hardware, such as a higher end graphics card. 2017 was a great year for consumers in the CPU market. And for the second year in a row, in the realm of desktop processors… consumers still win.

I can’t help but love that.

We want to know: how have you been liking these reviews? Have they helped influence your decision on buying PC hardware? Still on the fence or unsure about what to buy? Feel free to let us know in the comments below! Alternatively, you can reach us on Twitter, Facebook, or in the PC Hardware forums!



from xda-developers https://ift.tt/2HADLys
via IFTTT

Publié par à
Libellés : ,

Aucun commentaire:

Enregistrer un commentaire

Inscription à : Publier les commentaires (Atom)