CPU Intel and their outrageously funny claims...

Some good PR for Intel:



Just kidding of course, pretty much nothing survives an analysis of the testing methodology.

 

Steve followed up on what he said about paying Principled Technologies a visit.

 

This ^. I felt super awkward watching it. But yeah they've messed up, Steve makes some very valid points, I just hope PT take the feedback and learn from it, no matter how much we think we know, sometimes there are better ways to do things.

 

Would you class a co-founder as 'management-level'? I agree though, fair play to him fronting up for a spur of the moment interview.

I actually found it refreshing, he made it clear he wouldn't have all the answers available, understandably, but nor did Steve go after him for the sake of it.

 

What's quite sad is that if it kicks off that bloke is likely to be the fall guy, too.

 

https://forum.overclock3d.net/showthread.php?p=993908#post993908

More BS. They seem so arrogant, yet seem to smack of "the hare" from the hare and the tortoise. Silly boys....

 

Understood, he did make it quite clear that he wasn't hands on with the testing. He did get points for fronting up rather than pushing staff out there.

Doesn't change the problems with testing though.

 

I guess so as by all accounts his company has been caught up in an internet pitchfork mob as he said he's been trying to deal with the fallout ever since Intel published the results so i guess he's spent the last few days answering all sorts of questions.

I'm not sure I'd go as far as saying i feel sorry for him as he's a co-founder and co-owner of the company so ultimately he's responsible for what and how they go about their business, but i do respect the fact that he was willing to answer questions, to the best of his ability, on camera.

Personally i don't think PT should re-run the test as unless they showed massive advantage to AMD the mob would still be calling for a lynching so IMO it's probably best to use it as a learning experience and allow the mob to move onto their next righteous indignation.

I also don't believe it was malicious, either before or after that video, however having answers makes it easier to dismiss, or refute, the mobs claims that it was.

In the end although the video may have made for awkward watching i feel it serves to draw a line under what could have dragged on for ages, not that it will stop some of the more extreme members of the mob as Steve said he received death threats for a bad review of the 1800x, but i feel it serves to take some of the wind out of the mobs sails.

 

Can't say as I'm really sympathetic truth be told, if your looking to ultimately proclaim bragging rights for an uber product it takes facts rather than just "We're Intel and we're the best" - the fact Intel have published said flawed tests as gospel.

It doesn't really say a lot about Intel's integrity or business practices (which isn't a first, no names mentioned former Intel CEO) - and the fact the company they used for the testing is clearly clueless as to quite how to test said products.

 

IIRC one of the Youtubers did mention they found it surprising that Intel didn't pick up on the flaws in the testing before publishing the results.

 

The people at Intel who would be capable of finding the flaws in the testing are highly unlikely to ever have even said hello to the people who approved publishing the results (which undoubtedly was a decision made at management level in the marketing department).

 

Probably but if you're paying money for independent testing you'd think someone would run an eye over what you get back just for the sake of not paying for something that isn't fit for purpose.

EDIT: If anyone is interested PT have published an official response in the form of this PDF.

 

Just as I found it surprising nobody at Intel picked up the fact that there is already an 8c16t chip on the market.

I find it difficult to trust what Intel say about anything at the moment or what they do or do not know.

 

PTs testing was well documented and correct, they have fallen over on AMDs over complex system for Ryzen/TR, I doubt it was malicious, Ryzen master has game mode, using game mode when testing games seems like the obvious thing to do.

Stock cooler OK no great, but it's not going to change the outcome by a significant margin, the game mode mistake is the biggest culprit but you need it for big TR and not for R7, more established review sites have made that error.

It would not surprise me if there are people out their unknowingly gaming on Ryzen with half the chip disabled.

 

I wonder if somewhere in the bowels of PT's email server is a short exchange between a possibly banana-short-of-a-fruitbasket tester and a low-level AMD PR employee with words to the effect of :
"We're benchmarking some AMD CPUs, do we need to replace the cooler that came with it with an aftermarket one for full performance?"
"Oh no, the included cooler is perfectly adequate for top performance!"


The really interesting thing in all this is that Threadripper's 'game mode' setting (intended to disable one die and eliminate the high latency Infinity Fabric substrate-level inter-die link from memory access times) has the bizarre effect of killing off half a single Zeppelin die, rather than the more logical function of simply having no effect (or being greyed out entirely). Why on Earth would that even be an implemented function in the first place‽

 

I wouldn't say correct, they were deeply flawed.

Also saying that the system for Ryzen/TR is overly complex seems rather subjective, i mean AMD recommend in their quick reference guide and full user guide that game mode should only be used with TR.

If PT took AMD's advise about the out of the box 2700x cooler being sufficient then why not the advise that game mode is only recommended for TR.

It would take forever if we were to discus all the flaws as no scientific 'test' is perfect, not only is it a constantly moving target but there's so many variables that it's neigh on impossible to account for them all, having said that PT's testing was deeply flawed IMO, that's not a reflection on PT as I'm fairly sure they had *very limited time to run these test.

* I say very as going on what BIOS versions they used, when new versions were released, and when they published i estimate they had something like 1-2 weeks with the 9900K, to test 16 system in 1-2 weeks is almost super-human IMO.

 

​

It is implemented because some older games and software get very crash happy or does not run with too many cores and this was a solution to work around.

Of course when you only have 8 cores loosing half is undesirable.