The 2700X makes me want to build a DESKTOP!

Guys, the 2700X is BEASTLY! Whos with me? This 2700X is truly impressive!

Passing the 8700K in R15 by 30%.

I mean sure, I'm a intel fan and all. But, sometimes you just got to get off the pot?

 

Nah, an 8700K at 5GHz is still a more well-rounded performer. Sure the 2700X beats it in productivity apps, but we all knew that and it was the case with Zen 1 as well.

 

For now at least. Intel are starting to push game developers to optimise their engines for more threads rather than sheer clock speed and IPC. In the next few years the gap between Intel and AMD will start closing up from games being optimised better. I'm certain Intel are planning 8 core CPUs as well within the next year or two.

 

Considering that CPUs matter less and less for gaming, The ryzen smacks the intel CPU.

Intel is inferior to ryzen in almost anything. Ryzen CPU's are better for producitivty, are cheaper, more future proof and run much cooler. Once DX 12 and Vulkan are mainstream, in terms of gaming there will be no difference betweem a low I5 and a high end I7.let alone 8700K and ryzen.

(even tho currently the 2nd zen smacks the intel in terms of gaming lol)

EDIT:
Well, apparently ryzen does now even smack intel in terms of gaming and IPC. Guess there is no reason to go for intel anymore.

https://www.anandtech.com/show/12625/amd-second-generation-ryzen-7-2700x-2700-ryzen-5-2600x-2600/15

 

It’s incredible! Whenever threadripper 2950X comes out, it should be fun to watch.

The 1700X could never fully grasp my interest for some reason. The 2700X has changed that.

I don’t think there will be a 2800X.

 

Sure... And i7-8700K wasn’t even oc’d... https://www.techpowerup.com/reviews/AMD/Ryzen_7_2700X/20.html
Compared with a stock clocked 6 core i7-8700K (33% less cores).

 

Zen+ certainly does not smack Intel in terms of gaming. Intel is still king for high refresh date gaming.

The AnandTech review is utter BS (2700X performs 200% faster than 1800X in Rocket League, really?). They're clearly shilling Zen+ seeing as how it destroys Zen and CFL in that review, and their gaming numbers especially are different than every other reviewer out there. CFL power consumption magically increased by 40% compared to their initial CFL review as well.

I've blacklisted AnandTech because of this article as they're no longer a trustworthy news source. How the mighty have fallen. Anand be rolling over in his grave.

 

Anandtech's review isn't the only one showing the 2700x vs 8700K. Here's an updated selection of benchmark results with the new AMD Ryzen 2700x vs. the 8700K:

AMD Ryzen 7 2700X outpaces the Intel Core i7-8700K in more ways than one in first benchmarks

"The AMD Ryzen 2700X was given a test drive ahead of its official announcement and the CPU has managed to put up an impressive show. The Ryzen 2700X has managed to beat or come in close proximity to the Intel Core i7-8700K in gaming benchmarks and was absolutely stellar in multi-threaded performance.

https://www.notebookcheck.net/AMD-R...ys-than-one-in-first-benchmarks.296674.0.html

There are a many pages of print reviews and video reviews being posted here comparing Ryzen 2 vs Ryzen 1 vs Intel CPU's under many configurations, applications, and gaming, it's just getting started

AMD's Ryzen CPU's (Ryzen/TR/Epyc) & Vega/Polaris GPU's
http://forum.notebookreview.com/thr...zen-tr-epyc-vega-polaris-gpus.799348/page-446

 

Intel pushing game developers to improve gaming multi-core support will also benefit and uplift AMD 8 core, 16 core, and 32 core CPU's - shipping now and shipping in the future.

 

In AT's review, Zen+ beats CFL in every single gaming test including GPU-bound and single-thread reliant titles and including such anomalies as tripling Zen 1 performance in Rocket League. The Zen 1 and CFL numbers are fine, but anyone with a rational mind can see that the Zen+ numbers are an anomaly.

 

Yet, Anandtech ran the tests, and got the numbers, so there is a rational explanation.

Also, there are lots of other reviews that replicate the Anandtech results, if not in whole then in part, which is why I replied with links to other results.

I'm sure Anandtech will come back with a rational discussion and if necessary amended results.

I hope Anandtech improve's the cooling and tuning, OC's to a stable 4.3ghz+, and re-run's the tests for even higher results.

 

There is nothing rational about a benchmarking script gone haywire if that is indeed the culprit because the Meltdown and Spectre mitigations do not have a profound impact on gaming performance. They should've pulled the article long ago instead of leaving it up with that "we're reviewing the data" disclaimer and continuing to mislead people. I guess nothing drives traffic like shock value, but it's scummy.

 

You are assuming it is wrong, but it is correct according to their test bed and test plan, you may not like it, but it's what Anandtech came up with and published.

And, like I said, others are getting similar results in whole and in part, so there may be something to be learned from Anandtech's AMD setup.

I've been avoiding quoting Anandtech's results due to the controversy, that's why I linked all the others in the discussion thread on Ryzen / Vega / etc, if you spend some time with them you will learn much more than you know now, unless you've read or watched them all already.

 

A colleague of mine is building a new rig, but with the 2700 (not X). Reason being, is can be overclocked to nearly the same specs as the 2700X, uses less power, runs cooler, and is cheaper.

 

It can't run cooler, but it's specification rates the TDP dissipation lower, that maybe a problem with OC'ing as well, affecting it's reach into the top end.

In this Ryzen release, the X CPU's might be a better purchase than the unlocked non-X CPU's, time will tell. You can still OC the X CPU's, and they come with a nicer cooler, but for serious OC'ing you'll want an AIO or closed loop, or maybe try a Giant Noctua DS15

The PB2/XFR2 seems to eek out almost the maximum amount of performance when the BIOS is set for Enhanced Level 3-5 and the cooling is excellent. Then the limit is 4.35ghz - with better voltage setting - it can reduce voltage quickly and increase as necessary, on a core by core basis - perhaps better than a manual OC with a fixed voltage.

IDK, it's all too new - no fixed or certain recommendations can be made based on PB1/XFR1 generation Ryzen CPU's.

Please post your friends OC results, or ask him to come and visit us here and post the results:

AMD's Ryzen CPU's (Ryzen/TR/Epyc) & Vega/Polaris GPU's
http://forum.notebookreview.com/threads/amds-ryzen-cpus-ryzen-tr-epyc-vega-polaris-gpus.799348

 

They published invalid results and are leaving them up while they investigate, which is damaging their credibility.

 

Actually it stands by their credibility. They believe they have done good testing, and don't doubt their results.

You can't have it both ways. Either the many years of reliable reporting means something to you, or it doesn't.

If a little matter of questionable results from a known reliable site gets you to denounce and abandon them so quickly, I'd not think you very reliable either.

You could show a little patience and decorum, and wait for their update.

 

They do doubt their results, that's why they've been retesting everything since last week and have that disclaimer on every page. But not pulling the article in the meantime is bad form. AnandTech used to hold itself to the highest professional standards. If there is even a shred of doubt, then don't publish it. A little something called journalistic integrity.

I couldn't care less what you think of me.

 

You don't erase the results - you cross them out and include an explanation as to why.

Science man, lab bench results require disciplined recorded keeping.

Keeping the article online is how a professional does it. Taking it down and hiding from it wouldn't be the professional way to handle it.

Besides, this is the only disclaimer I found on the article, it doesn't sound like they are doubting their results to me:

"Please note that this review is still a work in process. There might be placeholders for tables, graphs, and text still to be uploaded. We're working right to the last minute (and then some more!). Please stay seated as we input your usual programming.

Update: A number of comments have noted that some of our gaming numbers are different to other publications. To clarify, we used the latest ASUS 0508 BIOS (on X470), full Windows RS3 + updates, Spectre/Meltdown patches, and updated gaming titles. We are reviewing the data."

Mellow out, you're getting upset over nothing, really.

 

So, quick history lesson on AnandTech. When Anand was running the show, instead of trying to rush their reviews out the second launch embargoes lifted in an unfinished and unpolished state like every other site, they took their time and released days or weeks after. But these AnandTech deep dives were worth waiting for. Now they're like everyone else, if not worse, since they can't be bothered to fact check obvious anomalies like the Rocket League results before publishing.

 

Yes, knew that as well, been around a very long time myself.

For now I am happy to let them work on it, and I will wait to hear their update, it should be interesting.

 

I’ve seen a few guys running the Threadripper 1900X at 4.3Ghz and some at 4.4Ghz with custom loop, and the good TR blocks that offer full ihs coverage. Not that EK block of junk..

Micro center was selling the 1900X TR for $349 so, I always thought it was the best (8) core ever released, next to when the 5960X came out.

But 1900X was pulling about 1910-1949 in R15 at 4.4Ghz!

So I’m hoping the 2700X could do the same with better cooling, firmware updates, and a little time.

 

Here you go, high OC 2700x scores including CB's Single core 210, multicore in the 2200-2300 range, sorry no index, have fun

AMD's Ryzen CPU's (Ryzen/TR/Epyc) & Vega/Polaris GPU's
http://forum.notebookreview.com/thr...ga-polaris-gpus.799348/page-445#post-10716704

 

That’s a very fast cpu! Wish more people could see that..

So many people are just in denial about the 2700X.

Sometimes it feels like the 8th gen intel owners just turn a blind eye.. The 2700X could be free for goodness sake ! And everyone’s like; meh! My cpu is a better daily option lol.

It must be the 8(7)00K contender, as AMD even named it 2(7)00X

 

@Danishblunt





https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results/4

https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results/5

So the moral of the story is, AnandTech effed up:

1. Their benchmarking scripts forced HPET on in Windows, which is not the default configuration (called it @hmscott)
2. HPET increases system calls, plus Intel's HPET implementation runs at a higher frequency than the specification requires
3. SysCalls are the operation most affected by the Meltdown+Spectre fixes
4. As the result of points 2&3, HPET in combination with SMeltdown fixes degrades Intel performance massively
5. AnandTech's Ryzen+ results are abnormally better than their Ryzen 1 results because they re-used their launch Ryzen 1 numbers instead of re-testing them for this review (Ryzen 1 has improved significantly since launch)

 

This is a much wider reaching issue than a huge drop in gaming performance popping up out of nowhere.

It's all due to the new Vulnerability Mitigations that cause system calls to be much more expensive and cause Intel's CPU's to have this huge performance hit, as seen through the 1080p (Higher CPU requirements) gaming benchmarks.

Considering Anandtech has been setting HPET on for a while now - no doubt to get better and more accurate timing for more accurate benchmark results, and it is just now getting noticed, it's a sign that Intel now has a serious problem when high precision timing is enabled (not just HPET) that require additional system calls that will incur a performance hit due to the new mitigations.

I need to consider this some more, but my initial reaction is that Intel CPU's are in serious trouble in the professional datacenter market, and any installation that requires HPT for accurate timers.

Note that AMD CPU's didn't incur a disastrous performance hit by enabling High Performance timers, that's key to seeing how little AMD has been affected by the mitigations, as they have stated time after time.

More studying is going to be needed in Datacenter environments where high precision timers are important. Different HPT implementations use a different mix of system calls for their implementation, with HPET using 2x as many as TSC, this may be why there was a much larger hit for datacenter services, VMware in particular - even though HPET is no longer on by default in VM's, the servers would have it enabled - and might indeed need it enabled in the VM's, for a multilayer hit.

In the past I've had high speed messaging systems not run well in VMware, timing isn't accurate enough - the resolution was below the threshold of the message timing, so accurate timestamps weren't possible, that changed over time, but is still a key consideration in new service implementation.

Basically the mitigations are the problem, not Anandtech's standard for testing. They have been using the same HPET enabled the whole time then their results are consistent from their testing perspective.

Which is also as I thought it was going to be found once Anandtech researched the source of the variance:

I think Anandtech has done an excellent job, and they have done a service to everyone in the community by discovering that Intel is in trouble performance wise - far larger than normally expected - when using HPT's.

It looks like Intel needs a new architecture much sooner than I thought, once this discovery hits the datacenter people they are going to be all over this - and quantifying the problem in terms that won't be flattering to Intel.

For the Datacenter AMD Epyc has another feather in it's cap, and Intel has another "aww ****" to deal with.

 

You don't need HPET unless you're BCLK overclocking on W8/10, and who does that anymore.

 

There are certainly other requirements for HPT, and in this case Anandtech's benchmark suite / tools require HPT enabled to remain consistent with historical results for accurate comparison.

This is a very public embarrassment for Intel CPU's, and a validation that AMD CPU performance is basically unaffected by the vulnerability mitigations.

And, I forgot to say it in my last post because I was focusing on the wider issues, but I think Anandtech's test results have been proven to be real and factual, and should stand published as tested.

Intel's vulnerabilities introduce a situation where benchmarks need to be run with the mitigations installed and enabled - or set the wrong example to the public by disabling them to boost Intel CPU performance results.

The Intel CPU system call performance hit is huge when requirements for the testing include high system call operations for the application being tested, monitoring tools, and / or test suite.

I don't think it would be appropriate to have multiple test runs to improve Intel CPU's test results with HPT's if required disabled, mitigations uninstalled / disabled, as with other settings tweaked to favor Intel.

Increasing the test load on the reviewers seems unfair, requiring them to make 3x-4x as many test runs just so Intel can post more favorable results.

What a mess...

 

Benchmark results should reflect real-world usage. Average joe systems do not have HPET forced in the OS.

 

That's actually not the point, the point is Anandtech has had it enabled all along, so why should it be changed now?

I do agree that the HPET shouldn't be enabled unless it's needed, tuning for deployment it's a consideration as to whether the service needs it enabled or not, if not then leave it off - but note that in the configuration notes so in the future if someone installs something that requires HPT features they know the state of the configuation.

If Intel didn't have this new lodestone around it's neck, this wouldn't be an issue, Anandtech test results would have continued to be fine, AMD too. It wouldn't matter.

Having HPET enabled turns out to be the canary in the system, and that canary keeled over nicely.

Which game's performance hit accurately reflects the performance hit your application will experience?, should you choose an Intel CPU for your next project?

 

The point is you have an axe to grind with Intel.

 

To be fair, we all do, you just don't realize it

Intel has a problem, it's affecting it's choice in new systems, a serious performance lodestone, and it is an unpredictable performance hit that shows up in use, varying in it's effect depending on your software and hardware configuration specifics.

That's a problem for everyone.

 

There's a difference between knowing and caring. These scare tactics can lick my rear, as the French would say.

 

You have me all wrong, and I know you don't care about anything, because you keep telling us that. I do care, that's why I am aware of what is going on, while you seem to want to keep everything in a particular focus - one you are comfortable with - without requiring any new awareness or understanding.

I am not trying to scare anyone, I am pointing out reality, if that is scary to you, if that is your reaction - then you need to react instead of sticking your head in the sand and muttering to yourself "everything is alright, Intel is still fast, Intel is still fast".

Everything is not alright, Intel has broken CPU's, and until Intel comes out with fixed CPU's, every CPU they sold, sell now, and come out with with the same broken architecture are all going to be worthless when the fixed CPU's finally arrive.

You do realize that don't you? The fear is supposed to grab your attention so you that you can evaluate reality and change your actions to reduce the danger to you? That is how fear is supposed to work. Let it happen.

Don't argue with the fire, run away from it or burn. It's that simple.

Learn not to burn.

 

Take it easy on the religious proselytizing, man. I prefer to believe in the things that are real. Not the big bad wolf that's out to get you.

 

All you have to do is ask me about the 2700X. Ive upgraded to it from the 1700X. I have not done a ton of benching with it yet, as there is no headroom above stock basically. It does perform better than my 1700X, but I dont know if it was worth the upgrade or not. I like to tinker with clock speeds, and this thing is extremely picky about any settings other than stock. At least as far as multiplier clocking is concerned. My board does not have the option to play with the FSB, so I cannot comment on that. At stock speeds, it performs better in 3dMark Fire strike, than my OCed 1700X @ 4.1ghz.

 

Now days you should believe in both at the same time, if you have any modicum of sense. The big bad wolf that is out to get you, is slowly becoming reality.
Not saying you dont have any sense, but if you believe major corporations, political bodies, or anyone for that matter, has your best interests in heart, then you should be scared.

 

You've just gotten it running so you may not have had the time yet, but when you do it would be fun to monitor the vitals with hwinfo64 logging on a 1 or 2 second interval to see how high the PB2 / XFR2 take the core clock speed, and how that relates to temperature vs tuning and cooling improvements.

I look forward to your feedback on performance results, on tuning and cooling, and other items of interest.

It looks like a lot of people are starting to push the limits of the 2700x / 2700 / 2600x / 2600 and are getting faster and faster results from manual tuning the CPU and memory speeds, timings, and voltages.

Congratulations on the new 2700x

 

I know they don't, but why should I be scared. Fear is a strong word and those things are trivial in the grand scheme of things.

 

XFR2 is only available on the X470 chipset, but I am still on the X370. Maybe the board upgrade is a little more crucial than previously thought?

 

Yeah, that's interesting, since stock comparisons on x370 and x470 show little performance difference, but XFR2/PB2 do show a big improvement in performance - almost as high as OC'ing manually if you are under 4.35ghz on the OC.

Early days yet, so many things to test for, and I think a lot of the reviewers aren't experienced enough to go into the depths needed know how to test for the differences, or even realize they are possible.

There are a number of posts of video's from real OC'ers starting to play with x470 / 2700x seriously, going into deeper BIOS settings than just voltage and multiplier, it'll take some time for the details to come out on what matters and doesn't matter most for tuning:

http://forum.notebookreview.com/thr...ga-polaris-gpus.799348/page-445#post-10716704

Go backward and forward from there for fun tuning sessions

 

This happens quite a lot at "auto" setting in the bios. The only thing I changed is memory clocking, and disable AMD's cool and quiet, so the cpu runs at turbo frequency. Note, that it rises above rated max frequency...

 

Bingo! You nabbed a shot at top PB/XFR performance, it was said to be 4.35ghz, but I don't think I've seen anyone else post an image showing that. Cool.

 

Awesome! The 2700X is the future of processing!

And if people are still “Ive gotta have an intel”

The architecture issues with HEPT should be resolved with the next gen $349 dollar (8) cores that are forced to be released soon enough.

Intel was playing dodge ball with a team that was standing still! It seems AMD has done more than starting to move

 

If you haven't already, these HPET details on the Anandtech article are good to read:

https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results/3

https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results/4

Both recommend leaving HPET on and not disable it. And to make sure to have that setting consistent across systems being tested / compared.

I doubt Intel has had time to implement changes to the CFL 8 core CPU architecture, at best Intel could burn the new microcode into the CPU, but that might not be a good thing - if you couldn't disable it.

That's not an architecture change, that's codifying the patches in hardware.

Here is the last from Intel on the issue of new architecture that fixes the vulnerabilities. Intel says this year, but for a real architecture change my estimation is it's going to be next year before anything comes out in production for consumers and professionals:

Hardware-based Protection Coming to Data Center and PC Products Later this Year
https://newsroom.intel.com/editorials/advancing-security-silicon-level/

 

AMD has taken both the x470 and z490 names now, placing them right in Intel's naming path.

AMD Readies Z490 Chipset with More (and Faster) PCIe Lanes
https://www.techpowerup.com/243419/amd-readies-z490-chipset-with-more-and-faster-pcie-lanes

 

That got derailed rather quickly, please keep on topic!