Principled Technologies Says it Messed Up Intel 9th Gen Testing

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
If you want to argue that ryzen should be tested with the stock wraith cooler then the only acceptable solution is to test the intel chip with its stock cooler :)
 
Ok, so someone asked if AMD's cooler allows the Ryzen 2700x to hit its max boost, which shows and ignorance of XFR. The max boost of the chip is 4.2GHz, which it will hit with the AMD stock cooler. XFR is above and beyond the max boost, and it works on all cores where as the boost is a single core thing. If there is thermal or power overhead left on the table, XFR boosts past stock settings and past boost settings on as many cores as it has the resources to do. So, testing with the Noctua cooler would have allowed single and multi threaded performance on Ryzen to be even better than with the stock cooler. This would be why the crippled 4 core/8 thread version of the 2700X performed relatively well. There was a lot of power and thermal headroom for XFR to work with. It probably took all cores past 4.2 as often as it could (which would make for a really cool article actually!).

Would it have beaten the 9900K? No, most likely not, and certainly not with 4 cores disabled. However, testing it will all cores enabled, and with the Noctua cooler, it would have closed the performance gap considerably.

The Wraith cooler is a good cooler in the way that the Hyper 212 is a good cooler. They are far better than Intel's offerings and cool things pretty well. Both are still cheap coolers that offer ok-ish overclocking headroom on higher TDP chips, but are not as good as the Noctua, which a much better quality cooler that handles higher TDP chips, and therefore offers improved overclocking headroom.
 
This reporting is all well and good but which Tom's Hardware reporter interviewed them to get this admission. I noticed you didn't attribute this to anyone. That is bad Journalism, credit those with enough drive to go and talk to them. Steve Burke @ GanersNexus is the one who should be credited in BOLD TYPE. Don't forget to print
the follow up reply that Steve Received.
 

Can you stop spreading outright misinformation please. The "professional" reviewers downloaded overclocking software from AMD and used it to turn on a non-default feature which disabled half the cores. AMD did not do that!

I agree that for a 2700X, the name "game mode" is misleading. It makes sense for TR CPUs, but not Ryzen. I can understand how a casual hobbyist or observer might get caught out, but it's hugely embarrassing in a professional review. If you listen to the interview with GamersNexus, "Principled Technologies" claim to be trying to represent real-world, out of the box performance. If that's my methodology as a professional reviewer, then I should make absolutely sure that there is a good reason for any non-standard, "out of the box" settings being changed.

I go could into a BIOS and select one of the pre-set overclocking modes for "extreme performance" (or whatever they label it as). If that results in the CPU de-clocking for stability and ruins performance, can I blame MSI or ASUS based on what the setting was called? Of course not!

Maybe it was an honest mistake by Principled Technologies. They say they're retesting now. But it's a really embarrassing mistake that this got all the way to being published in a commissioned report with a mistake like that. To try and construe it as AMD's fault it pretty ridiculous IMHO.
 
And thank you for keeping up on this story. Between using a cooler the size of an 18 wheeler to cool their demo model at computex, their vapor releases and now this, why in the world would anyone trust Intel?
 
AMD made software that disables cores on AMD products for "better gaming performance". That's laughable, but true. I know they made it for Threadripper, but when you download Ryzen Master software it is not explicitly for Threadripper. It's not misinformation. It's complicated, just like AMD likes to be.
 
... and while I'm on a rant, the argument being maintained above about the Noctua U14S being justified because AMD market their bundled cooler as the "ultimate cooling solution"... seriously!?

With any marketing fluff, surely we can agree there are implied product categories. A thin and light gaming laptop that claims to offer the "ultimate gaming experience" is obviously going to be objectively worse than a $5K gaming desktop. Intel's 9980XE product fluff claim "extreme performance", yet in highly threaded tasks they'd get absolutely crushed by a modern quad socket workstation. The point is, when marketing departments makes claims like "ultimate" and "extreme", it's implied that they're making comparisons to products in the same class/category.

A bundled down-draft cooler that has to be fully compatible with any stock build is in a different product class/category to a high end tower cooler like the U14S. To suggest it's a fair comparison because AMD's marking materials said so is akin to me asking for a refund on my sports car marketed as the "ultimate performance vehicle", because my lap time around Monza isn't as fast as a Formula 1 car.

In any case, I actually don't think the cooler is going to make more than 1-3fps difference in most gaming workloads anyway... particularly on the quad-core processor as configured!

I expect HardwareUnboxed to release a video in the next 12 hours with benchmark comparisons between the "game mode" and non game mode 2700X. They were able to recreate the wonky PT results by turning game mode on. That seems to account for most of the issues: https://www.patreon.com/hardwareunboxed
 

But it's overclocking software! If you go into any BIOS' overclocking sections, or any software based overclocking products, there are a mountain of settings and tweaks you get access to. Many of those settings "can" increase performance on certain systems in certain situations, but will absolutely reduce performance (or stability) on other systems or in other situations. That's precisely why those settings are only accessible through specialised overclocking software or BIOS sections and aren't just enabled by default.

Anyone changing overclock settings based purely on what they sound like they might do is running around a maze with a blindfold on.

I can't understand anyone defending that approach to published benchmarking.
 
I haven't had an AMD system in many years since they've been losing to Intel, so no I'm not familiar with the ins and outs. I do think it's humorous the situation their software caused here. I also think they should definitely reconsider that option in their software.
 
I honestly didn't know the Wraith cooler included with the 2700X throttled it at stock, else I might not have started out with this opinion. Yes, they should have used the same cooler on both CPUs. But, perhaps like me they weren't aware the Wraith was inadequate for stock clocks.

Here's an update from PT. https://wccftech.com/principled-technologies-replies-controversy/
 
Honestly disabling half the core on the 2700x is only going to affect it in games like ashes, and putting the a better cooler on it might add about 0.3% gain in fps due to better thermals, the results honestly aren't going to change much other than ashes lmao...
 


Yeah, that's what I was thinking about the cooler. I wasn't aware thermal throttling was an issue with the Wraith Prism.
 

The cooler isn't going to make much difference, but your claim about the core count making no difference is objectively wrong. Steve from Hardware Unboxed addressed that issue at around the 12:20 of this video: https://www.youtube.com/watch?v=25x2xPh7Kvo
You can see Far Cry 5 and Assassin's Creed: Origins hugely affected as well at 2:15 of the same video.


As has been pointed out to you in direct replies numbers of times in this very comment thread (including by the author of the article!), it's not about "thermal throttling" on the stock cooler it's about XFR. Just scroll up and read what people have been trying to tell you. In the end of the day it's not going to make a big difference, it's just poor benchmarking practice that casts doubt on the results.

In any case, we can argue about methodology forever. The real cause for doubt here isn't the methodology, it's the results. They just don't align at all with results from independent reviewers. Remember there's an 8700K included in the testing, and the gap between the 2700X & 8700K is significantly larger than any other reputable source has measured.

I'm done now anyway, but before I go, let me just recap what we're talking about here:
Fact 1: Intel paid Principled Technologies to run a set of benchmarks and produce the report we're all discussing
Fact 2: Intel approved the release of the report while all independent review results remain under embargo. These are the only publicly available performance figures for Intel's new products
Fact 3: The report shows Intel's CPUs performing significantly better relative to AMD than what we have seen from many reputable, independent review sites. Particularly the 8700K to 2700X comparison, which is well established and widely understood by now.
Fact 4: Many independent reviews have serious questions about the methodology (including Hardware Unboxed (& Techspot), Gamers Nexus, Paul's Hardware and here on Tom's Hardware).

We can argue about the impact of #4 all we like, but it's #3 that made everyone start digging into #4 anyway. If the performance figures were in line with what was expected, we wouldn't be having this conversation.

You trust the Principled Technologies report if you want to. In the end, there's little doubt that the 9900K will be the best gaming CPU on the market. But if we trust the reputable reviewers who are speaking out on this, it's an equally safe bet that the margin between the 9900K and 2700X will be much smaller than this report claims.
 
"Excellent reporting Paul. Sounds like AMD's Game Mode is a bit misleading for users of Ryzen 7. Which brings the question, why do they even include it on Ryzen 7 systems?"

It isn't included on Ryzen 7 systems, or any system at all. Ryzen Master is an overclocking utility you specifically have to download and install. To use Game Mode you specifically have to disregard all instructions on how to use it and enable game mode, ignore the warning when you select it and apply it anyway, and then reboot to make it take effect.

It is there to help people with a Threadripper that have problems running a specific game due to it not being compatible with high core count processors.

The obvious choice for Principled Technologies would have been to skip installing Ryzen Master at all. In the games they tested it would have had marginal, if any, effect on the Threadripper performance. And it would have save them the public relations nightmare they're now in where it has also been uncovered that they are actually an affiliate of Intel.
 
"Well this is the real work where Intel sells CPUs not benchmarks. Knowing that, no one should listen to their results and they shouldn't even be sharing results because they are biased."

That is actually wrong. Intel controls BAPco that makes SysMark. They even share the same physical address. They also commissioned the XPRT test suite developed by... Principle Technologies... well, now that's a funny coincidence, isn't it?
 


You would have to use smoke and mirrors to say XFR doesn't have to do with thermals or throttling. If it's not boosting based on too high of temperature then that's thermal throttling.

No, nobody trusts Principled Technologies based on those results. As I've held in all my posts here, I would never trust a benchmark before a release. I wouldn't expect anyone else to do so either.
 


Yes, they made some very poor choices.
 




Thanks. I don't know everything.
 


I have a 2700X. While waiting for the SecuFirm2 mounting kit for my old Noctua NH-U12P, I used it - in the middle of summer with no AC in an open case vertical build.
It didn't throttle once, even during a Blender render, it kept to the published all-core frequency : 3700 MHz. It even on occasion went up to 3725 MHz, keeping right under 80°C.
Now, with a Noctua mounted on it, it does go up to 3900 MHz sometimes, meaning that XFR2 works pretty well at auto overclocking the CPU.

  • ■ Does it mean the Wreath cooler causes the CPU to throttle? No - the CPU stays at the frequency it's supposed to run at (no 1 GHz downclock like on Intel's CPUs using their stock cooler).
    ■ Does it mean mounting a better cooling solution makes Ryzen faster ? Yes.
    ■ Is it exactly like specified by AMD? Yes.
    ■ Is their Wraith cooler good? As good as any $30 compact, blow-down cooler could be.
    ■ Is the Noctua NH-U14S much better? Hell yes!
But Intel doesn't bundle their CPU with a Noctua cooler, so it has no excuse being used on Intel builds only.
 


So you buy a CPU with a max boost speed, that doesn't boost to that because temps don't allow it. And you want to tell me that's not thermal throttling?
 
Thermal throttling is the control of a CPU's speed based on core temperature. XFR is thermal throttling (& boosting) by design. You can't boost without a throttle unless you never throttled and ran at max boost the entire time. Prove me wrong.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom