Linus Has Some Things to Say about Core i9 and X299

[TheLostSwede]

I've tested motherboards that wouldn't even get past POST despite multiple BIOS updates. One of the worst boards I ever got was a 955X board from Winfast shortly after Foxconn took over. It never actually worked, regardless of various processor and memory configurations and several new BIOSes.

I've tested boards from VIA that was supposed to work with I think it was DDR 266 memory or something along those lines, but it wouldn't have anything to do with the stick of Winchip memory that VIA supplied for testing.

I've tested IDE controllers from Promise that they shipped with non working firmware for no sensible reason.

I've tested ATI graphics cards (Rage Fury MAXX) that I don't think ever worked right due to poor drivers.

I've tested so many products over the years that I've forgotten about half of them, but at some point or another, every single company out there have released flawed products that simply doesn't do what they're supposed to do. Being an early adopter brings with it certain challenges, sometimes they're minor, sometimes you end up with a chocolate teapot...

 

[cdawall]

That's a bit premature IMO: how do you know the I9 works as intended right out of the box? Do you have a sample to test?

Until it's out, you shouldn't choose to go with I9 just because I7s work @ rated specs and Ryzen is a bit broken. When it's out, if indeed works @ rated specs, then you can go with I9 all you like.

In the history of Intel when have they released something that didn't work?

Even as far back as the early phenom days amd had massive issues (tlb bug).

I have a strange nagging feeling threadripper won't correctly clock ram, on top of other issues like behaving like it is a 4P system. Time will tell, but one of the two companies definitely releases a more finished product.

 

[HTC]

In the history of Intel when have they released something that didn't work?

Even as far back as the early phenom days amd had massive issues (tlb bug).

I have a strange nagging feeling threadripper won't correctly clock ram, on top of other issues like behaving like it is a 4P system. Time will tell, but one of the two companies definitely releases a more finished product.

One of two companies is (was) so far ahead they could afford to give partners 6+ months (if need be) to ensure compatibility for upcoming launches.

Not so far ahead anymore ...

Perhaps threadripper will have it's own set of problems, perhaps not: the point of my original reply was to not throw away a choice because it has a few issues and directly opt for another that is months away from being launched. Instead, wait for said choice to be launched and then, if the choice serves his needs, then choose it.

 

[TheLostSwede]

In the history of Intel when have they released something that didn't work?

Even as far back as the early phenom days amd had massive issues (tlb bug).

I have a strange nagging feeling threadripper won't correctly clock ram, on top of other issues like behaving like it is a 4P system. Time will tell, but one of the two companies definitely releases a more finished product.

Because Intel didn't have similar bugs?
https://en.wikipedia.org/wiki/Pentium_FDIV_bug
https://en.wikipedia.org/wiki/Pentium_F00F_bug
https://arstechnica.com/security/20...ecution-bug-that-lurked-in-cpus-for-10-years/

Why make AMD look bad when Intel is just as bad?

Both companies have had and will continue to have issues, can we leave it at that?

 

[FilipM]

People need to understand that there is no such thing as a perfect product. If we had a perfect CPU we wouldn't be upgrading at all.

Some products are more refined than others, but not perfect.

 

[trparky]

but honestly.. if you had 1200USD for a cpu would you get ryzen and have spare or spring for an i9?

If I had the money I would definitely go Intel, no doubt about it.

Intel might have to get even more desperate on pricing if too many people pass on these chips.

Considering that the whole X299 platform seems like Intel is giving the finger to their users I have a feeling that it's not going to sell as well as Intel would hope. The limitations of the platform alone are enough for me to stay clear of it. Things like the inability to use anything but Intel NVMe SSDs with their new special storage controller. Not only that but don't forget about the special "RAID Keys" that you have to buy in addition to the motherboard itself. The "keys" are nothing but a cash grab on an already expensive platform.

I certainly hope that Intel gets desperate on pricing, we need better pricing since some of Intel's prices are beyond stupid.

that the Coffee Lake S model 6 core will do much better sales wise. It just might prove to be the go to chip for those disgusted with Intel's X platform elitism and restrictions, yet not up for all the trial and error tweaking needed to make a Ryzen competitive.

I'm one of those people, I don't want to have to deal with the fact that you have to tweak this or tweak that to get Ryzen to work with memory. I pretty much want to plug in my RAM and have it work.

I also don't want to have to buy expensive system RAM because of some CPU interconnect (Infinity Fabric) that I feel has been gimped out of the gate. Why the hell they tied the speed of that thing to half the speed of the system RAM I have no earthly idea but it's just plain stupid.

I think it won't have issues because it's Skylake based. It's essentially a super-polished Nehalem core. This is why Intel almost never has bios problems. AMD braved new waters with Ryzen, and it's got both benefits and downsides.

I understand that when a new platform is brought about there's going to be teething issues but AMD exacerbated many of them by being so damn secretive with their partners. These RAM, UEFI, and other various issues should have been solved before the launch of the platform. I can't help but to think that AMD used their buyers as beta testers.

Many complaints about Ryzen platform out of the gate... memory is still sketch...

And that's a major reason why I'm not buying into Ryzen, that and the less IPCs that Ryzen has when compared to that of Intel. I have a few applications and games that require as much CPU power as I can throw at them and even then I need more simply because they're that damn un-optimized under the hood. Only more CPU power will be able to make them run better.

But you can see the stir it has caused with new people buying the platform being concerned about memory compatibility and speed.

That and the fact that you're pretty much forced to buy more expensive, higher speed system RAM just because AMD stupidly tied the speed of the Infinity Fabric to half the speed of the system RAM. Stupid.

 

[cdawall]

Because Intel didn't have similar bugs?
https://en.wikipedia.org/wiki/Pentium_FDIV_bug
https://en.wikipedia.org/wiki/Pentium_F00F_bug
https://arstechnica.com/security/20...ecution-bug-that-lurked-in-cpus-for-10-years/

Why make AMD look bad when Intel is just as bad?

Both companies have had and will continue to have issues, can we leave it at that?

So the bugs for the pentium happened before most of our new members were even born. The remote execution bug didn't touch performance. Intel has issues yes, not nearly as bad as amd. Go figure with an R&D thats larger than amd. Maybe I am just tired of half finished shit hitting the market or I'm still waiting for Vega. Something I was actually excited for.

 

[FordGT90Concept]

Really late to this party but I got what he was talking about within the first third of the video. He's absolutely right: Intel is making a bad situation worse. Trying to figure out why something on even an X99 board isn't running as it should is a PITA. There's so many asterisks on those boards based on which processor you choose to install in it; moreover, that documentation often becomes obsolete as new processors are released. Intel is making a bad situation worse with X299. The fact that DIMMs may be non-functional because of the chosen CPU takes the same principle to new extremes. We can no longer look at how many DIMMs are empty and decide what upgrades are available. We have to look up the CPU, look up the motherboard, find the manual where it references that specific CPU, then figure out which DIMMs will actually work.

I'm glad motherboard OEMs are lining up to skip the platform. Consumers should to. Intel is doing it on X299 for the same reason they did it on X99: encourage people to spend more money on a processor and not for processor performance. It needs to stop.

 

[trparky]

I think you misinterpreted what IPC is. What you described is NOT IPC (IPC is Instructions Per Cycle, the average number of instructions executed per clock cycle at any given clock speed) what you described above is instructions per SECOND.

In theory, one could get around lower IPC by increasing the clock speed since that would be increasing the average amount of instructions that are executed per second since there would be more clocks to do more work per second. At least in my mind it seems like it would work.

 

[cdawall]

One of two companies is (was) so far ahead they could afford to give partners 6+ months (if need be) to ensure compatibility for upcoming launches.

Not so far ahead anymore ...

Perhaps threadripper will have it's own set of problems, perhaps not: the point of my original reply was to not throw away a choice because it has a few issues and directly opt for another that is months away from being launched. Instead, wait for said choice to be launched and then, if the choice serves his needs, then choose it.

Intel also hands off an almost complete bios to companies hence why we don't normally see weird ram compatibility issues with those and why they all seemed the same.

I don't see us having too many issues with X299. If they don't sell Intel will adjust pricing so it does. I am excited for both platforms to hit the market and have zero faith after the ryzen launch that x399 will go well, but you know I'm going to test it lol.

In theory, one could get around lower IPC by increasing the clock speed since that would be increasing the average amount of instructions that are executed per second. At least in my mind it seems like it would work.

Ask Intel and amd how that worked with netburst and bulldozer.

 

[Frick]

Ask Intel and amd how that worked with netburst and bulldozer.

It worked for Pascal, and current Intel.

 

[FordGT90Concept]

Even if X399 doesn't go well at launch, you can bank on the fact they'll fix the hiccups eventually. At this point, I have zero interest in X299 and a lot of interest in x399 just because it vastly simplifies set up by all processors having all 64 PCIE lanes available.

 

[cdawall]

It worked for Pascal, and current Intel.

Pascal isn't pure clockspeed. IPC is still good and current Intel IPC is stronger than amd?

It is a balancing act and always has been you cannot go too far either way or you either have a useless heat box that needs to be clocked to the moon to do anything (bulldozer) or you can't overclock at all (amd Hawaii and later)

Even if X399 doesn't go well at launch, you can bank on the fact they'll fix the hiccups eventually. At this point, I have zero interest in X299 and a lot of interest in x399 just because it vastly simplifies set up by all processors having all 64 PCIE lanes available.

I thought it was 44? I have seen a couple of different postings on usable pcie now for them...

 

[Slizzo]

Intel also hands off an almost complete bios to companies hence why we don't normally see weird ram compatibility issues with those and why they all seemed the same.

I don't see us having too many issues with X299. If they don't sell Intel will adjust pricing so it does. I am excited for both platforms to hit the market and have zero faith after the ryzen launch that x399 will go well, but you know I'm going to test it lol.

You've got some awesome rose tinted glasses... The platform that you and I are currently on? X99? Guess what, it was a shit show when first launched.

From our very own TPU here: https://www.techpowerup.com/207979/intel-facing-hedt-chipset-troubles-again

Also, X79 had issues.

Just because Intel is a large company with a lot of money, does not mean that they always release a flawless product.


EDIT: Also, X299 is 28 or 44 PCI-E lanes depending on processor, and X399 is 64 PCI-E lanes across the entire Threadripper lineup.

 

[cdawall]

You've got some awesome rose tinted glasses... The platform that you and I are currently on? X99? Guess what, it was a shit show when first launched.

From our very own TPU here: https://www.techpowerup.com/207979/intel-facing-hedt-chipset-troubles-again

Also, X79 had issues.

Just because Intel is a large company with a lot of money, does not mean that they always release a flawless product.


EDIT: Also, X299 is 28 or 44 PCI-E lanes depending on processor, and X399 is 64 PCI-E lanes across the entire Threadripper lineup.

I had prerelease ryzen, wrote a review on it here. I will probably have a prerelease threadripper. Again I am excited for both, however amd can't see to do a proper release of anything. My h1 release vega card still isn't out.

 

[64K]

I had prerelease ryzen, wrote a review on it here. I will probably have a prerelease threadripper. Again I am excited for both, however amd can't see to do a proper release of anything. My h1 release vega card still isn't out.

Well H1 isn't over until June 30th but I don't expect any miracle to take place and AMD meet that deadline. Last they said was end of July. If they put it off again at that point it's going to get ugly on tech forums.

 

[Slizzo]

I had prerelease ryzen, wrote a review on it here. I will probably have a prerelease threadripper. Again I am excited for both, however amd can't see to do a proper release of anything. My h1 release vega card still isn't out.

Yeah, launch of Ryzen was rocky for sure; but I am willing to give AMD a little leeway here as they haven't launched a proper product in, what, 6? 7? years? They did rush it and didn't communicate with their board partners well, that's for sure. But they did correct issues quite quickly. I don't expect that X399 will go the same way; should be a good launch.

 

[cdawall]

Well H1 isn't over until June 30th but I don't expect any miracle to take place and AMD meet that deadline. Last they said was end of July. If they put it off again at that point it's going to get ugly on tech forums.

I have zero faith we will have anything more than a paper launch of the FE. This is getting old quick, why can't they release them so I can get the mining?

Yeah, launch of Ryzen was rocky for sure; but I am willing to give AMD a little leeway here as they haven't launched a proper product in, what, 6? 7? years? They did rush it and didn't communicate with their board partners well, that's for sure. But they did correct issues quite quickly. I don't expect that X399 will go the same way; should be a good launch.

Ryzen still has a multitude of bugs. I was playing with some non-hynix/samsung -b and still can't get the ram above 2666 with more than one board. The IMC had better be vastly improved with threadripper and I am hopping it isn't just two dual channel IMC's that are running in a ganged mode across the infinity fabric between the two sets of CCX units. That will be crap on a stick and is exactly how BD/PD worked for their higher core count MCM parts.

 

[TheLostSwede]

So the bugs for the pentium happened before most of our new members were even born. The remote execution bug didn't touch performance. Intel has issues yes, not nearly as bad as amd. Go figure with an R&D thats larger than amd. Maybe I am just tired of half finished shit hitting the market or I'm still waiting for Vega. Something I was actually excited for.

I think we're all tired of half finished shit, but it's also what has become acceptable in the world of early adopters for whatever reason. Customers are now beta and sometimes even alpha testers, because we live in a world where shareholders have to be pleased first and foremost and this means pushing out new products as quickly as possible. As long as it can be fixed in software, this is considered acceptable praxis. A lot of companies have actually gone bust before they've fixed all their software issues and a previous employer of mine is heading that way as well.

 

[Aenra]

How can some of you 'experts' have had so, so much trouble with Ryzens and memory freqs when a clueless person such as i did not? In two different Ryzen rigs thus far?

/rhetorical

Anyway, assuming one left all the immaturity (combined with some incredibly uninformed opinions) aside..
People spend thousands more for an extra tiny percentage of performance (say in cars) or an extra tiny percentage of quality (say in music). Has always been so and so it will remain. Very logical.

- If you have it, you get an Intel (as things currently stand) knowing you are being milked for it.
- If you don't have it or are unwilling to spend it, you get a Ryzen, knowing you are a few percentile units behind, but with half the cost.

It's really that simple folks.
(if you have managed not to fall into either category, congratulations, lol, either for fanboying or living in denial)

Back to my /popcorn

 

[Duality92]

I'll just leave this here.

 

[R-T-B]

How can some of you 'experts' have had so, so much trouble with Ryzens and memory freqs when a clueless person such as i did not? In two different Ryzen rigs thus far?

Timing. It was really bad early on I understand. Now? I bought one a week ago and had no trouble either.

Ryzen still has a multitude of bugs. I was playing with some non-hynix/samsung -b and still can't get the ram above 2666 with more than one board.

What AGESA were you playing with out of curiosity? 1.0.0.6 seems to have solved damn near any issue I can find.

 

[Aenra]

It was really bad early on

Nooo no no.. not that easy, not with me.

It's --your-- fault if you rushed, it's ---your--- fault for rushing and having expectations. Entirely yours.
(yours as in figuratively, am uninterested in targetting individuals)

Also, again.. this doesn't work with me. You paint the full picture, or quote me not. May i?

- People buying an 8core to run it on 'Balanced' mode. Whose fault is that. Now before you say it, no, i don't care about the new age trends and hug-me-a-tree mentalities and save the planet; putting mentality on top of logic is still a fault. Your fault (again figuratively).
- People eschewing disabling power states and wondering why or why not. Can link you post after post; yet they too had an "opinion" on Ryzen's bad launch.
- People not taking the time or lacking the knowledge to ensure some semblance of compatibility between their components. They too had an "opinion".
- People neglecting to switch to Gen2 (or entirely unaware of this) when tweaking their BCLK. They too had an "opinion".
- People not having done their reading and as such incapable of grasping SoC's import for both core and ram freqs.
- People so.. useless, so used to "we do it for you" mentality of today that are unable to even upgrade their own BIOS properly... enter a gazillion of issues i've never had simply because i updated the proper way; what a surprise (again, a gazillion posts i could link you of just this).
- People rushing to buy brands they shouldn't be rushing to buy except they were anyway, because such is the market, its mentality and its victims nowadays.
- People being so cheap.. so petty.. they HAD to save 30bucks, buy the non-X version of a cheap as peanuts CPU; and then complain about how it did not quite reach the 'X' variant's frequencies.. why really.. yet they too had an "opinion".
- People buying a certain brand's mobos even after they released a version that fried the mobo, lol.. and i can link you their posts, complaining it was Ryzen's fault for that too!
- People (can name you some we have here, as members), buying the wrong RAMs, entirely, and then making threads here asking for help. Which is O.K., unless they then complain, you guessed it, that this platform has "issues".

I could go on, so really, spare me.
I have made mistakes few here have. I have less than half the knowledge most here have; and yet twice thus far, nothing. All it took was reason. No AGESAs out, no beta mobo BIOSes out, just reason.

So while i respect your opinion (in regard to "early on it really was bad", aka their fault, not ours), i'm really not the one to help you with solidifying it

 

[Slizzo]

What AGESA were you playing with out of curiosity? 1.0.0.6 seems to have solved damn near any issue I can find.

This, yeah I'd like to know. 1.0.0.6 has been around for a little bit and from all reports has cleared up pretty much all issues with memory.

 

[EarthDog]

In theory, one could get around lower IPC by increasing the clock speed since that would be increasing the average amount of instructions that are executed per second since there would be more clocks to do more work per second. At least in my mind it seems like it would work.

.....and like fractions, do to one side what you did to the other..intel overclocks too, and more/higher clocks so.....