169 comments

[ 3.2 ms ] story [ 104 ms ] thread
My next CPU will be an AMD.
Mine too. I'm giving my current rig another two years.
Is the socket really zero insertion force?

I've always been planning on buying an AMD CPU after dealing with an Intel processor.

Yes, as long as the pins are properly aligned it should literally drop in.
Yes. I remember my 1800X making a "thunk" sound[0] when it dropped into the socket. It requires as much insertion force as water into a cup.

[0] https://youtu.be/QdNISYm0w-U?t=756

If AMD advertised that their CPUs can be installed gingerly as opposed to Intel which felt like thirty pounds of pressure was required, then I would have opted for the older, but AMD chip.

Had a false impression from all the PC part guides that somehow each were equal difficulty.

Most AMD heatsinks use a clip retention mechanism, which definitely require significant force.

Aside: at least new CPUs have heatspreaders. The first heatsink I installed was on an Athlon XP 2500+. Those things had the bare die exposed! Do it wrong the first time, and you had an expensive tiny brick.

Retro tip - copper shim.
I've installed a couple of the Intel LGA CPUs, and I don't remember there being that much force required to put the lever down; although it's a bit less comfortable than the traditional ZIF. Speculation is that ZIF sockets are harder to ruin in manufacturing and shipping than all the pins in LGA sockets, allowing AMD to have more confidence in the supply chain.

For higher end chips, like ThreadRipper and Epyc, AMD uses LGA like Intel -- the pin count is too high not to.

Yes. Typically you lift the lever on the side of the socket, drop the CPU in, and then lower the lever. The lever causes the holes in the socket to tighten, clamping down on the pins in the chip, holding it in place. ZIF sockets have been a part of PC builds and a feature of motherboards since at least the early-mid 90s.
Can’t recall a CPU socket that isn’t lever locked ZIF since Coppermine...
i5-9400F looks amazing though. Powerful and cheap.
I put a 2700X in my most recent desktop and a 2400G in my HTPC. For the price they can't be beat.
Are you gaming with the 2700X by chance? How is it?
Not GP, but gaming on 2700x+1070ti w/ 144hz 1440p monitor. It's suboptimal, but usually I can play games on high/ultra settings ~60fps or full 144 on low-medium. Not having any problems with VR either.
> 1070ti

LOL, cries in RX580 at 60hz on 1080p.

I am doing that. I can play most games very well, including Battlefield, even though it isn't optimized for more than 6 overall threads. Never had a real issue so far.
I game with a 2600x / 2080ti @ 1440p/165hz, on average I see 150 FPS in Apex Legends. Not an optimal setup, but not too shabby either. Planning on upgrading to the 3900x once it's released.
I game with my 2700X/RTx2080 and it’s phenomenal, Battlefield 5 is gorgeous in 4K (though I have to drop to 1440 if I enable RT and ultra settings, that’s the GPU though not the cpu).

It’s so good I’m probably going to defer my upgrade for Zen 2 til Zen 2+ or whatever it’s called.

It works great. With a 1070 as my video card I'm fairly certain the processor is not a bottleneck.
I bought a Ryzen 7 1700X and have been running Ubuntu on it for over a month now. It's insanely fast. (32 GB Ram, 1 TB Samsung EVO 970 nvme, Radeon 580)... I've even been able to overclock it some (3.4Ghz base speed -> 3.77 Ghz). Although I added some additional cooling and I think I could reasonably get 4Ghz out of it.

I miss several of my core daily-driving Mac apps (Querious, MailMate, and Messages) but for ~$1,300... it's hard to beat!

You can run Mac under AMD (compatibility's not too bad now)
This requires a custom kernel though. I wouldn’t reccomend that route for a Hackintosh system.
You could do it via KVM-based virtualization without a ton of trouble.
I'm on a 1700 and I completely agree that it's fantastic. Around the time I bought it, Unreal Engine was released for Linux, so naturally I built that for Linux and the cores made it to a lot quicker.

Over the last 6 months or so, I've had lockup issues, and I thought I'd have to buy a new chip or RMA it, but it seems to have been fixed in a recent kernel as I haven't seen the problem in the last few months.

I'm looking around for decent AMD laptops and they're definitely making their way to the market. Lenovo's A and E series laptops look pretty good (though I've heard of throttling on the A series), and I'm waiting for more reviews on the new T series line. I need a laptop soon, but I may be able to wait until Zen 2 chips are available, depending on launch date.

AMD is making a strong comeback, and I'm excited to see what they release over the next few years and that Intel's response will be.

I’ve used a laptop as a desktop for years but honestly I think having my laptop as a primary machine gave me /too much/ freedom.

Having a desktop makes me finish what I’m doing or makes me otherwise check out. I still take a laptop when I travel but now I’m just SSHing into my machine...

There are many reports about problems with 1700s running Linux

* https://github.com/suaefar/ryzen-test

* https://www.reddit.com/r/Amd/comments/79z43a/ryzen_1700_rma_...

* https://www.phoronix.com/scan.php?page=news_item&px=Ryzen-Se...

* https://www.extremetech.com/computing/254750-amd-replaces-ry...

* https://www.reddit.com/r/Amd/comments/apw8im/ryzen_freezes_i...

Mine (bought a bit more than a year ago) got one random SEGV in maybe 5,000,000 processes when all threads are busy. That's a lot if you are doing full Yocto builds starting 3,000,000 processes each. And maybe one lock-up per month even without heavy load. Not sure whether there is only 1 problem or several of them. A year of kernel updates and one AGESA update promising to fix random freeze has never made any change. I returned my CPU for replacement this week, not sure whether I'll get a new one.

I'm been thinking of building a low-end AMD APU system solely for Windows gaming (all of my computers except an old Panasonic CF-19 are running Kubuntu). Even the Ryzen 3 2200G can handle almost any game I'm likely to throw at it.
I picked up a Ryzen 5 laptop last month (great sale at Costco, needed a new cheap machine) and I'm shocked at how well it performs, in both compute and GPU. It's not my desktop's GTX1080, but it can play anything I've thrown at it without needing an additional discrete chip like my older Intel/Nvidia machine.
My next CPU will hopefully be RISC-V.
Even wider price gap when you account for taxes. Crazy.
Any chance Apple will ever consider AMD chips for the MacBook Pro or any of their computers. I have to wonder how much of the $4k+ I just spent on my MacBook Pro went to pay for the Intel 8-core CPU...
Its unlikely and I wouldn't hold your breath.

If Apple eventually move to their own CPUs (same as what they use for iPhone/iPad etc) then they'll spend time working on that instead. If it means sticking with Intel for another 2 years or whatever until its ready then its worth it for them.

Switching to AMD wouldn't make sense if its only short-term and not in their endgame.

Do you think Apple was "dumb" to "fragment" their OS market even further?

I know they are working on getting apps running on iOS + Mac OS by using the same APIs/libraries, but now iPad OS is a third instance of what I'm guessing is mostly the same kernel / libraries?

eh, I think it is fine, maybe even necessary.

iOS was looking a bit odd with the release of iPhone X as you now started having splintering at least initially (I'm not sure when it happened, but my older iPad started getting iPhone X-esque swipe down gestures for command center and notifications which was a nice touch). iPadOS looks to be iOS with extra add in features and it was kind of needed if they're going for the PC replacement angle now. I'm glad that my new iPad is getting (has, via beta) mouse support. I don't see any value in my iPhone X having it though (unless there were some kind of meaningful reason for docking), so it'd seem to be bloat at that point.

I thought both iPhone and iPad had mouse support now? And, it’s tucked away in the accessibility menu, so clearly not intended for general use, at least for now.
It may be part of iOS 13 proper, I was just giving it as an example of a newer feature that makes sense with a larger screen like an iPad whereas maybe not as much with a (relatively) small screen like the iPhone.

Regardless, iPadOS is shaping up to be the latest iOS release with some extra features that maybe don't make sense on iPhone (at least until they prove great in iPadOS and there's a push by users to include it in iPhone as well). It'll most likely get anything that iOS/iPhone gets but then has its own so they rebrand it as a new OS (seems iOS+ or something denoting they're inherently the same would be more helpful). Having iPad's home screen grid be limited to the number of icons the way it did was apparently keeping it in parity with iOS/iPhone. Allowing that aspect to be "extended" by iPadOS to include docking the notification widgets/Today View on the primary home screen (wish it could stay for all home screens, but I digress) is a net gain.

There were a lot of rumors recently about Apple working on custom desktop ARM CPU. I think that it's very likely that they'll introduce it into their macbooks at least. But powerful Macs is where it'll be really interesting. They recently released Mac Pros. Spending a lot of money on computer that might become kind of obsolete in a few years sounds not very good.

Overall my impression is that Apple does not care about compatibility. They'll just force every software developer to recompile their program with ARM instruction and pack it into installer and if you need some old program, you either out of luck or may be they'll implement some kind of translation (and if performance is bad, it's even better, because that would force software developer to release "optimized" build).

(comment deleted)
Apple is probably going to switch to their own chips before they switch to AMD.
So from a recent standing start Apple are going to beat Intel and AMD - I could see Apple using their own cpu's for tablets and phones but for work stations I don't think so.
You're probably right. But don't underestimate Apple. They have $250 billion in cash reserves.
Supposedly a new 7nm fab in Taiwan is going to cost around 17 billion dollars just for the FAB.

Not sure Apple share holders will think trying to compete with intel and AMD is good use of their money

That's not relevant. Apple isn't going to be manufacturing its own chips, it doesn't do that currently. What the person was talking about was the same thing it does for the iPhone chips, which is design them and then get TSMC to fabricate them.
$17 billion is peanuts for Apple. However, Apple uses TSMC for fabrication and they are already at 7nm. No need to own fabs. AMD doesn't own fabs either. The Apple designed A12X Bionic processor contains 10 billion transistors and is already competitive with desktop processors from AMD and Intel. Do you think it would be that difficult for Apple to design a desktop or laptop version? They have more than enough money to hire all of the smartest engineers on the planet. The problem would be backwards compatibility, like the transition from PPC to x86 in 2005.
But it is the share holders peanuts - is a quixotic and expensive charge to over take AMD or Intel the best use of our money?
I thought I made this clear in my previous response. Apple is already highly competitive in processor design. The A12X matches the mid range processors from AMD and Intel. AMD does not own a fab. Apple doesn't need to own a fab either.

It turns out they are going ahead with this plan already https://www.bloomberg.com/news/articles/2018-04-02/apple-is-...

(comment deleted)
(comment deleted)
(comment deleted)
What do you mean you “could see it”? Apple have been using their own cpus for iPad and iPhone, Apple TV, Watch and iPod touch, for a decade.

They’ve been calling them “desktop class performance” for a long time now, and they’ve included an Apple processor in their laptops in recent years for driving the touchbar and TouchID.

The A12X Bionic gets roughly the same Geekbench score as an Intel i5 8400 in both single and multi core benchmarks, and with an Nvidia Gt 1030 equivalent GPU performance and lower power use.

The idea of them switching entirely isn’t infeasible, it seems to me that they’re pushing hard to be able to have it as an option soon - a few years.

The Ryzen mobile CPUs are quite far behind Intel's offerings at the moment - they're manufactured on a 12 nm process and use the Zen+ uarch, instead of 7 nm/Zen 2 the new desktop chips use.

They have the advantage in GPU performance, but Ice Lake chips are coming in the next couple of months and they should slightly surpass the performance of the Vega GPUs, plus the CPU itself should perform quite a bit better, at least in terms of raw IPC.

How come techquila is publishing benchmark results while noone else is? Is this a controlled leak or a publisher that broke the rules?
Stuff always leaks and sites like WCCFTech or Techquila aggregate those leaks.
More mainstream tech sites almost certainly already have Ryzen 2 CPU's under NDA. Publishing any results ahead of launch would mean they wouldnt get samples again in the future.
Any independent verification here or runs of different test suites?

There are many ways to put your finger on the scale for this kind of test (say, inadequate cooling on the intel part)

Or could it be PCI-E 4.0 significantly increasing the performance?

Tons of reviews will be released a week from now; be patient.
Official reviews are embargoed until the 7th. PCIe 4.0 doesn't change anything for a CPU/Memory benchmark.
Out of curiosity, has anybody had anything in the AMD Ryzen instruction set missing that required you to purchase an Intel 9xxx instead?

Any significant differences in regards to scientific applications or machine learning?

+1. I've heard some optimizations in scientific computing (e.g. Gaussian 16, from someone who does chemistry) is not available without AVX, but I'm not a practitioner of scientific computing, and know nothing about the current situation. Could anyone share your experience? Also, what is the performance figure of multimedia/encoding performance?
The rr debugger had problems with Ryzen. AMD doesn't have AVX512 but that's more of a performance difference; code shouldn't be written to require it.
in ZEN2:

"The key highlight improvement for floating point performance is full AVX2 support. AMD has increased the execution unit width from 128-bit to 256-bit, allowing for single-cycle AVX2 calculations, rather than cracking the calculation into two instructions and two cycles. This is enhanced by giving 256-bit loads and stores, so the FMA units can be continuously fed. AMD states that due to its energy aware scheduling, there is no predefined frequency drop when using AVX2 instructions (however frequency may be reduced dependent on temperature and voltage requirements, but that’s automatic regardless of instructions used)"

"In the floating point unit, the queues accept up to four micro-ops per cycle from the dispatch unit which feed into a 160-entry physical register file. This moves into four execution units, which can be fed with 256b data in the load and store mechanism.

Other tweaks have been made to the FMA units than beyond doubling the size – AMD states that they have increased raw performance in memory allocations, for repetitive physics calculations, and certain audio processing techniques.

Another key update is decreasing the FP multiplication latency from 4 cycles to 3 cycles. That is quite a significant improvement. AMD has stated that it is keeping a lot of the detail under wraps, as it wants to present it at Hot Chips is August. We’ll be running a full instruction analysis for our reviews on July 7th."

https://www.anandtech.com/print/14525/amd-zen-2-microarchite...

> Any significant differences in regards to scientific applications or machine learning?

The PCIe4.0 is good for Data Science ( NVMe Disk / GPU machine learning )

"Corsair MP600 SSDs will offer up to 4950 MB/s sequential read speed as well as up to 4250 MB/s sequential write speed when used with a PCIe 4.0 x4 interface, which is substantially faster when compared to modern PCIe 3.0 x4 drives."

https://www.anandtech.com/show/14584/corsair-mp600-pcie-4-ss...

Darn, I just bought a Samsung 970 Pro 1TB NVMe drive for my new Ryzen zen2 build. I probably should have waited to buy this PCI4.0 drive instead. Thanks for sharing
> Corsair MP600 SSDs will offer up to 4950 MB/s sequential read speed as well as up to 4250 MB/s sequential write speed when used with a PCIe 4.0 x4 interface

WOW! Looks like they haven't announced pricing yet. I wonder what those drives will cost.

$250 for the 1 TB version and $450 for the 2 TB model, Corsai accidentally put the listings up early on Amazon a few days ago and then took them down
I'm considering Intel for my next workstation because I want to be able to run macOS at least in VirtualBox (or just run Hackintosh, did not decide yet). For some reason macOS does not like AMD CPUs.
You can run MacOS through KVM easy on Ryzen. Even GPU passthrough works fine.
iMessage and other features generally break when running amd macintoshes, intel ones are indistinguishable.
Why would they break if macOS itself runs fine on AMD? These softwares run on macOS and don't make any system calls directly by bypassing the OS, right?
They typically check for a network interface from a real Mac. Using a Hackintosh you can't spoof some of the NIC details that are detected and cause iCloud/iMessage to not work.
Oh, wasn't aware of that. So does iCloud / iMessage work when macOS is run virtualised?
Oh man, my suggestion buy a used dual socket Xeon V2 (or V3 if you can afford it) server with room for a GPU. The price for performance is pretty unbelievable since you get a whole system, which you can run ESXi or whatever flavor of hypervisor you want.

V2/V3 Xeons are dirt cheap on ebay and if you don't mind a little risk the ES versions are mighty cheap for what you can get.

I just got Dual 2667 V2, 96GB of ram, dual 1200W PSUs, and room for 20 SAS2 HDs for less than $1,000 (replaced the 2637 V2s that came with the system). I'll note power consumption will be quite a bit higher, but power is pretty cheap where I live.

.... I'll be really stoked in a few years when the Rome chips are available used.

I have a Dell Precision T7600. The only thing I hate is the almost full minute to boot with a perc controller and other servery type checks
With a little elbow grease this hasn't been true for Hackintosh builds nor virtualization for ages: https://amd-osx.com
There was a recent video on Linus Tech Tips where they set up a Threadripper workstation running macOS via KVM and it out-performed a top of the line iMac Pro.
Does it work with VirtualBox? I want to run Windows.
You can run Windows in KVM. I haven't used VirtualBox for years so my knowledge is out of date.
It won't work in a virtual machine? Any reason why?
The mozilla/rr debugger doesn't work on Ryzen CPUs - https://github.com/mozilla/rr/issues/2034

(rr stands for Record and Replay Framework, it's a gdb extension which allows for reverse execution)

Interesting that GDB had its own "time travel" support for longer that afaik works on AMD. Will have to see what rr brings to the table.
GDB's recording is extremely limited. It doesn't even work on multi threaded programs.
It's also about 1000x slowdown during recording, where rr is < 2x.
Yes official and guaranteed support for ECC but it was an i3-8100 instead of an AMD equivalent. For a FreeNAS build. I will upgrade it to a Xeon E-21XX or whatever supports my Supermicro C246 motherboard in the future. AMD doesn’t test ECC on Ryzen but doesn’t turn it off. Server grade components are not as common either.

Threadrippers are tempting but I like to play with EXSI and Intel’s virtualization is handy. My next desktop might be a Threadripper but my i7-7700K and Intel QSV is good enough for Plex and 4K.

I would like have a 10gb backbone with SFP+ so I can play with Fiber, DAC, and 10gb Ethernet.

What precludes you from using ESXi and virtualization on AMD's platform? The virtualization extensions to x86 are common to both AMD and Intel. Even direct IO and IOMMU. Is there something missing?
The virtualization extensions are actually different, but yeah, VMware supports both across all platforms. I think it's only AMD Hackintosh where you might have an issue?
I use the TSC for high precision timing, and turning TSC ticks into nanoseconds requires interpreting model specific registers, so my timing code would need AMD specific patches to work properly on those processors. Not necessarily a deal-breaker, but it would definitely be an annoyance for me if I did decide to switch. Probably not super relevant to scientific computing or ML though.
How do you use the TSC for timing when the frequency is changing unpredictably?
Recent processors have invariant TSC.
There might be meaningful differences in the hardware encode/decode for video. In the Plex community for example it seems like Intel CPUs are the best bet. They support a lot of codecs and can service multiple transcode streams. Consumer Nvidia GPUs are limited to 2. I don't think Plex supports hardware transcoding on AMD Ryzen or discrete GPUs at all right now. I haven't really looked into why yet, and could likely just be a software support issue in Plex.

This, unfortunately, would be a deal breaker for my home-server use case, although I'm very happy AMD is finally competitive again, overall.

The Zen2 chips being released and discussed here don't have a GPU. The Ryzen 3000 series APUs being released soon are using last year's Zen+ CPU architecture; and are only 4 cores, so not super comparable to a $2000 Intel CPU :)
I have too frequently had problems with hardware support in OS outside Linux and Windows for non Intel chipsets.

One that sticks out is virtualization support in freebsd and others. I don't know if it's still true but the time risk isn't worth the cost benefit for me any more.

I just assembled a new computer: amd 2200g ($90) + MSI B450M MORTAR ($90) + Intel 512GB SSD M.2 ($68) + 4 x 16G 2666 memory (4 x $72).

I installed the debian testing system. After the installation, X hung at starting up. But after installing amdgpu driver by following this tutorial: https://wiki.debian.org/How%20to%20install%20official%20AMDG... , it works like a charming now.

Do you happen to know if cpu and case fan controls work on Debian on your machine? I have a server with a very similar setup, but cannot control the fans.

    ~ sensors
  No sensors found!
I had to install the out of tree version of it87 to get fan control working on my ASUS Prime X370-Pro.

Unfortunately, the driver was abandoned last year, but copies of the code are still around on github.

The sensors command returns the following info:

    $ sensors
    amdgpu-pci-3800
    Adapter: PCI adapter
    vddgfx:           N/A  
    vddnb:            N/A  
    temp1:        +36.0°C  (crit = +80.0°C, hyst =  +0.0°C)
    
    k10temp-pci-00c3
    Adapter: PCI adapter
    Tdie:         +36.5°C  (high = +70.0°C)
    Tctl:         +36.5°C 
I don't know how to detect fan temperatures.
I had to double check your price on the SSD. When did SSDs get so cheap??
That was my reaction when I looked again.

You can get a 1TB WD Black nvme m2 ssd for less than half the equiv a year ago, bonkers.

Just this past month every local semi-tech store (think Staples, Best Buy, etc) had sales on SSD's that were nuts. I picked up a stack of 120 Gb HP drives for ~$20 each, and the 256's were not even $10 more. It let me put back in to use a bunch of good-except-for-dog-slow-spinning-rust systems I had lying around. Not complaining!
SSDs were held up in prices artificially for a while related to the memory price inflation that happened throughout 2018 and some of 2017. QLC was mentioned but prices are down across the board and everything in consumer land is down sharply YoY from that period (50%+ discount is fair). I’d expect some price increases coming up due to investor pressure for revenue.
What about SSD reliability? Did it improve in the last year or so?
QLC literally drops write endurance by an order of magnitude but most consumers won’t be hit by it. Most of the performance is from the SLC cache and it tanks once it gets full or with sustained usage. The bet is that won’t matter for 99% of users but most of HN is not that 99% so caution must be exercised when buying these SSDs.
Wow - nice combination. I am going with an ASUS B450 prime myself.
Why a B450 board over an older cheaper A320 board if you're using a 2200g?

Edit: NM, probably RAM...

Yes, one reason. The other is the board supports two m.2 slots.
_almost_. We always pay dearly for that last few points of performance! (Intel is 4% better, comparing the multi-core number)
In September the stars should align when the 16 core part drops (4 more cores) for still less than half the price of the 9980XE.
My ryzen 2700x was probably the best purchase I did last year :) it's just fantastic and I would really want to try out these new CPUs, just can't find an excuse to buy them :)

Takes like a second to cross compile Go app for all architectures, many cores ftw!

Same boat, bought 2700X as stop gap because couldn’t wait for Zen 2 and it’s so damn good I can’t find a reason to upgrade to a 3900X or 3950
Take with a grain of salt. Comparing a single benchmark against an aggregate isn't particularly meaningful (although getting within a stone's throw definitely is). You can go through and cherry pick benchmarks and find plenty that obliterate that measurement.

Better article title would be, "3900x appears comparable to 9980XE in single threaded workloads."

But single treaded was really the only area Intel used to have a lead in - which is interesting
Those nanometers are not just numbers. TSMC 7nm is ahead of Intel's 14nm. It's not really surprising. Even if AMD can't catch up with frequencies, they have more transistors, so they can increase many things. Better comparison would be with 10nm Intel, but Intel failed there, so I expect AMD to lead until Intel will release mainstream 7nm CPUs.
The 3900X already crushes in multi-threaded. That stone-throw multi-core result result is 12 AMD cores on vs 18 Intel cores. That's 50% more cores and 300% more cost for basically the same results.

Single-threaded is the only thing Intel's been touting as they get steamrolled. This is particularly important because a big market for this class of CPU is desktop gamers and for better or worse games tend not to be particularly well multi-threaded so single-core performance is king in that market. Obviously this will change over time as engines continue to adapt.

To your point though it's better to look at an overall trend of AMD vs Intel over the last couple of years, and that's equally devastating.

What I mean is that it's a sample size of 1. I can go through the 9980xe results and cherry pick a dozen iterations that crush this single sample.

Even looking at the leaked 3600 benchmarks listed on Geekbench and there's some pretty gnarly variance. The benchmarks are extremely promising, and I'm likely going to be an AMD customer in a month, but these articles are not grounded in statistical significance. Wait until the embargo is lifted before posting stuff like this, imo.

> Obviously this will change over time as engines continue to adapt.

They're already fairly well adapted - Xbox One and PS4 both have 8 cores for a reason.

Calling Geekbench a "single benchmark" is misleading because it runs 25 different types of workloads. So AMD doing well here means it does well compared to Intel in 25 different areas when they are all averaged:

  • Cryptography Workloads
  - AES
  • Integer Workloads
  - LZMA Compression
  - JPEG Compression
  - Canny
  - Lua
  - Dijkstra
  - SQLite
  - HTML 5 Parse
  - HTML 5 DOM
  - Histogram Equalization
  - PDF Rendering
  - LLVM
  - Camera
  • Floating Point Workloads
  - SGEMM
  - SFFT
  - N-Body Physics
  - Ray Tracing
  - Rigid Body Physics
  - HDR
  - Gaussian Blur
  - Speech Recognition
  - Face Detection
  • Memory Workloads
  - Memory Copy
  - Memory Latency
  - Memory Bandwidth
Source: https://www.geekbench.com/doc/geekbench4-cpu-workloads.pdf
Sorry for the confusion, I meant "single benchmark" as in "single run of Geekbench." As in comparing one measurement versus an aggregate of measurements isn't particularly meaningful, you need to compare aggregate to aggregate to get something valid. It's still a sample size of 1, even if the sample has 25 dimensions.

(I don't know if Geekbench uses an average or what they're weighting is e.g. with memory speeds and operating systems, which is why I'm saying "aggregate").

I'm all for reproducing results, but given the relatively deterministic nature of computers (especially for new hardware) I feel this is an unwarranted criticism. How often do people get significantly different results when running identical tests on identical hardware?
Just looking at the first page of results for the 9980xe, it's pretty common, especially for multithreaded benchmarks:

https://browser.geekbench.com/v4/cpu/search?q=9980xe

But the big qualifier is "on identical hardware." The geekbench score in the article is not a measurement on the same hardware, it's an aggregate of measurements across lots of hardware (different memory clocks, motherboards, etc). That's why a single run is not particularly useful, it just tells us what that user got on that day with that hardware. We need lots of data points to tell us more or less how it will compare on most builds.

Again, not to diminish how promising the score is for the product. Hats off to AMD. I'm just waiting until there are more data points to be confident.

Fair enough. Thank you for taking the time to explain that to me when it was my fault for not understanding the tests (or really the test conditions).
Can you disable frequency scaling reliably? In general clock speed is not deterministic - it depends how hot the CPU is, which is affected by all kinds of things (second run, fifth run, AC on, winter, etc.).
This is comparing an unreleased CPU to a two year old one.

But more importantly, it gives no information on how the Intel CPU was configured and we have no idea what kind of memory it was paired with.

It's just comparing two random user submissions. Here's another i9-9980XE where Intel scored 7700 in single core.

https://browser.geekbench.com/v4/cpu/12111208

> This is comparing an unreleased CPU

It comes out in 10 days, does it really matter?

> to a two year old one.

The 9980XE launched Q4 2018, did you mean 2 quarter old? This also seems to imply Intel has something else better is due for release in the next year, there is nothing on the roadmap.

> It's just comparing two random user submissions. Here's another i9-9980XE where Intel scored 7700 in single core.

Because that's an extreme OC bench at +800 MHz and this was a stock run. Check the .gb4

> It comes out in 10 days, does it really matter?

Yea it does, because no one in the public can reproduce the results to verify them.

The 9980XE is just a really large/fast Skylake CPU. Skylake is years old. If you wanted this level of performance from Intel you could have ordered the Xeon Gold 6154 way back in 2017. It seems fair to call this a 2-year-old part.

A more comparable part would be the i9-9980HK.

> A more comparable part would be the i9-9980HK

not really, that's an eight core CPU (and also a mobile part). I'm very interested to see how the lower core count Ryzen 2s compare to their Intel counterparts in single-thread, but part of what's so impressive about the (early) 3900X/3950X results is that they offer this kind of performance with more cores and at a consumer-friendly price point.

you're right that Skylake is an old design at this point, but intel invites this kind of comparison when they can't manage to ship an HCC version of their latest tech. entirely fair, imo.

True, Intel's part numbering scheme has passed beyond all reason. Think about the i9-9900K/KF. Existing SKU, in stock at Newegg for $479. You don't get quite as many cores but you get a lower price and higher single-core performance, and it exists right now instead of being unobtainable hypeware.
like I said, I'm very interested to see the head-to-heads, especially the overclocking results. my use case is c++ compilation and csgo so it looks like it'll be a tough choice.
Seems like for C++ builds you are always going to want the most cores for the money, even if the single core performance is a little worse.
> The 9980XE is just a really large/fast Skylake CPU. Skylake is years old.

Unfortunately larger and slightly refined Skylakes is all that Intel ships these days on Desktop. If it’s unfair to compare Ryzen CPUs to that, then I’m afraid there is no fair comparison that can be made until Intel gets around making actually new CPUs.

(comment deleted)
The deal is that Intel is still trying to sell expensive CPUs and there's no reason for a customer to be buying them - they could be putting that money into other components - or using it for something else.
Older CPUs generally are a better deal because they have seen a few iterations of price decreases. A 1800x costs $200, a 2700X costs $260. Both have the same core count. Most likely the latest 8 core will cost more than $300.
Never bought an Intel CPU after learning about their illegal anti-competetive deals with OEMs. It's been probably 15 years with AMD CPUs.

Happy to see AMD on top once again.

Never bought Intel CPU for a desktop after Pentium 133MHz that powered my first PC.

I always considered Intel for my builds but AMD always came out as best bang for buck in my price range in the end and I had no way of justifying use of Intel CPU.

I bought one in my last laptop, by that was over 5 years ago. There just hasn't a good selection of decent AMD-based laptops out there, though that situation is finally getting better (excited about Lenovo offering AMD-based Thinkpad T-series, though I'll probably buy an A or E series for now).

In fact, I think the only time I've purchased Intel has been for laptops (except renting through VPS providers). If you don't particularly care about power, they've dominated the low end for years, and now with Ryzen, they're dominating pretty much everything including power (provided you need something more than Intel graphics).

I'm excited for real competition vs Intel on power from AMD, and I hope this little boost helps them continue to compete when Intel eventually catches up.

Yeah but I can't find a motherboard and RAM that won't hardware reserve half my memory without RMAing 4 iterations of each.
This was expected because the 3900X is optimized for fewer faster cores (12 at 3.8GHz) while the 9980XE is optimized for more slower cores (18 at 3.0GHz.)

A better and far more impressive comparison would be the 3950X vs 9980XE which have comparable number of cores (16 vs 18). And in that instance AMD still beats Intel in both single-threaded and multi-threaded workloads! And by wide margins! And with AMD being 62% cheaper ($750 vs $1950)! And with AMD's TDP 36% lower (105 vs 165W)! And despite having 2 fewer cores! And despite the 3950X engineering sample below being clocked under final specs (3.29GHz vs 3.5GHz)! See this leaked 3950X engineering sample codenamed "Myrtle": https://browser.geekbench.com/v4/cpu/compare/13612629?baseli...

Intel has never been so thoroughly defeated since the 1999 K7. Interesting times...

Also doesn't Intel have a lot more known security flaws?
One more, assuming you mean meltdown. But that's the easy one to fix.
Not true. Copy-pasting from my Ryzen terminal:

   $ spectre-meltdown-checker | grep "NOT VULNERABLE" | wc -l  
   12
So? The same script reports 11 lines on my 4 year old Haswell. It should read as "mitigated or not vulnerable" instead of simply "not vulnerable".

Edit: oh wait, it's actually 12 lines, the same as on your system. The reason it reported 11 on the first run is because KPTI has been disabled manually here.

:) I grepped for the wrong thing. Here's a better version:

   $ sudo spectre-meltdown-checker | grep "Not affected" | wc -l


   On Skylake: 0
   On Ryzen  : 10
I don't have an Ivy Bridge CPU to test this on, but AMD has suffered from _significantly_ fewer number of vulnerabilities than Intel, not just "one more" as the top commenter suggested.
What about the recent one ZombieLoad?
Does the old “there are more software written with Intel in mind than AMD” still hold? Just like Nvida.
A critical difference is that GPU architectures differ far more, AND different GPUs (well, their driver's) effectively use different compilers with different optimizations, etc. (Yes, the input is an IL with some optimizations already done, but ultimately codegen is done by the driver.)

You can micro-optimize for a particular CPU- say, not using AVX-512 on Zen, or hand-placing instructions so the latencies all line up juuust right- but it's not nearly as important as for a GPU.

Plus, GPU vendors will work with software devs to optimize the driver just for that software (mostly via a collection of weird workarounds, hacks, feature flags, etc. in the driver, automatically applied when it detects that software running); often the software dev will take, essentially, a bribe from a GPU manufacturer to only work with them.

Then you have stuff like PhysX, CuDNN, etc...

This is what happens when the 3900/3950X are on TSMC's 7nm FINFET while Intel's still on 14nm base Lithography
You're forgetting Intel's trouncing in the Pentium4/Athlon XP days also. It's been a while, but I for one am looking forward to a very competitive marketplace for desktop CPUs.
He was referring to the K7 architecture which includes the Athlon XP brand.
Does anyone know which of the new Ryzen motherboards support PCIe bifurcation?
Ryzen is an incredible processor. I'm still happily running my 1700 like the champ it is.