AMD Ryzen 5 2600X Review: Spectre Patches Weigh In

Status
Not open for further replies.

bbertram99

Prominent
Dec 11, 2017
4
0
510
i was wondering if reviews will now be posting if they checked if it was on or not. I think they should since we don't always know when it being forced. Its not evident until you look. If they don't state its not forced on then we are left wondering.
 

toyo

Distinguished
Sep 24, 2011
36
0
18,530
What's the point of this? Where's the GTX 1080ti? The 1080 simply result in every CPU being able to feed it enough data so scores are almost similar. Hence anomalies like having the 8400 or 8600k often being better than the 8700K, which should be impossible considering the higher clocks. I mean, this CPU performed the best for 3-4 months, even after Meltdown/Spectre patches/BIOS, and now it suddenly has issues competing with its own family of CPUs that are half that price, really?
Then there's the gaming suite chosen. Old Far Cry? The 5th is out. Where's AC: Origins, notoriously CPU hungry? Overwatch? FFXV?
Hell, even the older Deus Ex or Kingdom Come: Deliverance would have made more sense to test CPUs.
But yes, this shows that for anything below 1080ti, you're good with pretty much all of these CPUs. Yet it doesn't tell the whole story, and soon a new GPU generation will be released, probably introducing many here to GTX 1080ti levels of performance, so testing with it does make sense.
 

PaulAlcorn

Managing Editor: News and Emerging Technology
Editor
Feb 24, 2015
857
313
19,360


HPET has been disabled by default in Windows for a decade or so now. The OS can call on HPET if it needs it. The performance overhead of HPET is a known quantity, which is why the OS doesn't use it if possible.

We test without HPET disabled, which is enforced by our test scripts to ensure it stays that way.
 

Agreed. I'm still using an i5 760 (@.3.4GHz) which was released in July 2010. I have had multiple GPU upgrades over the years (as of this moment I'm on an R9 390), so I also would very much like to know if a new CPU is as "future proof" as possible with regards to GPU upgrades.
 
Mar 30, 2018
2
0
10
Would the 8700k cost what it does today without Ryzen? They would still be feeding us quad cores on the same 14nm process.
 

bbertram99

Prominent
Dec 11, 2017
4
0
510
Don't see how to quote you PAULALCORN.

Considering Anandtech got caught by the HPET bug and you never see it mentioned in any reviews until now. So now i question each review I have seen and will see unless in mentioned. The credibility of all benchmarks are in question unless it clear HPET is disabled. Good thing you script handles that, thank you for let me know.

Keep on providing great content, I've loved Tom's reviews for a LONG time.
 

PaulAlcorn

Managing Editor: News and Emerging Technology
Editor
Feb 24, 2015
857
313
19,360


No one mentions HPET because it is disabled by default in the OS. If we listed every single feature that we leave alone and do not alter...that would be a long list :)
 
Except you're wrong Paulalcorn.

HPET is ON be default in the Bios, and it's treated on a case by case basis by the OS. Generally it's off, but if it's available and software asks for it, it's allowed. What bit anandtech was they KNEW of the OS's unpredictable treatment of HPET, and in order to "standardize" their results they were forcing HPET ON for everything at the OS level. This is good scientific process, you have a variable (a feature like HPET) which is used by the OS in a non-controlled way and has been shown in the past to negatively (or positively) affect performance, you simply turn it into a constant.

The problem is HEPT, since the Specter2 patch is now HAMMERING intel performance in a wide range of benches. I don't agree with ANANDTECH's decision to no longer "force" HEPT. Not because it helps INTEL, but because it was good science to FORCE it on, because of how the OS and software treat it and apply it being sort of case by case. anandtech and all other reviewers should either test with it forced off for everything or force it on; and the reviews should make it clear which way their system was tested. Leaving it on but not forced only makes their results less consistent, and is probably the worse way to address the problem.
 
Apr 28, 2018
1
0
10
Why aren't you checking the Ryzen 7 2700. I'm very interested in the gap between the 2700 and 2700X. If you plan a quieter setup how much do you loose on this change? I'm missing the 2700 on those charts.
 

toyo

Distinguished
Sep 24, 2011
36
0
18,530

What "case by case"? Give me ONE example where HPET is enabled by the OS automatically depending on an application. Just one. You know it takes a reboot for that to happen? How is that normal OS behavior when you require reboots to enable/disable features depending on user activity? It's not and it never happens.

I've personally had the HPET On from BIOS, but it was never active in Windows for 6 months now. I had to force it to see its effects, and I'll be damned, it's a disaster for my 8700K. And NO, Spectre/meltdown patches do not matter, this is true even before them being applied.

It's incredibly dumb to force a Windows flag that is by default Always Off, especially when it has a dramatic performance impact. How many Intel users in this whole world force HPET on when it makes games unplayable? Not even an exaggeration.

So why in the world a reviewer would force it, when it has VISIBLE framerate drops and hiking in games? Why would you tank performance intentionally for half your CPUs?

Yeah, I thought so,
 

msroadkill612

Distinguished
Jan 31, 2009
202
29
18,710
I love scatter plots, and wow, 2600x sure dominates the excellent ones on the last page.

It sits down there in the corner in ~every category, with little much to the right of it, at any price above.

What a bargain?

It has the makings of a popular classic.

Its no longer version 1.0 of the radical zen ryzen architecture and platform. 12nm, meh, its a great chance for a good tidy up and a harvest of low hanging fruit.

AMD will always be a bit hungry for those extra 2 cores, so 6 cores will remain attractively, relatively cheaper.

Big picture/strategically, they show a seriously troubled intel.

For me it's moot. AM4 is a better platform imo. Vitally, most intels with a proper dgpu, give you exactly zero spare io bandwidth beyond the 4GB/s shared by the chipset, with it's overheads also.

A single nvme (960 Pro e.g.) would be too fast for it / could saturate it. Other chipset connected resources like sata/network/usb ... would be choked.

AMD is niggardly also, but adds another 4GB/s, or as much as the entire chipset again.

A strategically utilised nvme on those extra amd lanes, relieves the chipset of its major drain on it's limited bandwidth - the main system drive.

An am4 w/ 16GB ddr4 = ~40GB/s (40,000MB/s) bandwidth - thats the max the system can offer.

16 lane pcie3 gpu link = 16GB/s in theory

Then we have, still mainstream, storage like:

sata ssd = ~450GB/s (550MB/s in theory)

sata hdd - ~50-150MB/s

Quite a gap, no? It is the slowest important resource by a huge margin, and inevitably hindering their performance in many subtle ways.

Yet when affordable mainstream nvme offers a chance to strenghthen this grave weakness (like 3000MB/s sequential reads), folks buy intel & effectively turn their backs on nvme's real power.

Its surreal. Folks are offered an affordable option for their slowest component, thats ~6x faster, & they not only pass, they close future options for want of bandwidth and lanes on their intel platform.
 

msroadkill612

Distinguished
Jan 31, 2009
202
29
18,710
Superficially it seems OT, but comparison would reveal a lot, if it included AMD's zen vega apu product, which deviates from the zen norm of pairs of CCX, and the associated inter ccx latency.

Its amdS only true zen equivalent to intel's 4 core cpuS afaik.
 

randomizer

Champion
Moderator


Forcing the HPET on will give you results that are consistent and accurate but completely meaningless. Forcing it off will in all likelihood do nothing at all, except perhaps cause an application to behave incorrectly in the rare instances where it needs the HPET. It's the better of the two options, but I still don't think it's a good option. Leaving it available on demand is the only way to have results that represent reality.
 
First gen Ryzens had acceptable performance with an awesome upgrade path, so people could justify it to help make the industry competitive. This doesn't need any argument to support a purchase other than benchmark performance, which is pretty cool, considering I've been waiting a decade.

Is it as cool as getting HL3 would be? Note quite...but it measures on that scale.
 


Yep. there's no reason under pure gaming scenario I need to upgrade from my nearly four year old i5 4690K/Z97 Haswell build pushing a 1440p monitor since it's mostly on the GPU. However as I get more involved with video rendering (Handbrake, Vegas Studio), I need more CPU power and threading. That's where Zen wins over Intel.

AMD also wins in supporting their chipsets for longer. Since my Haswell/Z97, we've seen three new generations of chipsets from Intel, and only two of those three have been compatible. And most early 7th-gen Kaby Lake motherboards required a BIOS flash with a 6th-gen Skylake chip.

Regarding Half Life 3, that's the Area 51 for the gaming community. UFOs may or may not be real. But we can always dream. IMO that's the biggest devoid/lost opportunity in the gaming industry to this day. Such a shame.
 

alextheblue

Distinguished

Actually I've seen HPET fix stuttering in games for a number of users. Whether it helps or hurts, and how severely, depends on the hardware and software involved. Regardless there definitely ARE reasons to force it on. Also HPET does have a small impact on AMD's CPUs... it's just far more severe on Intel's recent platforms. Make of that what you will.

A lot of overclocking and monitoring software require HPET to be forced actually. I guess if you had done a little research before railing against HPET you'd have known this. It requires a reboot... what's your point? I installed a new driver the other day... I had to reboot. I installed an OS patch, I had to reboot. Perfectly normal OS behavior.

With that being said I absolutely don't care that Tom's and others test with it at "default". That's the majority experience and it's fine. But putting an "HPET Not Forced" line in their software setup page wouldn't kill anyone.


That's not entirely accurate. Read the AnandTech analysis of HPET's effect. Intel-specific Meltyghost microcode patches exacerbate HPET's impact. They've been using it for a while, and their tests and Intel's testing guidance indicated no issues... prior to the recent fixes which DO impact HPET usage.

https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results/4
 
I'm still running an i7-2600K on my main desktop. It doesn't noticeably bottleneck anything I do frequently.

 

PaulAlcorn

Managing Editor: News and Emerging Technology
Editor
Feb 24, 2015
857
313
19,360


Clarification here. HPET has long produced sometimes-severe performance problems, even before the patches. This is not new information.

The patches likely exacerbate the issue, but until we see HPET patch- versus non-patched performance, we can't be sure of the true impact.
 


Lol can you just imagine intel bringing todays lineup out back in the days as haswell CPUs?
Oh the carnage, just imagine a 6/12 haswell at 4.7Ghz for $370.00 against the FX-9590, AMD would just close up shop right then and there...good thing intel didn't do something like that.
 
Status
Not open for further replies.