Synthetics and Scientific
SiSoft Sandra 2017
AMD's advantage in memory can be seen in the Sandra memory bandwidth test. With it's adherence to the JEDEC DDR4-2933 memory standard, the Ryzen 2000 series processors provide a 20% improvement in memory bandwidth over the Intel Coffee Lake parts. AMD has done impressive work with memory compatibility on Ryzen in the last year, and an out of box memory frequency of 2933MHz is quite impressive.
Geekbench 4.2.0
As expected, the additional threads of the Zen+ processors add up to large gains in heavily multi-threaded workloads. However, in the Geekbench single-threaded tests, the i7-8700K pulls away from the R7 2700X with a 17% higher score.
With the addition of 2 cores (4 threads), the R7 2700X is able to overcome a 14% deficit in single-threaded performance and turn it into a 4% overall gain in multi-threaded performance. The R5 2600X sees a similar gain over the i5-8600K in multi-threaded workloads.
Generationally, the R7 2700X sees a 5% improvement in single-threaded Geekbench performance and a 10% increase in multi-threaded performance over the previous flagship R7 1800X CPU.
Euler3D
While the i7-8700K outperforms the Ryzen 7 2700X in the Euler3D benchmark, the AMD option remains competitive, with just a 5% performance gap with the benchmark running at 16 threads.
So how come the i7 8700k got
So how come the i7 8700k got the full 3200MHz in its review? Also why didn’t you test the i7 8700k and the i5 8400 with their stock coolers?
We actually tested the 8700K
We actually tested the 8700K at 2400MHz in the initial review, I'll have to correct the memory section of that article to more accurately reflect that despite using a DDR4-3200 kit, it was running at a slower frequency.
As for stock coolers, the i7-8700K does not come with one, and the i5-8400 we recieved from Intel near the Coffee Lake launch was not in retail packaging, so we don't have the exact stock cooler it shipped with.
Cool then.
Just an idea for
Cool then.
Just an idea for some future testing: I think a comparison of 8400 vs 2600 (non X) with a B3xx motherboard and the stock coolers would be really interesting, if you are up for it. For some reason AMD didn’t sent the 2600 (with the small cooler) to any reviewer. I think something could be hiding there.
Did you do the tests with all
Did you do the tests with all security patches and bios updates for Spetre and Meltdown?
Windows 10 was fully patched
Windows 10 was fully patched with all security updates until the beginning of April when we started this testing. Additionally, all motherboards were on their latest revision of UEFI, which for Z370 and Z270 included the Spectre patches.
I’ve updated the article with
I've updated the article with exact BIOS revisions, as well as to reflect the patched state of Windows 10. Hope this helps clear up any potential confusion!
This confirms testing was not
This confirms testing was not performed with Microsoft’s April Patch Tuesday (April 10) Spectre V2 mitigations for AMD CPUs then. I know as well some AMD motherboard OEMs have not issued accompanying microcode updates/mitigations as well, even today (Andandtech mentioned MSI as one OEM who hasn’t yet).
I think there’s quite a bit of variation in different sites reviews due to this mess of who has what update and when, but at least it doesn’t change the overall takeaway. I think I’ll probably go AMD for my next build, but my 4670k is still going strong.
It appears while we did not
It appears while we did not have the patch from April 10th, it shouldn't have been an issue. If you look at the Microsoft support page discussing the Spectre/Meltdown patches, they have a section about the AMD Meltdown 2 update (scroll down to the bottom of the page.
Here, they give instructions on how to enable the Indirect Branch Prediction Barrier (IBPB) in the registry. It appears that even if you have the patch installed, this mitigation factor isn't enabled by default (at least not yet).
I'm going to run some tests this afternoon with IBPB turned on to see what kind of difference it makes to be sure though!
You’re right I think,
You’re right I think, Microsoft’s wording is somewhat unclear to me. AMD’s own security page flat out says “Microsoft is releasing an operating system update containing Variant 2 (Spectre) mitigations for AMD users running Windows 10 (version 1709) today.” Not ‘it’s disabled by default and you need to enable it in the registry.’ Hmmm… Anway, please do make note if any real difference is seen. Thanks.
We got some additional
We got some additional clarification from AMD about this earlier today.
With the April 10th patch, IPBP is enabled on platforms that also have the appropriate firmware support (which all X470 boards have). The MSDN document is apparently referring to enabling IBRS, which is not recommended by AMD as a necessary mitigation and provides a potential hit to performance.
Quick testing in both of these states saw no performance effect on our normal CPU and gaming benchmarks as far as we could find. Going forward, we will be using the April 10th patch without the additional registry keys enabling IPRS for Ryzen testing.
Hmmm but what about the
Hmmm but what about the Spectre2 patches?
It looks like over at Anandtech that a secured Intel system doesn’t do so well?
What about XFR2?
2 remarks :
1) I might have
2 remarks :
1) I might have missed it but which bios was used on platforms , i think you need PR AGESA 1.0.0.2a for amd-spectre , and well on intel I have no clue.
2) something silly , but really the 2700X blue graph bars and the intel part orange , xD
Curious that the idle power
Curious that the idle power consumption hasn’t improved one bit.
I wonder if the new fabrication process actually does have a smaller feature size, or if it’s a tweaked version of the old process with feature size measured differently.
It’s pretty well understood
It’s pretty well understood that this is more “12nm” than 12nm. Individual feature sizes haven’t really changed that much for a few processes now, I think.
As for the idle power consumption, it’s so low for modern CPUs (in the <10W range) that the system idle power is likely to be dominated by other components, meaning that any gains by the CPU are unlikely to be noticed.
What format was the Handbrake
What format was the Handbrake test output? H.265?
The final export was H.264. I
The final export was H.264. I updated the article with clarification of this. Thanks!
I’m going to guess from the
I’m going to guess from the context that whenever “inter-CCX” is used, what is really meant is intra-CCX.
The information for the 2600X
The information for the 2600X is wrong or you actually tested a 2600 and not a 2600X. For example, X is not 65W TDP and doesn’t come with Wraith Stealth, it comes with the Spire.
Good catch with the Wraith
Good catch with the Wraith Spire/Stealth confusion, I misspoke when I referred to the name of the cooler, and this has been changed in the review. The pictured cooler in the review is, in fact, the Wraith Spire.
You are also correct about the TDP of the R5 2600X. It appears that AMD's reviewer's guide has a typo in one of the specs tables, calling it a 65W part while another specs table lists it as 95W. The review will be updated to reflect the accurate TDP of the 2600X. This TDP change does not change the outcome of our review.
Thank you for bringing these issues to our attention!
In “For Honor”, why do lower
In “For Honor”, why do lower end processors outperform higher end processors? The Ryzen 5 has what looks like significantly higher performance compared to Ryzen 7. That looks like it is true for intel parts and Ryzen 1xxx parts also, although those may be in the margin of error. This points to maybe some multithreading problem in this game or perhaps some problem with your test.
Also, I would like to see benchmarks with an AMD graphics card. The game testing is, in large part, testing the Nvidia driver on AMD CPUs.
Just getting started on the
Just getting started on the article – but a correction for your specifications table: the TDP of the Ryzen 1600X is 95 watts, not 65 watts (I know – I have one). See: https://www.amd.com/en/products/cpu/amd-ryzen-5-1600x
They haven’t increased the TDP between the 1600X/2600X.
Thanks for the correction,
Thanks for the correction, noted and fixed!
good review and all, give
good review and all, give more insight over the many many sites that are doing this with ryzen 2 (crappy naming from AMD, they should have made sure folks called it as Ryzen 2xxx so that when the proper Ryzen 2 comes out it can be called Ryzen 2 likely model 3000)
not sure why some sites show these walking all over the various intel models whereas others (including this one to a point) are really painting the new chips far more power hungry and less performance then they likely actually have.
bias much?
anyways, I myself quite like the look of 2600 over the 1600, ~10% more expensive but easily a good 15% quicker as well , 2600x vs 1600x ~12% more expensive seems about that much quicker as well so it works out.
also for the limited overclock potential, I do not know about this as many other sites claimed to be able to get closer or above 5Ghz with a substantial cooler but the included coolers at least pretty much allow them to get near clock wall vs pricey AIO coolers compared to first gen which seemed that much more limited.
kudos to the mainboard makers as they did a bang up job on clock tuning for memory where you are pretty much just use it as it is manually tuning the memory seems to just not be worth doing compared to gen 1 (seems to reduce performance for nothing compared to just plug it in and go)
seems x400 vs x300 and Ryzen 2000 vs Ryzen 1000 they really have done a terrific job of fine tuning them in not even 1 year for a “refresh” kudos to AMD.
anyways, they “seem” power hungry, maybe but take that Intel vs AMD bias away (which is very hard to do) and the performance given, the build quality of them, the ability to run at decent temperatures vs frying themselves in the process Ryzen/Ryzen 2000 are AMAZING, full stop.
You wouldn’t have a link
You wouldn’t have a link handy to those 5Ghz? Or nearly 5Ghz.
All I am aware of is this one guy using LN2
Which doesn’t means much since the 1800X was clocked nearly as high using LN2 back in the day
Who runs their i7-8700K at
Who runs their i7-8700K at stock? While I understand there needs to be some standard basis in comparison the i7 easily boosts to 4.7 with just 1 setting in bios. Considering the new Ryzens are already hitting the speed ceiling it’s clear the testing differentials are understated.
I’m often told many people
I’m often told many people keep they’re K parts at stock speed.
Might also be a reasonable assumption, given that K parts have a higher boost than non K parts
Btw
A few hundred Mhz more are not going to matter a lot (4.3 vs 4.7)
What would is a 5.2 OC on a 8700K, cause that’s a cool Ghz more on all cores than Ryzen 2 can manage (it seems)
though this is even more niche than normal OC’ing since you need to delid; thx Intel (or buy it that way for an even bigger premium price)
Stock on the i7-8700K is 3.7
Stock on the i7-8700K is 3.7 so it makes a HUGE difference. This is a CPU begging to be overclocked. I’m pulling 4.9 air cooled.
Max boost for 1 core is
Max boost for 1 core is 4.7Ghz
I know using more cores you get lower boost, but at worst its 4.3 for all core boost
Using multicore enhancement can lead to instability if the voltage doesn’t scale up with it
3.7 might be base frequency, but it’s always boosting as much as it can.
That’s stock behavior.
K parts have higher/more aggresive boost than non K parts
That’s a good reason for many to buy a K CPU and to leave as is
It’ll boost to 4.7 when you need it the most
Btw
Leaving everything at stock also means base frequency is actually 800Mhz 😀
In a way
When the CPU is idle it clocks down and ramps up to 4.7 when needed
MCP is no big deal with
MCP is no big deal with proper settings and cooling. Despite some AMD fans who would dismiss a 400 mhz clock difference as negligible the fact is it is significant (some are extolling the 200 mhz diff with Ryzen2 over Ryzen1) plus 4.7 is pretty much an easy speed to begin with. The benchmarks here show a nearly crippled K series cpu capable of more against a cpu with little to no extra overheard.
Of course all i7 cores clock at the same speed when pushed with the right settings and I can’t wait to see overclocked scores against the Ryzen. Probably no contest except for some multi-thread tests. I’m fully aware of power throttling but it’s not relevant when it comes to this article. I’m glad to see AMD competitive again, just show me the tests when both are pushed.
Also get yourself a rockit 88
Also get yourself a rockit 88 or some other means of de-lidding and you pull 5.2 on high end air
These CPU’s are begging to be OC’ed higher than 4.9 😀
Nah, there’s diminishing
Nah, there’s diminishing returns over 4.8 in cooling costs and effort over performance. 5.2 stable is no guarantee either but hey, we can always hope. Then again one can always pimp their RealBench scores and say it was stable for a few hours when in fact no one but you knows it blue screened in a minute lol.
That would make comparing
That would make comparing them difficult due to the extra cost of cooling. Ryzen 2 parts all current come with a good quality cooler, much better that the junk coolers intel ships with their parts, if they even ship one If you start overclocking the intel parts, you could drive the cost up significantly. How much does a water cooler cost these days? Also, overclocking is of questionable value in many cases. Even at stock speeds, you will often be GPU limited.
Are all Ryzen 2 parts using
Are all Ryzen 2 parts using TIM instead of solder this time?
As far as I know the 2400g isn’t soldered.
Obviously might hinder OC ability
No, the Ryzen APU’s were not
No, the Ryzen APU’s were not soldered solely for the purposes of cost savings. Regular Ryzen CPU’s are all soldered from the factory my friend, none of this “toothpaste TIM” garbage to see here 🙂
Can I have the 2700x,
Can I have the 2700x, motherboard and RAM when you’re done? I want to confirm your results.
…
Ken, did you use the standard
Ken, did you use the standard Windows 10 power profile for the Ryzen 2000 parts?
Are there any plans for clock
Are there any plans for clock for clock comparisons? Thank you.
So what I get from this is
So what I get from this is for an extra $20 from AMD 2700X to Intel 8700X I can get better performance (for most standard stuff) all around with less power consumption. If the price difference was say $100 that would be different but since the difference is so marginal I will stick with Intel. Sorry AMD you did not make me jump ship on this one!!
You conveniently forgot to
You conveniently forgot to factor in the cost of the cooling solution that comes boxed with the The 2700X. And that’s rather disingenous with that “(for most standard stuff)” for the Intel part as the 2700X offers 2 more cores and 4 more threads and did you go look at GN’s game streaming videos at higher resolution on the Ryzen 7 2700X and see that Intel part stuttering like a Model T.
Really you do not appear to be able to do the most simple of Cost/Benefit analysis and factor in that boxed cooling solution or the Cost/Per-core math that is lower still on that Ryzen 7 2700X SKUs. Even the productivity scores are there with the Ryzen 7 2700X as well as some aother usual workloads where the 2700X comes out ahead performance and value wise.
What kind of cooling solution comes with that Intel Part compared to what AMD provides with the 2700X for that very same MSRP, at no extra charge.
So what I get from this is
So what I get from this is for an extra $20 from AMD 2700X to Intel 8700X I can get better performance (for most standard stuff) all around with less power consumption. If the price difference was say $100 that would be different but since the difference is so marginal I will stick with Intel. Sorry AMD you did not make me jump ship on this one!!
You conveniently forgot to
You conveniently forgot to factor in the cost of the cooling solution that comes boxed with the The 2700X. And that’s rather disingenous with that “(for most standard stuff)” for the Intel part as the 2700X offers 2 more cores and 4 more threads and did you go look at GN’s game streaming videos at higher resolution on the Ryzen 7 2700X and see that Intel part stuttering like a Model T.
Really you do not appear to be able to do the most simple of Cost/Benefit analysis and factor in that boxed cooling solution or the Cost/Per-core math that is lower still on that Ryzen 7 2700X SKUs. Even the productivity scores are there with the Ryzen 7 2700X as well as some aother usual workloads where the 2700X comes out ahead performance and value wise.
What kind of cooling solution comes with that Intel Part compared to what AMD provides with the 2700X for that very same MSRP, at no extra charge.
(Did you reply twice to a
(Did you reply twice to a double post… really? XD) Your logic is funny, what about if he already owns a good cooler and couldn’t care less about streaming?
Just to make sure my reply
Just to make sure my reply was not deleted along with one of your superfluous posts. Game Streaming is just one area where Ryzen 7 2700X is better. And your other nonsense was also called out like your specious price comparsion that did not factor in any Ryzen 8 core SKU vs Intel 6 core SKU Price/Per-Core metrics or the Value of that Free cooling solution that came boxed in with the Ryzen 7 2700X SKU when compared to an Intel SKU where the user has to purchased at extra cost a cooling solution.
Buy Hey I’m also very quick to jump on any Threadripper for workstation usage folks also for not doing any proper cost/benefit and cost/feature metrics that claerly show that AMD’s Epyc True Server/Workstation CPUs/SP3 motherboards are the better solution for workstation/encoding/other non gaming tasks than any Threadripper/TR4 MB solution.
So I’m more about using a proper cost/benifit analysis for any product be it Intel or AMD. I’ll always be commenting on any Threadripper for workstation usage where the owner does not also want to game as Epyc is the real solution for server/workstation usage! And Epyc/SP3 is more affordable on a cost/feature basis than any Threadripper or Intel consumer CPU/MB HEDT options.
AMD’s Epyc/SP3 options are so damn affordable that no person wanting True server/workstation hardware need to ever be forced to substitute any consumer grade parts like Intel’s users are forced to do simpley because Intel’s real Server/Workstation options are too damn expensive!
If you already own a cooler then you should have mentioned that as part of your cost/benifit comparsion while also mentioning that Ryzen 7 2700X came with a boxed cooler included in that SKU’s MSRP. And even there if you already own a cooler then maybe that included Ryzen 7 2700X boxed cooler SKU can be readily sold on eBay or other sites and that figured into the overall price savings comparsion.
A wall of text for no reasons
A wall of text for no reasons at all, congrats!
P.S.
My what? You failed to note that I’m not the author of the original post you were replying at, yet I easly realized he probably already own a decent cooler and may not care of streaming… as for obtaing anything worth by reselling the stock cooler, the important thing is that… you believe that.
That Wall-O-Text was there
That Wall-O-Text was there just to piss you off and that other poster is just being disingenuous.
There is much more to a cost/benifit analysis than just price and an 8 core SKU at lower cost by $20, or even at some resonable higher cost, compared to a 6 core part is still going to have the 8 core CPU be a better feature(Per core)/cost. That included cooling solution has value so that’s has to be included for Ryzen’s value eguation for more than just one person that may or may not already have a cooling solution.
Intel 8700X at $349/6-cores = $58.16(rounded) per core.
Ryzen 7 2700x at $329/8-core = $41.23(rounded) per core, add the cost saved for the included cooling solution and the Cost/Core is even lower than that $41.23.
It’s just fine if that other poster is looking at what he wants but that does not erase the fact that over that GN, Steve’s review tested gaming while streaming as that’s what some folks do. There are other all around workloads that like the extra 2 cores/4 threads and the latency improvements for Ryzen 2 in addition to the lower voltage needed to achieve higher clock speeds on Ryzen 7 2700X.
Sure that 12nm process is somewhat improved for voltage inside the regular clock envelope(3.7 to 4.3) GHz up to aroud a little above 4.3 Ghz and that’s a good thing for Ryzen 2. And Yes the overclocks above that 4.3 GHz range come at an extra cost. Ryzen 7 2700X and Zen+ are still an improvment and maybe Ryzen 2 will also improve over time as much as Ryzen 1 if you look at the 1800X and lower Ryzen 1 SKUs and those SKUs improvments over time.
GF will be tweaking its 12nm over time just like GF tweaked that licensed from Samsung 14nm process over time to get the diffusion tweaked and the Ryzen 1/14nm chips performing better. The diffusion QA/QC will improve over time at 12nm and that will result in better transistor leakage metrics ans well as other improvments.
Zen+ also got those latency improvments IF tweaks and that had an even better affest on some gaming workloads than higher clocks, ditto for any non gaming workloads that are latency sensitive. Faster memory and better 400 series MBs are also an improvment over the 300 series MB/memory speeds.
Actually you are just adding to that other poster’s specious argument so much so that just: you believe that!
Edit: better affest
to:
Edit: better affest
to: better affect
one hell of a review.
one hell of a review. curious to know why you guys keep running old benchmarks written for intel chips?
i5-8400 has always higher FPS
i5-8400 has always higher FPS than i5-8600K, it’s hard to believe… looks like you have some mistake in your graphs…
Why would you make the Intel
Why would you make the Intel cpu orange/red and the AMD cpu blue?