Review Index:
Feedback

The AMD Ryzen Threadripper 1950X and 1920X Review

Author: Ryan Shrout
Subject: Processors
Manufacturer: AMD

NUMA and UMA - the memory locality concern

The structure and design of the Threadripper creates an interesting situation for AMD. While having two Zen dies on a single CPU works, it means that there are distributed memory controllers and cores, and communication between them is more latent in some instances. Users who are familiar with the intricacies of NUMA on multi-socket systems are already aware of what this means.

To quote from Wikipedia: Non-uniform memory access (NUMA) is a computer memory design used in multiprocessing, where the memory access time depends on the memory location relative to the processor. Under NUMA, a processor can access its own local memory faster than non-local memory (memory local to another processor or memory shared between processors). The benefits of NUMA are limited to particular workloads, notably on servers where the data is often associated strongly with certain tasks or users.

Essentially, because the 8 cores on die 1 (as an example) can access the memory attached to controller on the same die more quickly than it can access the memory on the controller on die 2, it makes sense in some cases for the operating system and applications to be aware of that and to adjust the workload accordingly. But not always.

If you have ever dealt with multi-socket systems (which Threadripper closely emulates) and have had to work around non-NUMA aware applications, you will know of the potential headache it can cause.

Luckily for us, AMD thought ahead on this one and has enabled two different memory access modes to help address any issues of performance or compatibility that might arise for the memory design of Threadripper. Set in either the BIOS or in its Ryzen Master software, AMD will offer users a Distributed (UMA) or a Local (NUMA) memory mode.

  • Distributed Mode
    • Distributed Mode places the system into a Uniform Memory Access (UMA) configuration, which prioritizes even distribution of memory transactions across all available memory channels. Distributing memory transactions in this fashion improves overall memory bandwidth and performance for creative applications that typically place a premium on raw bandwidth. This is the default configuration for the AMD Ryzen™ Threadripper™ CPU reflecting its primary purpose as a creator’s powerhouse.
  • Local Mode
    • Local Mode places the system into a Non-Uniform Memory Access (NUMA) configuration, which allows each die to prioritize transactions within the DIMMs that are physically nearest to the core(s) processing the associated workload. Localizing memory contents to the nearest core(s) improves overall latency for gaming applications that tend to place a premium on fast memory access.

View Full Size

In NUMA/Local mode, the system will report to Windows as having two distinct NUMA nodes (0/1) with 8 cores each. The operating system then attempts to keep workloads that share memory on the same nodes, hoping to reduce the impact of higher latency memory accesses. However, spill over can occur, in both memory capacity and thread capacity. When you exceed the amount of memory on the memory controller on a single NUMA node (say you have 32GB total, 16GB to each die, but your workload uses 20GB), then some memory on the other die will need to be used at the expense of higher latency. If your application can use more than 16 threads (from the 8 cores on a single Zen die), then it will also spill over onto the other die. This situation actually is worse than the memory spill over as it means half of the threads will be accessing memory on the OTHER die the entire time (assuming the workload uses less than 16GB in the above example).

View Full Size

In general, keeping the system in UMA/Distributed mode will result in the best experience for the consumer, especially one that works with highly threaded applications that can utilize the power of the CPU. In this mode, memory is evenly distributed to both memory controllers on both die, meaning that some threads will still access memory across the die (at a higher latency), but on average it will be lower for highly threaded applications.

View Full Size

The primary pain point that AMD hopes to address with the NUMA mode is gaming, where they have identified (as has the community) instances where games can suffer from the longer latencies associated with threads that happen to be placed across the die by Windows or the game itself. AMD says that over a testing regiment of more than 75 games, putting a Threadripper system into the NUMA mode nets an average of +5% in average frame rate, with occasional peaks of 10%. Our testing mirrors that implication, though we didn’t have time to go through 75 games.

There is a third mode for users to be aware of as well, though not directly related to memory access modes. In Legacy Compatibility Mode, the number of available cores is cut in half, with each die having access to 4 cores on the 1950X (its 3 cores each die on the 1920X). AMD says this will give the Threadripper processors performance equivalent to the Ryzen 7 1800X or Ryzen 5 1600X, though you do so at the expense of half the cores you paid for. (At least until you change the setting and reboot.) If you think you will find yourself in this mode for the majority of the time, you’d be better off saving some cash and just buying that Ryzen 7 1800X processor.

AMD found a few games, notably Dirt Rally and Far Cry Primal, which have bugs preventing the application from loading correctly when more than 20 logical cores are detected. You can either enable this legacy mode to play them or disable SMT as well.

Complications around high core count processors will not be unique to AMD and Intel will deal with the same types of issues when its own 12+ core CPUs hit the market later this year. Intel will not have to deal with significant memory access concerns though thanks to its single, monolithic die design. I am interested to see what advantages this may offer Intel Skylake-X.

Testing Core to Core Latency on Threadripper

During the release window of the Ryzen 7 processor, we at PC Perspective used some custom applications to test the real-world latency of various architectures. We found that because of the design of the Zen architecture, with its CCX and Infinity Fabric integration, core to core latency was noticeably longer than in previous multi-core designs from both AMD and Intel. Because the two CCX (core complexes) on each Ryzen die communicated through a unique fabric, the latency between them was higher than cores that exist on each individual CCX. The memory latency between the four cores on each CCX was around 40ns, while the latency between any two cores on opposing CCXs was near 140ns. It was because of this latency that 1080p gaming performance and some other similar, latency dependent workloads took a hit on Ryzen that they did not exhibit on other CPUs.

With Threadripper (and EPYC actually), AMD has another potential hop of memory latency between threads running on different physical die. Let’s see what that looks like.

View Full Size

Okay, there’s a lot going on here and it is reasonable to assert that it’s near impossible to follow every line or every data point it showcases. What is most important to understand is that there are four distinct levels of latency on the Threadripper CPU: per-core, per-CCX, per-die, and cross-die. When running at DDR4 2400 MHz memory speeds (which directly relates to the speed of the Infinity Fabric), the memory latency for threads sharing the same core is ~21ns and for threads on the same CCX about ~48ns. When we cross from a CCX to another CCX on the same physical die, latency jumps to ~143ns, identical to what we measured on the Ryzen 7/5/3 family of CPUs. However, once memory accesses need to cross from one die to the next, latency jumps to over 250ns.

View Full Size

Increasing the memory speed to 3200 MHz shows considerable decreases in memory latency. Our four latencies drop to 20ns for on-die and 45ns for on-CCX; these gains are smaller as they aren’t impacted as much by the Infinity Fabric implementation. Crossing from CCX to CCX though we see latency drops to 125ns (14% faster) and going from die to die shows latency of 203ns (23% faster). These are significant performance gains for Threadripper and indicates that we will see performance advantages to higher clocked memory on multi-threaded workloads that have high memory latency dependencies.

View Full Size

For comparison, here is the same tool run on a dual-socket Xeon E5-2680 v2 platform we happen to have in the office. Based on Ivy Bridge-E, this 4 year old machine has surprisingly similar metrics to the Threadripper processor when it comes to memory latency. Notice that there are only three distinct levels of performance (though there is plenty of variance at the top), showing us an on-core latency, on die, and cross-die result. The QPI interface used to connect the two Intel Xeon processors averages somewhere around 240ns of latency to cross between the two physical sockets.

View Full Size

Finally, here is the look at latency from thread zero across to thread 31 (limited to keep the graph readable, the Xeon remains the same after that). The architecture, die layout, and Infinity Fabric design clearly present a unique arrangement of memory for software and OS developer to work with. AMD will continue to fight the issues around memory latency for its platforms and the move to a multi-die configuration has increased that latency by one more, but still significant, step.


August 10, 2017 | 09:18 AM - Posted by Daniel Moran (garwynn) (not verified)

Pretty much sounds like the messaging was spot on for how they were communicating to consumers what to expect on Threadripper. Just like Ryzen it will offer a great value by leveraging moar cores at a lower price per core.

Now will come the questions on how to put this all to use. What I see coming will be a consumer version of virtualization. I've got more than a few ideas on how to put this system to work, looking forward to getting my hands on it down the road and seeing how much of that actually can happen.

August 10, 2017 | 10:15 AM - Posted by Ryan Shrout

Virtualization is a good way to think through using all these cores. If you want to use 4 cores for a HTPC box, 4 for a NAS, etc., it could be a very flexible system.

August 11, 2017 | 02:24 AM - Posted by quest4glory

Exactly. I can run more simultaneous VMs on a machine with more cores and more memory bandwidth (and PCIe lanes for additional SSDs) than are reasonable on my 5930K. Probably not a 2017 purchase for me, but 2018 is going to be here before you know it.

August 15, 2017 | 11:48 AM - Posted by DiagramsRUz (not verified)

This wikichip refrence materal/link on the Zen microarchitecture has become very detailed and is a great refrence materal compared to any wikipedia entry on the subject matter. The Zen block diagrams are as detailed as I have ever seen concerning Zen functional units and it will make for a great refrence source to have bookmarked. Plenty of cross refrences to all AMD's and others similar products with web based methods of sorting by attributes.

https://en.wikichip.org/wiki/amd/microarchitectures/zen

August 10, 2017 | 09:32 AM - Posted by Thisguy (not verified)

Very good write up, any reason your not including Ryzen 5 in your Perf/$?

Only reason I ask, your including the I5 which holds its own in the Perf/$ agianst just about everything else.

August 10, 2017 | 10:14 AM - Posted by Ryan Shrout

Just spacing and timing honestly. I think the Core i3 and Pentium would hold up well there too but they aren't really in the same performance class.

August 10, 2017 | 10:45 AM - Posted by Thisguy (not verified)

Thank you for the reply, and thank you for the great coverage.

August 10, 2017 | 09:33 AM - Posted by Anonimo (not verified)

Is this the i9-7900X review ?

August 10, 2017 | 09:38 AM - Posted by Jabbadap

Heh had the same thought when reading synthetic benches.
But probably it's just still in writing over old i9 7900x review, reviews are usually rushed to make in deadline(i.e. anandtech's review was pretty mess too right after nda lift).

August 10, 2017 | 09:52 AM - Posted by Honza (not verified)

They probably forgot to replace the commentary text that goes with the graphs. Or didn't manage to do it on time.

August 10, 2017 | 10:14 AM - Posted by Ryan Shrout

Every editor has been fired, but the inline text has been fixed. ;)

August 10, 2017 | 09:39 AM - Posted by John H (not verified)

Need context. 37% and 30% ... faster ...

August 10, 2017 | 09:45 AM - Posted by derz

smells like freedom

August 10, 2017 | 10:06 AM - Posted by dagnamit

Welcome back AMD and thanks for stopping the race to the bottom by taking a stab directly at Intel's high-end profit margins. I knew that Intel holding back extra cores for nearly a decade would come back to bite them in the ass. If AMD's next gen can increase IPC and 15% clock boost we'll have some truly ridiculous competition in this space again.

Goddam I can't remember the last time there's been this much excitement about a computer component release. It's the first truly competitive part AMD has put out since probably the HD 7900 series GPU's, over 5 years ago. I'm due for an upgrade within a year, so the timing of all this is simply outstanding.

August 10, 2017 | 10:19 AM - Posted by WhereIsThreadripperz (not verified)

From: "Media Encoding and Rendering" page!

"A classic test of multi-threaded capability, the Core i9-7900X sees gains of 15% over the Core i7-6950X and 35% over the Ryzen 7 1800X." [It's a Threadripper review what about TR's performance]

"Here is another instance of strong single thread performance improvements allowing the Core i9-7900X to match the performance of the Core i7-7700K, the previous single-threaded king. A 25% improvement over the 6950X gets us a score of 197. On the multi-threaded result the 7900X has an advantage 15% over the 6950X and a 34% lead over the Ryzen 7 1800X." [It's a Threadripper review what about TR's performance]

"Similar to the CineBench R15 scores above, POV-Ray gives the new 7900X a 22% lead over the 6950X and a 38% lead over AMD’s Ryzen 7 1800X." [It's a Threadripper review what about TR's performance]

"We run two different Blender workloads, and both show interesting data for the Core i9-7900X. While in the BMW render we see a 15% advantage for the multi-threaded performance of Skylake-X over the 6950X, that lead is only 6% in the much longer Gooseberry workload, as it spends more time in procedural calls." [It's a Threadripper review what about TR's performance]

This is a Threadripper review and all there is on the "Media Encoding and Rendering" page Text Wize is Ryzen 7 1800X against Intel, or Intel against Intel comparsions!
What is with these comparsions in a Threadripper Review and these the non related to Threadripper notes!

August 10, 2017 | 10:31 AM - Posted by Ryan Shrout

This has all been fixed and updated. Sorry!

August 10, 2017 | 10:30 AM - Posted by KamoteMan (not verified)

why didn't you put the overclock power consumption of intel in your chart with oc threadripper power consumption?

in hardware unbox review threadripper is -10% lower power consumption compared to intel when overclock.

August 10, 2017 | 10:38 AM - Posted by Johan (not verified)

Thanks for giving a much better all round review than what Anandtech did. They are clearly in Intel's pockets. That is such a biased review.

August 10, 2017 | 11:09 AM - Posted by cpucrust (not verified)

Thanks for the awesome review - PCPer is continually improving.
Great to see all this wonderful competition from AMD.
I'm firmly in the content creator use case and gaming secondly.

Was close to pulling the trigger on a 1700/1700X/1800X system to replace my Intel 2500K system. I may go TR 1920X/1950X now, however the price of X399 is a bit high (new product pricing).

Looking forward to continuing TR coverage - thanks!

August 10, 2017 | 11:12 AM - Posted by Ryan Shrout

Thanks for reading!

August 10, 2017 | 11:11 AM - Posted by Kilobytez95 (not verified)

285 watts? Really? Can motherboard VRMs even supply that kind of power? Surely these TR4 platform boards know that thread ripper consumes allot of power so i guess they maybe build in a VRM that can supply a little bit extra than spec just incase but 285 seems crazy. I don't know if I'd be comfortable putting that much power through my VRMs every day. I probably just wouldn't overclock the chip until after i upgraded it years later. Maybe I'm wrong and the boards really can handle that kind of power. I also wouldn't run any AMD VEGA GUPs unless I had a very beefy PSU like 1250 watts or higher.

August 10, 2017 | 04:34 PM - Posted by Anonymous (not verified)

You wouldn't want to run that on a motherboard with lower-end power delivery, but that kind of wattage is not anything special for overclocking. For a year or so, I ran a 6-core Phenom II at 1.6v core/1.4v NB which used over 300w on just the CPU, and it survived fine even on a midrange 8-phase motherboard (Sabertooth 990FX). However, someone else with similar hardware blew out their VRM with 1.65v so this is probably about the limit for 8-phase boards. The ASUS threadripper board used in this review seems to be 8-phase, but there's no reason why there couldn't be more robust x399 boards for heavy overclocking.

The larger concern, I think, is how much voltage the CPUs can take before silicon degradation becomes a serious issue, which is a lot harder to test and a lot harder to address, but 1.4v shouldn't be too bad.

August 11, 2017 | 12:47 AM - Posted by James

How many graphics cards take close to 300 Watts? I don't know why people don't think of this. That is just like everyone suggesting that you must have water cooling for TR. Most video cards, even high end cards close to the 300 Watt range, just use an air cooler. A CPU socket can actually use a much bigger air cooler than a graphics card. When you get into overclocking, the power density can get high enough that you need the temperature differential that only a water cooler can supply. Standard clock TR is just two Ryzen die, so the thermal density is actually the same as Ryzen or lower if the clocks are lower. Water cooling should not be required unless you are overclocking significantly.

I am curious as to what the power consumption limits of the Epyc socket are. There has been talk of an HPC APU from AMD. That will probably be in an Epyc socket, so it will need to supply a large amount of power. The APU is rumored be two zeppelin die connected to a GPU with 4 links. The GPU will need to use HBM memory. I don't know if a full Vega die will fit on the MCM though. Four links (close to 100 GB/s?) with high bandwidth cache could make an exceptional HPC device. If you think about it though, it will essentially be TR combined with a powerful GPU in one socket. Even with lower clocks it is going to take a lot of power. It will also not be cheap.

August 14, 2017 | 04:11 AM - Posted by msroadkill612

APUs (ie. DIFFERENT processors linked by fabric) roots will be in low power mobile, meaning a minimalist single zen 4 core ccx & a single vega gpu on a single ~zeppelin type die. Cost effective Raven Ridge desktops are intended as minimalist 4C & 1x gpu.

If so, its a very different die from dual ccx zeppelin, and 4 core ryzens, which are 2x ccx w/ 4 cores disabled.

We cannot be certain HEDT APUs wont be multiples of this ~mobile die/mask.

August 10, 2017 | 11:34 AM - Posted by Clmentoz (not verified)

I'm wanting to see some workstation/productivity benchmarks done on any Epyc single socket SKUs and compared to Threadripper just to see what advantages there are to having 8 memory channels as opposed to only 4 for workstation/productivity workloads.

And the options for populating 8 memory channels with more DIMMS(at one DIMM per channel) across 8 memory channels for larger memory capacities able to be run at the maximum supported clock rates is what has me wanting to see some Epyc 16 and 24 core single socket SKUs directly compared to Threadripper on some workstation/productivity workloads.

The Epyc 7401P at 24 cores/48 threads should be price wise, CPU SKU to CPU SKU, a relatively direct comparsion as the Epyc 7401P costs $1075(Listed cost) compared to the cost of the TR(16 core/32 threads) 1950X's $999(Listed price). And it very much looks like AMD's Epyc customers can look forward to even better more affordable pricing from AMD on the professional Epyc SKU's compared to even AMD's consumer SKUs, and not only AMD's MUCH more affordable Epyc pricing compared to Intel's really costly Xeon high core count SKUs.

So feature wise and price wise there is not any economic reasons for any potential AMD workstation customer to be forced to purchase a consumer variant just to be able to have affordable options because the professional variants from AMD for single socket workstation/productivity workloads are actually the better deal ecomomically speaking.

That Epyc single socket 7401P SKU has with its 24 core/48 threads, 8 memory channels, and 128 PCIe lanes feature set somsthing that really needs to be looked at by anyone contemplating purchasing Threadripper if they only are needing to utilze the TR processor for workstation/productivity only workloads. Gaming workloads are another matter, but AMD has priced it's single socket Epyc workstation SKUs so low that it's damn incredible considering having that 7401P's 24 core/48 threads and the 8 memory channels and 128 PCIe lanes supported features for workstation/productivity workloads at that price point.

August 10, 2017 | 11:44 AM - Posted by cpucrust (not verified)

Some very good points - especially the potential price/performance of Epyc over TR. Perhaps my content creator and VM workstation plans might be better served with going Epyc over TR. Can just keep my 2500K for gaming use cases - rarely play AAA titles anyways when I have time for gaming.

August 10, 2017 | 12:18 PM - Posted by Mr.Gold (not verified)

Power consumption seem to be an ASUS motherboard problem ?
Your conclusions and chart might only be based and applicable to the ASUS motherboard.

check anandtech CPU power usage (cant link)

The 1950x consume 10w extra Vs 7900x at full load, and 3w at idle.

August 10, 2017 | 02:52 PM - Posted by VariablesAndUnknowns (not verified)

The 7900x is a 10 core/20 thread part while the TR 1950x is a 16 core/32 thread part. So that 10W extra at full load compared to the Intel part comparsion is a little more complicated than what you may be thinking about only the motherboards being different. And the TR 1950X is only using 10Watts more power uasge for 6/12 more cores/threads for TR(16/32) over the 10/20 cores/threads of the 7900x says some but not all about AMD's Threadripper power usage metrics.

Those big AVX units on Intel's SKUs can really drink up the power. And there is the question of the non linear scaling of power usage with increasing clock speeds. Threadripper is doing fine with the power usage for a processor with its core/thread counts and it's going to be more about what overclocks can be achived on each makers' respective 14nm process, Intel's 14nm process vs GF's(licensed from Samsing) 14nm process for AMD's TR.

Then there is the AMD vs Intel motherboard power usage question that is somewhat valid and there does need to be more attempts at getting at the ASIC power usage metric seperated from the other system power usage metrics in order to get a better look at things.

August 10, 2017 | 12:26 PM - Posted by vyvyvv6565898 (not verified)

In the fourth from last paragraph in the conclusion shouldn't " the 32-thread 1950X will likely over a sizeable performance advantage "

shouldn't over be offer?

August 10, 2017 | 12:44 PM - Posted by Power (not verified)

PCper can proudly go back to the name with AMD's moniker in it.

August 10, 2017 | 12:51 PM - Posted by Alansmithee (not verified)

Ryan, in regard to PCIe lanes, you mention 64 lanes for TR in the summary without the caveat that it's actually 60 net lanes. This perpetuates the myth that it's 64 TR lanes vs. 44 Intel lanes where it's really 60 vs. 44 since Intel's spec excludes DMI lanes.

While TR having 60 lanes is great in concept, if you look at all motherboards that exist for both platforms, TR has a maximum of 48 lanes to PCIe slots, while Intel has a maximum of 44 lanes to slots.

Both have 3 M.2 slots and yes, on TR the M.2 slots are plumbed to the CPU which is better for latency, but completely sacrifices RAID capability as the trade-off. And the slot arrangement for all TR launch motherboards is not very useful because two slots are hard wired to x8 and there are basically no x8 cards outside of server applications (now that 10GbE cards have gone to x4 3.0). It would have been more useful to split those lanes into more x4 slots and dispense with the chipset-connected slots.

Or better yet, it would be nice to have a PLX switched board for either TR or X299 along the lines of the Asus X99-E WS, so you could have 7 x16 slots with complete slot flexibility no matter what width of GPUs you were using.

August 10, 2017 | 03:56 PM - Posted by VariablesAndUnknowns (not verified)

That "maximum of 48" says a lot about Intel's platform with all of AMD's x399/TR SKUs having the same CPU supplied "maximum" with respect to any of AMD's Threadripper parts/SKUs offering the same 64 PCIe lanes directly from the Threadripper SOC/MCM with 4 of those PCIe lanes providing for the chip's Ethernet, USB, SATA, and other such functionality. AMD has no "Maximum" in the sense that all TR platform SKUs are not artificially segmented with respect to PCIe lane offerings on any of AMD's TR platform offerings. And that's 60 PCIe lanes out of 64 for the Motherboard makers to work with to provide for whatever slot requirements the motherboard's design is targeting.

Maybe for both the AMD platform and Intel platform/platform segments listed in the article those total and available PCIe lane metrics need to be included along with any Intel DMI connectivity that may support any Ethernet/USB/Other similar functionality. But keeping track of all the Different MB SKUs and their varying slots/other offerings will be a mostly ongoing process that can never be quite pinned down as far as to what exact usage/functionality is provided by the motherboard's design. So that leaves just the fixed/non fixed PCIe lane functionality provided from the CPU/MCM with metrics that can be pinned down and reported on with more certainty.

I'm also interested in AMD's Infinity Fabric(IF) off MCM connectivity options on any specialized motherboard SKUs that may be able to use the IF protocol for off MCM communication at a faster effective bandwidth than even PCIe 3.0 offers. I wonder if that off MCM Infinity Fabric usage is going to be limited at first to the Epyc CPU/motherboard ecosystem for any direct attached GPU accelerator usage in the workstation/server markets. I hope that some single socket Epyc MBs/"P"(single socket) branded Epyc CPU SKU systems will have that sort of Infinity Fabric off MCM functionality available for workstation usage.

August 10, 2017 | 01:26 PM - Posted by Johan (not verified)

Thanks for not being useless like Anandtech. They clearly don't understand Threadripper and it looks like they are paid by Intel to do so.

August 11, 2017 | 12:04 PM - Posted by Thatman007

Are you joking? Anandtech is actually reliable. PcPer are just being paid by AMD. ThreadRipper is crap. Its power hungry and isn't even good in gaming.

August 11, 2017 | 02:57 PM - Posted by Jeremy Hellstrom

Stop stealing my damn cheques.  I'd love to finally find out how much they are paying me.

August 20, 2017 | 04:39 PM - Posted by LG (not verified)

Funny, I would say the same about you. Eat some bran.

August 10, 2017 | 01:39 PM - Posted by Thedarklord

Liked the review, still very interested to see how Zen matures over the coming months and years. I do a good amount of virtualization and some content creation, but gaming is first, so might still be leaning on Skylake-X, but not sure yet.

One minor request, on the graphs comparing performance, can we label the bars or add a second legend to the top as well? Easier for viewing/comparing bars/colors.

August 10, 2017 | 01:54 PM - Posted by Thedarklord

How about a few tests comparing Threadripper vs Skylake-X running at the same frequency? (With no turbo boosting)

Would be interesting to see how they compare running at the same speed, and I am wondering if Intel's advantage in lightly threaded/single threaded is coming from a architectural benefit or more of a pure clock speed advantage.

August 10, 2017 | 06:51 PM - Posted by Your Name (not verified)

I don't understand this kind of analysis. We know Intel has better IPC. But nobody buys a computer and clocks it equal to some notional competitor!

August 11, 2017 | 01:03 AM - Posted by James

It is interesting that Intel is essentially competing on clock speed. I don't think the core IPC is really much higher. Intel probably still has slightly better caches that may be responsible for the apparent higher IPC that isn't explained by higher clocks.

August 10, 2017 | 04:04 PM - Posted by Oren (not verified)

I know you are busy, especially with the VEGA launch in a few days. But I hope you could do something for your readers. At some point soon, can you test ThreadRipper vs Ryzen?I (and perhaps many others) who are on the fence regarding Threadripper really want to know the real world difference the Threadripper brings vs Ryzen.

I wonder how much a difference running crossfire GPUs in full x16/x16 mode with QUAD channel memory really compares to running the same GPUs in x8/x8 mode with DUAL channel memory. You could set up a test where an 1800X competes with a Threadripper where only 8 cores are active. Same amount of RAM. See what real world differences the architecture brings in that tight of a scenario.

Good work on this review.

August 10, 2017 | 04:36 PM - Posted by skysaberx8x

I'm very curious on how will the two dies and memory modes affect virtualization? I've only experimented with VM in the past but is it possible to run two Hexa-cores windows VM and with each individual memory nodes assigned to each VM?

August 10, 2017 | 05:29 PM - Posted by SerinJ (not verified)

Are you setting the Blender tile sizes to 256 or 16/32?
Just wondering since an overclocked 5960x gets 1 minute 30 seconds on the BMW at 16x16 tile size. Significant difference that shouldn't just be a result of the OC.
For reference: 256 or 512 are for GPU and 16 or 32 are for CPU - at least for getting the best and generally more comparable results to what we get over at BlenderArtists.

August 10, 2017 | 05:30 PM - Posted by dan the grammar man (not verified)

When reading is not enough, the mistakes are OVER 9000!

"If you content creation is your livelihood or your passion, "

" as consumers in this space are often will to pay more"

" Anyone itching to speed some coin"

" flagship status will be impressed by what the purchase."

" but allows for the same connectivity support that the higher priced CPUs."

""""Editor""""

August 11, 2017 | 10:28 AM - Posted by Ryan Shrout

Now just point me to the pages... ;)

August 10, 2017 | 06:42 PM - Posted by Your Name (not verified)

Nice to see a review with more than a bunch of games tested. Keep up the good work!

August 10, 2017 | 06:49 PM - Posted by kenjo

Should not a test like 7-zip use 32 threads as max since that is what is presented to the OS??

now it only uses 50% of the threads on TR but 80% on i9-7900x.

August 10, 2017 | 09:21 PM - Posted by #650cores (not verified)

Silly performance, looking forward to the 1900X and maybe 1900.

August 11, 2017 | 03:50 AM - Posted by mat9v

I sometimes wonder why nobody ever points out that within CCX (4 cores that can allow a lot of games to run comfortably) ZEN has latencies of half those of Intel CPUs. Binding a game to those 4 cores (8 threads like any i7) has significant impact on performance. It does not change memory latencies of course but core to core is much better.

August 11, 2017 | 11:25 AM - Posted by Anonymous28 (not verified)

I'm glad someone else noticed this besides myself. I noted this during the Ryzen launch & quickly noted that by using CPU affinity along w\ CPU priority to force my games to run exclusively within 1 CCX & take advantage of using high CPU processing time on these same CPU cores I could take advantage of this up to a point.

What all this shows to me is that the OS & game developers software need to be revised to better handle this architecture at the core logic level instead of users\AMD having to provide\use methods to try to do this that cannot be used in a more dynamic fashion. I've ran some testing on Win 10's Game Mode & discovered that MS is actually trying to use CPU affinity to dynamically set running game threads to be run on the fastest\lowest latency CPU cores to "optimize" game output thru the CPU but it still tends to cross the CPU CCX's at times if left on it's own.

What I've found is by doing this my games run much smoother w\ a lot less variance which gives the "feel" of games running faster (actual FPS is the same) due to lower input lag & much better GPU frametime variance graph lines w\ very few spikes....essentially a fairly flat GPU frametime variance line which is what you want to achieve performance-wise.

Just to note....my box is running an AMD Ryzen 7 1800X CPU\Sapphire R9 Fury X graphics card w\ no OC's applied to either the CPU or GPU.

It's a step in the right direction but needs more refinement at the OS level......

August 11, 2017 | 12:06 PM - Posted by Thatman007

As expected, performance per dollar is crap in single threaded tasks, which most workloads are. Games don't even use more than 1 or 2 cores.

August 12, 2017 | 10:47 AM - Posted by ID10T (not verified)

Yea games only use 2 cores lol

http://i.imgur.com/Hg3Ev5p.png

August 20, 2017 | 05:59 PM - Posted by LG (not verified)

And "as expected", we have yet another Intel shill complaining about gaming performance on a production CPU, which isn't made for gaming (although it's not bad in the least and has a longer future as devs code for more than Intel's tiny core count (under $1000))..

-"performance per dollar is crap in single threaded workloads"...
Well, since these aren't sold as a single or dual core CPU, performance per dollar as a unit is beyond everything on Intel's menu.

- "Games don't even use more than 1 or 2 cores"
Well, I've been using a FX-8350 for 2 years now, and I always see all 8 cores loaded up on every single game I play (and I have many). Windows 10 makes use of these cores even when it's not coded in programs. It would work even better if devs started coding for at least 8 cores, and I believe they will start doing this in earnest now that 8-core CPUs are now considered average core counts (unless you're with Intel).

You would have been better off stating that core vs core is in Intel's favor on the 4-core chips and some others, but ironically the "performance per dollar", as you mention is superior with AMD.. in every way.

August 12, 2017 | 06:28 AM - Posted by Anonymous37 (not verified)

What memory are you using, and could you name a 64GB kit that works in XMP? And why 3200Mhz over 3600?

August 16, 2017 | 12:09 AM - Posted by Anonymousdfdf3 (not verified)

Intel is still superior both in raw performance and in perf/$. If you were being objective you wouldn't have given slapped an editor's choice on this inferior product.

August 20, 2017 | 03:24 PM - Posted by Rotorheadman (not verified)

In Handbrake the 1800x is 40% slower than the 1950x and in reverse the 1950x is 67% faster than 1800x.

August 20, 2017 | 06:09 PM - Posted by LG (not verified)

Open cinebench with a TR or even an 1800x. Show me any Intel chip that can come within 20% of the 1950x. The entire Ryzen 7 lineup is king of the "perf/$" category. 1800x = $365 on eBay right now. Look how close it matches with Intel products that are double the price or worse.

If you want to compare single core perf vs Intel, you can win an argument.. at the cost of very high power draw and even worse cash draw. Perf/$ is a dead argument for any Intel fanboy. Find something else. BTW, are you also commenting under "Thatman007" or something? Sound like the same Intel mouthpiece.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.