AMD’s Ryzen 7 3700X and Ryzen 9 3900X CPUs reviewed

AMD began its Zen-aissance with the first-generation Ryzen CPUs in 2017, proving that it was not to be discounted from the high-performance CPU race just yet. While those CPUs weren’t perfect, they offered high-end desktop core counts at formerly unheard-of prices.

Barely a year later, AMD launched the second-generation Ryzen CPUs. Smarter dynamic voltage and frequency scaling, a lower-latency memory controller, and higher peak clock speeds went a long way to make Ryzen more attractive to gamers and other folks who may not necessarily have had much use for a whole pile of processor cores.

Now, AMD is taking a step rarely seen in the history of CPUs: it’s migrating to a next-generation semiconductor fabrication process ahead of arch-rival Intel. With TSMC’s 7-nm foundries at its disposal, AMD has used this genuine, Moore’s Law-compliant advance in transistor density and performance to introduce a family of processors powered by its Zen 2 architecture.

The specific processors that AMD provided for us are the Ryzen 7 3700X and the Ryzen 9 3900X. You can see the most important specifications for these chips in the chart above. As you’re probably aware, these CPUs are not only the first releases with AMD’s new Zen 2 processor core, but they’re also the first CPUs from AMD assembled using multiple heterogeneous “chiplets” in a single package. Get used to that word—chiplet—because we suspect you’re going to be hearing it quite a bit over the next few years.

Chip(let)s’ challenge

So what is a “chiplet?” From the name, you can infer that it’s a little chip. Specifically, AMD refers to its new bits of silicon as chiplets because they are not traditional monolithic processors that function independently. Instead, one of these third-generation Ryzen CPUs is based on two or three chiplets from two different types. One type of chiplet is the CCD, or “Core Chiplet Die,” and the other is the IOD (the “I/O die”). The actual processing happens on one or more CCDs, while the IOD contains the memory controller, high-speed I/O, and other functions.


A diagram of a Socket AM4 third-generation Ryzen CPU. Source: AMD

This change was likely spurred by a number of factors. Notably, it allows AMD to use the same CCD chiplets for every single product across its range. While the company only explicitly names “client products” such as the ones we’re looking at today, all signs point to AMD re-using these same chiplets as one brick in the foundation for its next-generation Epyc CPUs, code-named Rome.

Along similar lines, there’s nothing stopping the company from sticking these same CCDs in everything from game consoles to ultra-mobile processors. In the end, this move allows AMD to improve yields, density, and scalability at the cost of drastically increased design complexity.

Snarky internet commenters have already pointed out that this is not really all that different from the way things used to work when we had both north- and south-bridge chips on our motherboards. The difference between a distant chip on the motherboard and a separate chiplet on the same package is monumental, though. AMD says that “from the perspective of cache and memory access” these new processors “behave monolithically” aside from 1-2 nanoseconds of wire latency on cache accesses. We’ll see if that’s true when we get to our performance testing, but let’s talk a little bit more about Zen 2 first.

Zen, once more with feeling

Don’t be confused: even though these are the third-generation Ryzen processors, the CPU cores inside are based on the “Zen 2” architecture. That design is itself a major revision of the Zen+ architecture inside the second-generation Ryzen processors. Despite the radical shift in processor design to a chiplet-based paradigm, the most pertinent changes in these processors, for our purposes, are those made to the cores and caches.


A diagram of a Zen 2 processor core. Source: AMD

I’m not a CPU architect, so some of the modifications that AMD made go over my head. Still, there are a couple of changes that are easy to understand. Firstly, AMD doubled the width of the core’s AVX units, allowing them to handle 256-bit floating-point data in a single operation. That change alone doubles the speed of Zen 2 cores on crunchy vector math operations compared to their forebears, and brings the core design up to par in that regard with Intel’s desktop chips. More and more applications are starting to use AVX instructions to accelerate vector math operations, so this is a big deal.

Even more importantly, AMD doubled the L3 cache on Zen 2. As with Zen and Zen+, Zen 2’s CPU cores are further subdivided within an 8-core CCD into two quad-core Core CompleXes (CCX). On previous-generation designs, each CCX had 8MB of L3 cache to call its own, but on Zen 2 each CCX now has 16MB of L3 cache. Doing simple arithmetic, that gives parts with one CCD 32MB of L3 cache, and parts with two CCDs fully 64MB of L3 cache. It’s possible AMD that could disable part of the cache for future CPUs, but all of the parts launching today are fully enabled.

AMD specifically says this change was made possible by the move to 7nm fabrication. That’s easy to believe, because 32MB of cache is a huge chunk of the CCD chiplet. Microprocessor makers don’t dedicate enormous swaths of silicon to specific features without a good reason, and we think the big block of cache is there to help mitigate a memory latency penalty brought on by these CPUs’ unique packaging.

AMD is so impressed by the potential difference that the extra cache makes that it has a marketing buzzword for the feature: AMD GameCache. If you see it in the future, know that it doesn’t really mean anything; it’s just branding born from the idea that Zen 2-based CPUs see a significant boost—14%, according to AMD—in game performance as a direct result of the larger cache.

Zen 2 also brings support for the UEFI Collaborative Power and Performance Control (CPPC2) interface. This is not unlike a vendor-neutral version of Intel’s SpeedShift technology. When CPPC2 is enabled, third-generation Ryzen processors take control of their per-core clock rates away from the OS and manage it themselves. Instead of requiring some 30ms to ramp up CPU core clocks in response to incoming work, the CPU can instead hit top speed within 1-2ms. This is a tremendous benefit for brief and bursty workloads such as webpage rendering and application launches. As it requires OS support, using this feature requires the very latest version of Windows 10.

Other major changes in Zen 2 include a double-size micro-op cache (up to 4K), an all-new “tagged geometric” branch predictor, an extra address generation unit, and a larger 180-entry register file. AMD also says it has improved fetch and pre-fetch capabilities, and generally improved load/store bandwidth throughput the device. Overall, Zen 2 looks to shore up the weaknesses of Zen+, and AMD claims that the new chips’ instructions-per-clock (IPC) is up by 15% over second-generation Ryzen. If you’re thirsty for more specific details about the changes in Zen 2, hit up Wikichip Fuse’s in-depth article.

Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14

165 Comments
    • Mat
    • 1 month ago

    Where’s the 5700XT review?

    Reply
    • FRED
    • 1 month ago

    AMD Ryzen 9 3900X & AMD Ryzen 7 3700X – > Cooling: AMD Wraith Prism
    Intel Core i7-8700K & Intel Core i9-9900K -> Cooling: Corsair H110i
    syrlysly?

    Reply
      • Voca
      • 4 weeks ago

      If they do that, then they should technically add the cost of the corsair into the price scatter-plot. It would only be fair since wraith coolers are free.

      Reply
    • jesuscat
    • 1 month ago

    The lack of a next page button is really hampering the readability of the article. I wonder if there would be a way to code it. the page numbers are too easy for me to misclick when i poke my finger at the screen.

    Reply
      • Jb
      • 1 month ago

      I liked the dropdown that lest you quickly see the title for each page and quickly switch between them.
      Would be nice to have something that achieves the same functionality back.

      Reply
      • JVee
      • 1 month ago

      AnandTech had that problem back in 2001 IIRC and fixed it.
      There’s a Print option that makes the entire review a single page.

      ScanAudio speaks very highly of the new AMDs and DAWBench tests show the 3700X on par with 9700/9900 and beating all Intel’s on IPC from the 7700k back. Pretty impressive.

      There’s also evidence of super low latency overclocked DRAM that actually speeds up certain workloads.
      This definitely is no reverse engineered Design.

      Best news, Intel is going to counter before years end, leaks no doubt. CES 2020 will be Intel’s revenge.
      I’m still getting the 3700X just to thank AMD for being itself again.
      Sure look forward to Intel’s angry reply.

      Reply
    • sluggo
    • 1 month ago

    “aside from 1-2 nanoseconds of wire latency”

    Hmm. Signals travel through bond wires at no less than ten inches per nanosecond.

    Those are some long bond wires. My guess is they’re including every possible gate delay and adding a few clock cycles into this “1-2 nanoseconds” quote.

    Reply
  1. I was thinking about upgrading to the 3700x until I saw some benchmarks that had the i5-9600K in it. IMO the i5-9600K would be a better / slightly cheaper purchase.

    Am I wrong?

    Reply
    • Opps. I was not seeing my comment posted so I tried again and it said about a duplicate comment detected, even after refreshing the page so I entered another name and it posted and my original posted as well….

      not a fan of this new comment section on the main site.

      Reply
        • b00nd0ggie
        • 1 month ago

        I love how some incel thumb downed ur comments. u made an innocent mistake and attempted to correct said mistake. Thumbs up from me!

        Reply
      • f0d
      • 1 month ago

      i agree and in some areas (australia at least) the 9600k has had a price dump to in between the 3600 and 3600x but it seems usa doesnt have the price drop for them yet (at least newegg that i checked)
      a z170 motherboard is close to the price of a decent b450 motherboard too

      i just sold my 1700x and im actually going 9600k myself

      Reply
        • f0d
        • 1 month ago

        argh i meant z370 motherboard

        Reply
  2. I was thinking about upgrading to the 3700x until I saw some benchmarks that had the i5-9600K in it. IMO the i5-9600K would be a better / slightly cheaper purchase.

    Am I wrong?

    Reply
      • Aduljr
      • 1 month ago

      Frame times on the 9600k make it not that good.

      Reply
        • f0d
        • 1 month ago

        i have only seen that in one review and one game (gamers nexus and assassins creed)
        might be that one game that does it which is fine as i dont have any plans on playing it

        Reply
    • I am going off of a review I found on PCper. In the CPU Bench test the AMD 3700x outperforms the Intel i5-9600k but in the gaming the i5-9600 runs neck and neck with 3700x

      Reply
        • JustAnEngineer
        • 1 month ago

        https://www.techspot.com/review/1871-amd-ryzen-3600/
        [quote=”Steven Walton”] Ryzen 5 3600 offers a tremendous value. It smoked the Core i5-9600K in every single application benchmark we ran and worst case matches its single core performance. You get 12 threads opposed to just 6, so it’s no doubt going to age better, but this time you don’t have to roll the dice on Ryzen’s longevity, as it’s already faster today. [/quote]

        Reply
    • bonusbartus
    • 1 month ago

    Does anyone have a clue about what is going on with the AIDA Memory bandwidth results?
    I noticed the same performance on other review sites.
    The Ryzen 3 3700X has a very low write bandwidth, and also its read bandwidth is lower than all the other cpus.
    On the other hand, the Ryzen 9 3900x performs great. The latency is about the same between the two.

    Reply
      • enixenigma
      • 1 month ago

      Memory write bandwidth is halved on the single chiplet parts. This was an intentional move on AMD’s part.

      Reply
    • ushio
    • 2 months ago

    Out of curiosity does anyone know which of TSMC’s 7nm processes is used? Intel made a big fuss over millions of transistors per millimetre squared so I was wondering which one AMD used the 96MTr/mm2 or the 113MTr/mm2 one

    Reply
    • Morjens
    • 2 months ago

    Impressive. Well done AMD, again.

    Reply
      • derFunkenstein
      • 2 months ago

      Because physics. Power consumption scales exponentially. Ask any overclocker. They know.

      Reply
      • RAGEPRO
      • 2 months ago

      I think it simply comes down to AMD simplifying things for OEMs. By providing two TDP targets OEMs can safely make a “65W” cooler and a “105W cooler.”

      Reply
      • MOSFET
      • 2 months ago

      65W – 105W are targets, not set in stone (as one example)

      Reply
    • Gastec
    • 2 months ago

    SAY WHAT!? Where is the drop-down menu for selecting between the pages of the article?

    Reply
      • Cyber
      • 2 months ago

      There isn’t even a next button or link!

      Reply
    • Cuhulin
    • 2 months ago

    Zak: Great review of really interesting new chips! The look may be different, but the content is the TR many of us have come to know and look forward to! And now that it is above the fold for those of us using PC’s, I hope many others will give you the kudos you deserve for a job well done.

    Reply
    • Superscroller
    • 2 months ago

    So much whitespace, so much scrolling, so many page numbers without page titles not listed in a handy dropdown box.

    Reply
    • unknown-error
    • 2 months ago

    I very rarely criticise TR, but this new “look and feel” is absolutely hideous.

    Reply
      • Gastec
      • 2 months ago

      Yes, basic functionality is missing, like the drop-down menu for selecting between pages.

      Reply
        • Dennis Clay
        • 2 months ago

        I was looking for the most commented stories and most liked comment links. It may just be me, but couldn’t find them. I frequently learn more from the comments than the stories. Sometimes they’re even funny, though not as often as the people writing think they are.

        Reply
    • MrJP
    • 2 months ago

    Many thanks for the nice review. It’s great to have some new meaty content, and good luck with getting the redesign sorted out.

    Reply
    • Mr Bill
    • 2 months ago

    Zach, A nicely done review! The 3900X is really an upgrade. I like you putting all the traces in a grid. Its easier to compare them; good idea. Can one say that one advantage of the chiplet architecture is that the fastest chiplets can be reserved for the high end CPU? It seems it must be that way when a CPU with more cores can clock higher than one with less cores. It lets AMD turn the paradigm on its head. Faster CPU’s are assembled with the faster chicklets. Period.

    Reply
      • juzz86
      • 2 months ago

      Agreed Mr Bill – a top effort Zak, well done on a very special flagship review mate. It reads very well mate.

      Reply
    • srsly_bro
    • 2 months ago

    Let’s call this what it is…

    Comment gate

    Reply
    • Ninjitsu
    • 2 months ago

    Curious why 9-series i5s and i7s weren’t included?

    p.s. why do i need to input my name and email to post a comment, if i’m already logged in?

    Reply
      • Krogoth
      • 2 months ago

      They would be outclassed by the items in the current line-up more notably on the productivity side.

      Reply
    • Ivanek
    • 2 months ago

    Hardware Unboxed recently made a video comparing the 7700K and 1800X with more recent games.
    Could you please include the 7700K test results as well?
    Comparing it to 1800X could be interesting (and might indicate the gaming performance of the 9900K vs the 3900X a couple of years from now).

    Reply
    • I don’t have a 7700K on-hand to test, sorry. Thanks for commenting.

      Reply
    • JVee
    • 2 months ago

    Too bad the reviewer didn’t have time to do audio benchmarks.
    What’s the use of accepting professional tools from NI, RME and DAWBench if the CPU reviewers cant use the tools, don’t have them or don’t want to.

    Had high hopes for this review site.
    Well at least TR received nice hardware and software, hopefully you got to keep everything.

    Reply
      • Vince Rove
      • 2 months ago

      Really disappointed myself. Looks like another bunch of lazy reviewers who just wanted to cash in on whats hot right now rather than sticking to their guns and providing great, hard to find content.

      Reply
      • Mister Lemiwinks
      • 2 months ago

      I’m disappointed too. It’s not all about gaming and creating YouTube videos.

      Reply
    • Unfortunately, TR doesn’t exist in one big office. The Focusrite audio interface we have used in previous reviews was purchased out-of-pocket by Jeff Kampman, our previous editor in chief, and I don’t have access to any such device.

      There’s also the matter of the complicated licensing for the expensive software required to run DAW testing. In the end, no matter how much I wanted to do it — and I did — there was simply no way.

      Reply
        • sreams
        • 2 months ago

        Audio benchmarks can be run with onboard soundcards if you use ASIO4All.

        Reply
        • Nuno Pimenta
        • 2 months ago

        Daw benchmark was what took me to TR. i hope you manage to get Daw benchmarks again soon. There are actually a couple of people that care a lot for them.
        Asio4all may solve you problem. but i don´t know if it has an impact in audio performance that ruins your benchmarks. But it shouldn´t be hard to try

        Reply
        • enixenigma
        • 1 month ago

        Comments like the ones you are replying to make me quickly miss having to be registered to comment. Not that it stops such ignorance, but at least they would have to jump through more hurdles first.

        Maybe the community can come together to help TR acquire the necessary tools to resume DAW testing? I know I have consumed enough content here to warrant putting a few bucks up for the cause.

        Reply
    • Mostly Sane
    • 2 months ago

    Re: website design change
    Longtime lurker/fan here; TR (was) my favorite go-to tech site. The new change is awful, dumb, and a complete waste of space! At the *very least*, give us a switch at the bottom or top so we can *choose* between mobile and desktop formats. Oh, and if the purpose of the change was *increase* ad revenue, you’ve accomplished the exact opposite (can’t see much now)! Sorry to pile, but…. 🙁

    Reply
    • This is welcome commentary, but please help us out by sharing it in this thread where I’m trying to keep everything organized.

      Reply
      • srsly_bro
      • 2 months ago

      Srsly. The comments and regulars are the reason I come here. The comment section is unusable and I can’t see where my bros at.

      Send help.

      Reply
    • Dposcorp
    • 2 months ago

    I got beat on the web navigator comment 🙂

    Excellent review guys……….off to read the Radeon RX 5700 sutff

    Reply
    • Stephen Pampell
    • 2 months ago

    The page navigator with the summary of each page’s contents, along with the next and previous buttons are sorely missed. Please, bring those back.

    Reply
    • DPete27
    • 2 months ago

    Where’d all the comments go with the new website design?

    Anyway (if this is the right place) [url=https://www.youtube.com/watch?v=qk3PD-4zPN0]Here’s Der8auer’s video on X570 chipset power consumption[/url] (skip to 14:00 for full results table).

    Looks like AMD REALLY borked their attempt at an in-house chipset design. Power consumption is 2x what the X470 chipset needed. PCIe4.0 imparts nearly zero extra power consumption from the chipset. And he was able to keep the chipset at 73C under a comically tiny passive heatsink.

    Reply
    • The previously existing comments will be back soon. We’re working out some things. Use href for links now, BTW.

      Reply
      • Waco
      • 2 months ago

      The rumor mill is circulating the idea that the X570 is just one of the IO dies without any real power-gating.

      Reply
        • Waco
        • 2 months ago

        …and since I can’t quote myself, it appears this is common knowledge and was announced about a month ago.

        Reply
        • It’s also mentioned in the article. 😛 Specifically, it’s an IOD fabricated on 14nm instead of 12nm.

          Reply
            • Waco
            • 2 months ago

            Yep, I’m an idiot. 🙂

      • bonusbartus
      • 2 months ago

      I think that video skips a few important parts.
      Firstly, the older chipsets didn’t even have PCIe 3, which is a very important distinction. PCIe used a different encoding scheme than the older versions and has quite different transceiver requirements.
      To support PCIe 3 (or 4) the chipset also needs to support PCIe 2 and 1, meaning it needs all the hardware for both transceiver standards.
      As you know form all the benchmarks, tests, overclockin etc, Power consumption has a linear relation to switching frequency. as PCIe 4 switches literally twice as fast as PCIe 3 the dynamic power consumption will also double. PCIe 3 again is 1.6x the switching speeds as PCIe 2.
      Also PCIe uses line-encoding to enforce enough 1 – 0 transitions to be able to do clock/data recovery, making it switch almost independent of the amount/sort of data.

      That (including the video above) would just be the impact of 1 lane of PCIe. As the X570 has 4 lanes to the cpu, it would stand to reason at least those would be active most of the time. Then there is the amount of connections to the outside world from the X570. I count 16 endpoint PCIe 4 lanes, 8x 10GBit USB3.2 lanes, 4 SATA ports and some usb2 ports. Going from 4 lanes to all these outputs will need quite some internal switching, routing, buffering etc.

      Lastly, PCIe 3 has some stricter specs on signal quality and might need better (re)-drivers.
      That said, the high idle power draw, might just be an early version bug which keeps the chipset from power gating, or even just from downtraining back to PCIe 1 speeds when idle

      Reply
    • Pettytheft
    • 2 months ago

    What is going on here!! Well I’m glad TR is back in the saddle as I was wondering if there would even be a review up. All this brightness on my screen though! Also thanks for mentioning that all the current patches are installed for mitigations. Some of the reviews I don’t know if I’m looking at old or new data.

    Reply
    • Krogoth
    • 2 months ago

    [url=https://www.youtube.com/watch?v=otCpCn0l4Wo]AMD: You can’t touch this [/url]

    Reply
    • I’m asking about how to do links, I tested with HTML and it didn’t work.

      Reply
    • Jay
    • 2 months ago

    Dropped the ball by not doing a DAWBench test this time around. That’s literally the only thing I want to see. Improved memory and core to core latency is a huge deal for audio production and many of us have been waiting quite a while for this because we need cores and speed but don’t want to blow $1200 on a 9920X.

    Reply
    • As Zak mentioned in the comments that got wiped by the redesign launching, it wasn’t intentional. Moving the licensing around is complicated and he simply couldn’t do it in time. There’s a chance we’ll be able to run those test later, though.

      Reply
        • Nathan Ford
        • 2 months ago

        When do you hope to have the DAWBench results up? Would love to see them….

        Reply
        • We’ll try, but no promises or estimates at this time.

          Reply
            • albackwardz
            • 2 months ago

            I’m a music producer and I found this site by googling DAWbench. You guys give some excellent info for the music producer community. I hope you guys find the time after this big release to get the tests done again. Thanks again.

            • JustAnEngineer
            • 2 months ago

            3 characters per line are visible with Safari at 1920×1080, landscape.

            I.was really looking forward to TR’s gerbildom commenting on these new CPUs.

        • Jay
        • 2 months ago

        Here’s hoping you can get it done. Should generate some extra traffic too because not that many reviewers test audio software in their productivity tests(not that there’s a very straightforward and consistent way to do it lol).

        Reply
    • Austin
    • 2 months ago

    Great review as usual thank you so much.
    Memory latency was the only weakness i saw from AMD Ryzen 3000 series, other than that they are great cpus for any workloads!

    Reply
    • JustAnEngineer
    • 2 months ago

    It’s a great review.

    The layout change is a bit jarring.

    Reply
    • srsly_bro
    • 2 months ago

    Where are all the comments from earlier???

    Reply
    • They didn’t survive The Click. I’m hoping they can be ported over soon.

      Reply
    • Jigar
    • 2 months ago

    I like the new layout and nice review.

    Reply
    • Paul
    • 2 months ago

    You note a regression in the memory latency test. However, I note that the Ryzen 3000 CPUs are being tested with 16-16-16 timings vs. the 14-14-14 timings of the others. This is likely because the Ryzen 3000 is at 3600 while the others are at 3200. Can you tell us what the AIDA 64 memory latency is when the memory is at 3200 14-14-14 speeds? I think that memory latency tests are usually used to show theoretical architectural differences, so the other variables in memory speed should be held constant.

    On another note, the auto-scaling of the charts causes several of the fields to text wrap in the middle of product names or words (in a maximized 1080p browser window). I thing that perhaps you should just upload the charts as images and only allow the auto-layout tool to handle the text wrapping for the main body text of the article.

    Reply
    • BIF
    • 2 months ago

    This is of course not the first time it’s being done, but I’m not fond of the concept of “chiplets” for CPU tasks. Is AMD in trouble here? I thought they were starting to kick some ass with their recent offerings, but now I’m not so sure. Are they having trouble making the new fab process work for full-blown CPUs? Or have they hit a limit (cost, reliability, whatever) designing CPU stuff to all fit inside a single CPU chip? Or a combined CPU+GPU chip?

    Reply
      • just brew it!
      • 2 months ago

      I am guessing it was a combination of yield concerns on the new process, and wanting extra flexibility to mix-n-match compute dies in different combinations to meet different needs. They’d already started down this path with Threadripper and Epyc; splitting out the I/O was kind of a logical next step I guess.

      The smaller process node is also less of an advantage for I/O; you don’t have transistor-dense structures like cache and vector ALUs/FPUs, and you need larger transistors to drive the external buses anyway. So you might as well use a more mature process for that.

      Reply
      • TurtlePerson2
      • 2 months ago

      Multi chip modules (MCMs) have become a lot more popular in the semiconductor world outside of the CPU business. There are a lot of advantages of MCMs and AMD probably just realized that the performance advantage of using a single die wasn’t worth giving up all the advantages of an MCM.

      I wouldn’t be surprised if we see Intel do the same thing in the near future.

      Reply
      • Brother Michigan
      • 2 months ago

      Costs per wafer increased pretty dramatically with the move the 7 nm and, being a new manufacturing process, yields are going to be lower. The chiplet strategy allows AMD to produce smaller dies that increase effective yields for the same defect density, allows them to bin more aggressively because of the larger number of dies, and allows them to reuse the 7 nm parts across multiple product lines. It’s certainly an economical decision first and foremost, but it’s not without its benefits.

      Reply
      • Mr Bill
      • 1 month ago

      The fastest chiplets can be selected to make a very fast multi-core while leaving the slower chiplets for less expensive models. Otherwise the base and max clock could not be going up as core count goes up. This is a great new paradigm!

      Reply
    • just brew it!
    • 2 months ago

    *tap* *tap* *tap*
    Is this thing on?

    Reply
    • It’s on now. I have power to approve posts and will try to stay on top of things today as the dust continues to settle.

      Reply
    • rudimentary_lathe
    • 2 months ago

    Solid review Zak, thanks. Great to see TR with a launch day review again.

    I have a few questions:

    1. With the 3700X, are both CCXs on the same CCD? There was a pre-lease rumour that claimed there would be one CCX on each of 2 CCDs?

    2. The 6-core 3600 seems to get the same amount of L3 cache – 32MB – as the 8-core 3700X?
    Is that right?

    3. Why is AIDA64 memory write so low on the 3700X vs the 3900X?

    4. Crypto seems comparatively weak vs. Intel – any thoughts as to why and performance ramifications in real world use (Bitlocker, dm-crypt)?

    5. Any early indications on virtual machine performance?

    The high memory latency does look a bit worrying on the surface – will be interesting to see if this impacts performance meaningfully in any real-world workflows.

    Overall this looks like an impressive result here from AMD. They seem to have matched or slightly exceeded Intel in IPC. It’s just a shame they won’t OC to the high 4 GHz range – maybe that will come with Ryzen 4! Even still, the value for money here is undeniable, and the performance delta in gaming is now negligible in most titles after the significant improvements over Ryzen 2.

    A 3700X is now on my shopping list for Black Friday (what can I say, I’m not an early adopter). Since these CPUs don’t seem to overclock well, I don’t like tiny fans on mobos, and I don’t need the new features (PCIe 4.0, etc.), I’d forego the Rolls-Royce X570 to save some cash and go with a quality B450 board. I’ll wait to see more real world feedback and reviews with software I use regularly, but it may be time to replace my old Haswell 8 thread CPU.

    Reply
    • f0d
    • 2 months ago

    Pretty disappointed in ryzen myself
    With all the talks about high clocks it sucks seeing them pretty much at their max clocks out the box and still behind Intel in gaming overall

    It wouldn’t be so bad being behind if you could overclock them but they are pretty much max clocked out of the box and even with high voltage reports are they don’t go any higher than around 4.3ghz all core

    With the recent discounts on intel CPUs they are a better alternative at a better price

    [url<]https://www.pccasegear.com/products/46914/amd-ryzen-5-3600x-with-wraith-spire[/url<] [url<]https://www.pccasegear.com/products/44159/intel-core-i5-9600k-processor[/url<] $359 9600k vs $389 3600X

    Reply
    • GatoRat
    • 2 months ago

    I’ve been thinking about the AMD 3700X for my next build. One reason I hesitate is that I got badly burned by the chipset for the Athlon 64 build I did years ago.

    My youngest son wants me to update now since he’ll get my old system, but I’m in no rush and I was quite surprised by how well the Core i7-8700K held up.

    Reply
    • leor
    • 2 months ago

    I really would have thought my 7900x would have been more future proof. It’s delidded and running on an all core 4.6 OC, so it’s fine but 2 years ago 10 coars and 20 threads was SO INCREDIBLE, and it’s like that 1,000 CPU plus the extra 200 for delidding, and well it will be (and is fine). However when you drop that kind of coin you don’t expect it to be disrupted to quickly, I guess in another year I might be buying whatever comes after this from AMD should this process continue.

    Really excited to see competition back! (and TR too!!!)

    Reply
    • Keziwithchimi
    • 2 months ago

    I’m tempted to buy, but I do not trust the mothers with factory fans.

    I absolutely need a fan replacement article for those fans on the mother. Is it doable and practical? if you need to discard a mother because the fan failed, then it is worthless.

    I lost a lot of hardware to accumulation of carpets of dust.

    Reply
    • Srsly_Bro
    • 2 months ago

    X570 chipset is quite a bit more power-thirsty than earlier Socket AM4 chipsets—so much so that AMD’s partners equipped every single X570 mainboard with active chipset cooling.

    This is incorrect. The oarus xtreme x570 doesn’t have a chipset fan.

    Reply
    • albundy
    • 2 months ago

    i wonder how theHEVC encoding compares to the Turing NVENC with the 3900x.

    Reply
    • derFunkenstein
    • 2 months ago

    Really regretting my 9600K purchase right now

    Reply
    • DPete27
    • 2 months ago

    Interested to know why the 3700X is just 65W, but the 3800X needs 40W more (+60%) for only an extra 300MHz (+8%)

    Reply
    • kloreep
    • 2 months ago

    I’d be curious to see gaming numbers for the 3600X if y’all get your hands on one. From what I’ve heard there is indeed not much use for more than 4-6 cores in current games, seems like that may be the real MVP for a dedicated gaming box.

    Reply
    • ronch
    • 2 months ago

    Sometimes still hard to believe AMD is so competitive these days. Crazy.

    Reply
    • jensend
    • 2 months ago

    [quote<]Plus, every X570 board has active cooling for the chipset[/quote<]Not quite - the Gigabyte X570 Aorus Xtreme has passive chipset cooling, but AFAIK it's the only one so far. Wish there were an 'X560' or such bringing some of the IO benefits and futureproofness without the much higher chipset power consumption. I have bad memories of a tiny loud chipset fan from the early P4 era and it's not something I aspire to revisit.

    Reply
    • techguy
    • 2 months ago

    Wow, launch day content from TechReport!

    Reply
    • GurtTractor
    • 2 months ago

    No DAW Bench testing….? 🙁

    Reply
    • cygnus1
    • 2 months ago

    Under your Test Methods lists, can I suggest providing the specific build of Windows used? Win10 1903 has thread placement optimizations for these non-monolithic CPU designs. Would be curious to see a couple of select benchmarks comparing 1903 to 1809 too if there’s time.

    Reply
    • Unknown-Error
    • 2 months ago

    Did AMD just take the IPC crown? 😮 😮 😮

    [url<]https://images.anandtech.com/graphs/graph14605/111165.png[/url<] 10th gen Sunny Cove is thought to gain 18% in IPC, but until then, looks like AMD did the unthinkable.

    Reply
    • ptsant
    • 2 months ago

    Thanks for an extremely timely review. Would be curious to find out whether performance chances noticeably with good cooling (for example Corsair H115i or similar).

    Given that I can simply replace my 1700X with a 3900X on my X370 motherboard from two years ago, this seems like a very nice upgrade.

    Reply
    • Srsly_Bro
    • 2 months ago

    Mr. Wasson is the real MVP here.

    Reply
    • Prototyped
    • 2 months ago

    [quote<]Snarky internet commenters have already pointed out that this is not really all that different from the way things used to work when we had both north- and south-bridge chips on our motherboards. The difference between a distant chip on the motherboard and a separate chiplet on the same package is monumental, though.[/quote<] It is what Intel did for the dual core, quad thread designs for the Nehalem/Westmere generation -- Clarkdale desktop processors (Core i3 and below) and Arrandale laptop processors (most of the lineup). [url=https://hothardware.com/reviews/intel-clarkdale-core-i5-desktop-processor-debuts<]See the package shot here.[/url<]

    Reply
    • firewired
    • 2 months ago

    This was a great review and read, Zak, very well done.

    It goes without saying that, as always since the site’s inception, I came to TR first for the review and I am very pleased to see it on embargo day, and with the quality and content I have always enjoyed seeing from TR since the old days of the long long ago.

    10/10

    Reply
    • plonk420
    • 2 months ago

    it made me smile every time it beat a Threadripper, let alone the 9900K @ “time beyond 16.7ms”

    i don’t suppose you could test core-to-core cross-CCX latency? e.g. pcper.com/2017/06/the-intel-core-i9-7900x-10-core-skylake-x-processor-review/3/

    the context i’m looking for is if it’s low enough for RPCS3 (PS3 emulator) to utilize more than 1 CCX for SPE emulation

    Reply
    • Goty
    • 2 months ago

    A quick area under the curve analysis of the blender numbers (with the 8700K as the baseline) gives the following results:

    CPU — Efficiency relative to 8700K
    3900X — 254.0%
    2920X — 176.1%
    3700X — 173.2%
    9900K — 141.0%
    7900K — 130.8%
    2700X — 108.7%
    1800X — 108.4%
    8700K — 100.0%

    That’s an amazing increase for a single generation and fantastic performance overall. I’m interested to see what a 3800X can do with PBO.

    Reply
    • NTMBK
    • 2 months ago

    Good grief, that efficiency. If the Epyc/Xeon efficiency gap is that big, then Intel is in serious trouble.

    Reply
    • USAFTW
    • 2 months ago

    It all looked pretty impressive to me until I go to power efficiency, but then, MIND BLOWN!
    Just looking at the 3700x, it matches the 9900K pretty closely, but consumes ~ 26% less power, and that is for the whole system?
    Nicely done, AMD!

    Reply
    • Firestarter
    • 2 months ago

    Wow, very exciting! I bought an i7-9700K last year because at the time I didn’t have much confidence in AMD pulling this off but I’m glad to have been proven wrong

    Reply
    • thedosbox
    • 2 months ago

    Wow, the 3700X looks like the ideal mix of price/performance for me. I was planning on a 3800X, but may need to reconsider.

    Any plans on a follow-up with Ryzen 3000 on 4xx series motherboards? I’ll be curious to see whether there’s any performance penalty.

    Reply
    • Krogoth
    • 2 months ago

    AMD II: The Gluing[/url<] Meanwhile Intel will cancel Freedom Lake and start work on [url=https://www.youtube.com/watch?v=rw8RVcUyma0<]Battlefield Lake[/url<] It looks like Intels hold on the regular desktop market is going to take a massive hit. Theres almost no reason to get an Intel solution right now unless you got a good bundle deal or have a use case for the iGPU. Intel only hope in the desktop market lies on the Sunny Cove family.

    Reply
    • chuckula
    • 2 months ago

    1. In multithreaded apps the performance Delta from the 1800X (8 core) to the 9900K (8 core) is about the same as from the 9900k to the 3900X (16 coar). Clearly it’s impossible for Intel to EVAR catch up.

    2. Given how 16 miracle Coarz on 7nm are still will behind the core parts in games, we now know why AMD put on that ridiculously unrealistic “streaming” demo.

    3. We’re canceling…. ALL FPGAS! See, we have surprises.

    Reply
    • plonk420
    • 2 months ago

    so happy TR got this one! also, mountain time zone, best time zone… 7am 7/7

    oops, unintentionally posted at 7:07

    Reply
    • drfish
    • 2 months ago

    Zak, did you ever know that you’re my hero? A true Herculean effort if I’ve ever seen one. Thank you!

    Everyone, if this return to form is the kind of content you want more of, now is the new best-time-ever to [url=https://techreport.com/subscriptions<]show your support[/url<]. Also, please share this review with others on the platform of your choice. The bigger splash this piece makes, the better.

    Reply

Leave A Reply

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

Pin It on Pinterest

Share This