AMD Ryzen 7 2700X 3.7 GHz

[Shatun_Bear]

Now you have me really interested. The low end never gets love, and it's what sells the most.
AMD has those horrible 9000 APUs as the cheap options, the Celeron should destroy them.

They also have the impending Ryzen 2400GE and 2200GE, low-power APUs. Or you probably mean even cheaper than these? Let's not forget I can buy a 2200G for £75 here in the UK. 2200GE should be cheaper.

 

[GoldenX]

No, the low power options are always more expensive, they are extra binned.
AMD can release a dual core + SMT Ryzen APU, that would be a lot better than the FX era 9500-9800 APUs.

 

[HTC]

@W1zzard : have you seen this?

Anandtech discovered why their results were so different then other reviews and the implications of this can affect the review's results.

 

[W1zzard]

@W1zzard : have you seen this?

Anandtech discovered why their results were so different then other reviews and the implications of this can affect the review's results.

Not sure why anyone would ever want to force HPET in Windows. Apparently no other reviewer did that, so just an isolated failure.

 

[HTC]

Not sure why anyone would ever want to force HPET in Windows. Apparently no other reviewer did that, so just an isolated failure.

You can "thank" Microsoft for that. If they kept the "up to Windows 7" way of measuring time since then, this would not be needed.

As you can see from Anandtech's benches, the difference can be significant which IMO warrants a check to see if your results were in anyway affected. Just pick one of the "worst offenders" from their review and do what they did with the timer (assuming you can): if it shows no difference, it's safe to say your review is "timer error" free. If it does show a big difference, then ...

 

[DeathtoGnomes]

Not sure why anyone would ever want to force HPET in Windows. Apparently no other reviewer did that, so just an isolated failure.

it has to be enabled in the BIOS ( assuming there is an option in your BIOS) as well as Windows.

 

[W1zzard]

it has to be enabled in the BIOS ( assuming there is an option in your BIOS) as well as Windows.

You're not supposed to "enable" it in Windows. Doing so (bcdedit useplatformclock true) forces Windows to use HPET as ONLY time source on the system which causes the issues Anandtech experienced. Windows will automatically do the right thing if things are left at default.

 

[HTC]

You're not supposed to "enable" it in Windows. Doing so (bcdedit useplatformclock true) forces Windows to use HPET as ONLY time source on the system which causes the issues Anandtech experienced. Windows will automatically do the right thing if things are left at default.

It's the other way around, dude: they found out their issues were because the timer was having inconsistencies, which is why they forced the HPET timer to be the ONLY timer, since it's supposedly the most accurate one.

Once they enabled HPET only timer, their results were very similar in some cases but very different in others, and this is for both Intel and AMD's systems, though in different ways.

EDIT (damn auto merge)

It's the other way around, dude: they found out their issues were because the timer was having inconsistencies, which is why they forced the HPET timer to be the ONLY timer, since it's supposedly the most accurate one.

Once they enabled HPET only timer, their results were very similar in some cases but very different in others, and this is for both Intel and AMD's systems, though in different ways.

And i had a brain fart ... after reading the new Anandtech review i understood it backwards ... it's once they disabled HPET that their results became more inline with other reviews.

Still, this opens up another can of worms: whose to say that the HPET only timer results are "the wrong results"? I mean, if benches rely on timer to accurately give results and the timer is incorrect, the results should also be incorrect, no?

 

[JRMBelgium]

The performance per dollar, does that take in to account that you need to buy a cooler for the 8700K? Because I don't think it's fair to make a "performance per dollar" chart if you're not comparing appels to appels.

 

[oli_ramsay]

I still think my 3770K is OK by today's standards. I don't see the point in upgrading until next year when Ryzen 2 is out. I mainly game anyway.

Still good to see competition though!
Now if only the GPU side of things were the same! I need something better than a 480 for my freesync monitor and I'#m not paying >£500 for a vega 56 lol

 

[Zatanus]

Can anyone please tell how Virtualization with Vmware has been tested?

 

[W1zzard]

Can anyone please tell how Virtualization with Vmware has been tested?

1.. 10 VMware instances are started in parallel and the time measured up to a certain point

 

[Zatanus]

1.. 10 VMware instances are started in parallel and the time measured up to a certain point

Thanx for the reply.

Can I replicate that or is a TechPowerUp custom process?

 

[W1zzard]

Thanx for the reply.

Can I replicate that or is a TechPowerUp custom process?

it's a custom vm with custom test driver, nothing comes from anything that's available commercially or in public. so no

 

[Charcharo]

You're absolutely right I think about the onus being on the devs, but there are limitations to that as well. In budget, the engine in use and the limitations of that engine, and coding talent as well. And wherever the devs fall short, we use hardware to compensate for it. Its always been that way and always will be that way. Software is almost never 'perfect'.

The bottom line for a vast majority of games in my view, doesn't really change: single thread is and will for the foreseeable future be the limiting factor when it comes to CPU performance. You're right about Ashes. And there are other examples of new games as well that use cores extremely well, I see them too. And at the same time, even in those cases the performance does not scale linearly with core counts. There are many more things in the entire pipeline that can limit FPS some of which even faster hardware cannot overcome. But it does still help in lots of scenarios.

Physics simulations are easy to divide across cores and threads, and any non-sequential workload in games as well. New APIs provide easier access to actually implementing code in that way, and this will in the longer term (we're not really seeing it yet, today, let's be honest, almost every bench linked in this topic is proof of that, almost none of the results actually show a tangible FPS increase from core counts/thread counts) provide the headroom to make more complex games.



You keep harping on about a dead res and if you are not running into a performance bottleneck on your rig, more power to you. I'm not 'skipping' facts by the way, quite the opposite, I am supplying evidence that supports them and I will always keep asking questions when it comes to performance - all you've done is repeat yourself a couple times and link Google.com. And yes, there are a lot more multi-threaded programs but if you had taken the time to actually get into the supported evidence of single thread limited gaming scenario's, you would see most cores are loaded with nothing substantial, at least nothing that improves FPS. We're at 15 pages now... do scroll through them and take a look. The comment about FX and 2012... you do realize I just upgraded from Ivy Bridge, yes? I was around back then, just not here... I suppose I shouldn't tell you how old I am but you're quite far off the mark with 'kid'

Beyond that, lets agree to disagree, because this is going nowhere substantial, except for getting personal.

Problem is, it is not about average IPC. Ryzen 2000 is close to Intel there, around 5% difference so one would expect that it would be the same in games under equal clocks, yet Ryzen 2700X (which is clocked higher) sometimes can not win even vs the i5 8400. I honestly think it will take time till the architecture in general gets used better by the, let's be honest, mostly bad programmers that work on video games. And the Nvidia driver issues with Ryzen do not help in some older game benches.

Ashes has split AI across its threads too. Not just physics. Men of War has done neither, hell it is 32 bit still. Another low budget game beating Starcraft 2 on technological fronts though, makes one wish Blizzard would actually get good at their job.

*I say this as a person that plays Starcraft 2 and likes it a lot, but seriously its getting silly.


ANyways, game logic at 720p and at 1080p may have different bottlenecks. CPUs and games are not so simple, that is my problem with the tests. Still, for me personally, as a high refresh rate gamer, going from Haswell to Ryzen 5 1500X was a gigantic jump in all scenarios (though I am using a Fury on top so no Nvidia/Ryzen driver issues in older games). Now if only I had more cash I would plan an upgrade for 2600X.

 

[Shatun_Bear]

Seeing as the 9900K releases for £600 ( ), in hindsight buying this 8-core Ryzen was one of the best purchases you could of made earlier this year for £280.

 

[phill]

Seeing as the 9900K releases for £600 ( ), in hindsight buying this 8-core Ryzen was one of the best purchases you could of made earlier this year for £280.

I especially like a 1700X for £150... Bargain deal if ever there was one

 

[JRMBelgium]

I especially like a 1700X for £150... Bargain deal if ever there was one

The current Price/Performance ratio in Belgium/The Netherlands:



So you are correct. Also in my country the 1700x is a better deal then the 2700x when it comes to performance per euro.