AMD and Intel say "Trust me Bro"

[GeoIV]

Why did everyone throw a fit about Linus's warranty, yet it seems excepted what Intel and AMD say about XMP and EXPO?

 

AMD literally seeded reviewers with 6000 EXPO memory, not sure about Intel, but I'm sure they are relying on XMP to post their comparison charts.

I think all reviewers should test CPUs at JEDEC until Intel and AMD officially state that they support and warranty XMP and EXPO. (It would be fine to have separate reviews/comparisons of overclocked states, but they should make clear that the results are not garmented or warranted) 

[Zando_]

Just now, GeoIV said:

I think all reviewers should test CPUs at JEDEC until Intel and AMD officially state that they support and warranty XMP and EXPO. (It would be fine to have separate reviews/comparisons of overclocked states, but they should make clear that the results are not garmented or warranted) 

Nah. 

 

Realistically everyone is going to use XMP or EXPO so thus reviewers test with those. But also, there's too much variance in individual CPU IMCs and individual motherboards to guarantee that X speed above spec will work 100% of the time, thus it isn't reasonable to expect Intel/AMD to do so. Let alone the variance in RAM sticks from just different dies in general to individual kits. 

[Bad5ector]

@Zando_, The OP might be referring to the reports of AMD chips burning themselves up and the speculation is that it could be related to EXPO.

[YoungBlade]

9 minutes ago, GeoIV said:

Why did everyone throw a fit about Linus's warranty, yet it seems excepted what Intel and AMD say about XMP and EXPO?

 

AMD literally seeded reviewers with 6000 EXPO memory, not sure about Intel, but I'm sure they are relying on XMP to post their comparison charts.

I think all reviewers should test CPUs at JEDEC until Intel and AMD officially state that they support and warranty XMP and EXPO. (It would be fine to have separate reviews/comparisons of overclocked states, but they should make clear that the results are not garmented or warranted) 

There are some review outlets who test at the rated memory speed of the processors. You can see in this best-of round-up from Tom's Hardware that the non-OC results used memory running at the stock speeds of the CPUs: https://www.tomshardware.com/reviews/cpu-hierarchy,4312.html

 

If you don't like reviewers who test things the way most users will actually try to run them, then go support the outlets who test in ways that are effectively artificial based on Intel/AMD spec pages.

[Zando_]

2 minutes ago, Bad5ector said:

@Zando_, The OP might be referring to the reports of AMD chips burning themselves up and the speculation is that it could be related to EXPO.

Referencing something requires referencing it. OP references performance charts always being done with XMP/EXPO, despite those technically counting as "overclocking" and not being fully supported/warrantied by the CPU manufacturers. No mention of the chips burning themselves, just a gripe about how reviews are done. 

[Levent]

AMD and Intel sell their products with written warranties. Some middle aged guy going through midlife crisis didnt and that’s why he was called out.

 

[Bad5ector]

Just now, Zando_ said:

No mention of the chips burning themselves, just a gripe about how reviews are done. 

Sure, but then what's this about Warranty? But I agree with your statement:

Just now, Zando_ said:

Referencing something requires referencing it.

 

[seanondemand]

Because some viewers have a weird parasocial relationship with creators. 

[Bad5ector]

1 minute ago, Levent said:

Some middle aged guy going through midlife crisis

36 is middle aged?

[Zando_]

Just now, Bad5ector said:

Sure, but then what's this about Warranty? But I agree with your statement:

From the original post:

11 minutes ago, GeoIV said:

I think all reviewers should test CPUs at JEDEC until Intel and AMD officially state that they support and warranty XMP and EXPO.

Bold added by me. 

[Bad5ector]

1 minute ago, Zando_ said:

Bold added by me. 

Gotcha... I felt like there might have been a thread that OP was pulling on... Like they have heard about the burnt out AMD chips and figured why offer EXPO but not offer warranty for it. But re-reading, you're right, they are referencing both AMD and Intel. But not sure what their warranty and the way the reviewers review the chips have to do with anything. But hey it's end of day here and I'm not firing on all cylinders. 

[porina]

6 minutes ago, YoungBlade said:

There are some review outlets who test at the rated memory speed of the processors.

I know Anandtech do, and they get a lot of flak for doing so as some enthusiasts claim they're crippling the CPU in favour of their competitor or similar. 

 

12 minutes ago, Zando_ said:

Realistically everyone is going to use XMP or EXPO so thus reviewers test with those.

Not everyone, maybe if you limited the scope to "people who visit this forum" then I'd go as far as "most".

 

I suppose in an unrealistic world, we'd see several ram configurations tested.

1, officially supported speed, JEDEC timings.

2, "typical" affordable XMP/EXPO ram

3, "extreme" XMP/EXPO ram

 

I'd like to throw in 2 vs 4 sticks, 1R vs 2R, but as you can see, the number of tests grows rapidly and no reviewer has time for that. At least, not in one go. so they then have to decide what their target audience cares most about and hope for the best.

[GeoIV]

32 minutes ago, Zando_ said:

Nah. 

 

Realistically everyone is going to use XMP or EXPO so thus reviewers test with those. But also, there's too much variance in individual CPU IMCs and individual motherboards to guarantee that X speed above spec will work 100% of the time, thus it isn't reasonable to expect Intel/AMD to do so. Let alone the variance in RAM sticks from just different dies in general to individual kits. 

They manage to get JEDEC guaranteed for all motherboards and sticks of RAM right?  Also not everyone can use XMP or EXPO (try running 4 sticks)

I'm just saying, they sell the product with the idea that EXPO/XMP are part of its 'regular' performance, more so than normal overclocking, but you are still playing a silicon lottery amongst other issues. 

The current issues with AMD has brought this to the front of my thoughts, considering I am currently now running my memory at JEDEC out of caution. But, it has always bothered me a bit that they could not guarantee performance, yet still get reviewed based off of it. 

[Zando_]

2 minutes ago, GeoIV said:

They manage to get JEDEC guaranteed for all motherboards and sticks of RAM right?

Because that's an incredibly basic spec, anything that can't run at least that will fail QA testing. They can't test with every possible RAM speed/timing combo that exists on the market, so they only guarantee the basic spec. Same as the silicon lottery for CPUs really, that exists because OEMs bin for the lowest common denominator, as they guarantee every CPU can hit the rated base/boost clocks. Some, often most of the CPUs in X SKU line can go a good bit faster, but not all of them so thus overclocking isn't guaranteed. 

 

[MageTank]

9 minutes ago, GeoIV said:

Why did everyone throw a fit about Linus's warranty, yet it seems excepted what Intel and AMD say about XMP and EXPO?

 

AMD literally seeded reviewers with 6000 EXPO memory, not sure about Intel, but I'm sure they are relying on XMP to post their comparison charts.

I think all reviewers should test CPUs at JEDEC until Intel and AMD officially state that they support and warranty XMP and EXPO. (It would be fine to have separate reviews/comparisons of overclocked states, but they should make clear that the results are not garmented or warranted) 

Reviewers were sent multiple kits of memory, including JEDEC kits. The reviewers guide also referenced performance data was captured at 5600Mhz. Reviewers using 6000Mhz memory chose to use it. When I received my samples, I tested at 5600.

 

As for why they don't warranty EXPO/XMP, that is very simple. They do not control what board partners and memory manufacturers do with their memory. If board partners like ASUS want to load more than just primary/tertiary timings with their profile and start pumping unstable SOC voltages to support 6000mhz+ memory, why should AMD (or Intel if we are talking VCCSA/VCCIO) be forced to eat the cost on that?

 

Their terms are very clear, they are not saying "trust us", they explicitly state XMP/EXPO voids their warranty. However, they have no means of proving XMP/EXPO was loaded either, so as long as you do not admit to it, you'll likely never have an issue redeeming it.

 

10 minutes ago, GeoIV said:

They manage to get JEDEC guaranteed for all motherboards and sticks of RAM right?  Also not everyone can use XMP or EXPO (try running 4 sticks)

The JEDEC group is comprised of dozens of companies working together on the standards, including AMD and Intel. They are required to design their hardware around JEDEC compatibility which is why they work across the different IMC designs. XMP/EXPO is not subjected to the same standards. XMP/EXPO DIMMs are sorted and binned by third parties that advertise speeds beyond the design capabilities of the IMC, it's not on them if something out of spec doesn't work. No overclocks are guaranteed, period.

14 minutes ago, GeoIV said:

The current issues with AMD has brought this to the front of my thoughts, considering I am currently now running my memory at JEDEC out of caution. But, it has always bothered me a bit that they could not guarantee performance, yet still get reviewed based off of it. 

They cannot (and should not) have control over what third party reviewers are doing with their product. AMD made the IMC limitations very clear to reviewers and test labs prior to shipping them out and people chose to run them with higher speed memory. Your issue should be with the reviewers, not with AMD/Intel.

 

Unless you have evidence of AMD telling reviewers to use 6000Mhz memory with their platforms, I really don't understand the issue here.

[Bad5ector]

47 minutes ago, GeoIV said:

The current issues with AMD has brought this to the front of my thoughts,

Woohoo validation!!!!!  

[-rascal-]

2 hours ago, porina said:

I know Anandtech do, and they get a lot of flak for doing so as some enthusiasts claim they're crippling the CPU in favour of their competitor or similar. 

 

Not everyone, maybe if you limited the scope to "people who visit this forum" then I'd go as far as "most".

 

I suppose in an unrealistic world, we'd see several ram configurations tested.

1, officially supported speed, JEDEC timings.

2, "typical" affordable XMP/EXPO ram

3, "extreme" XMP/EXPO ram

 

I'd like to throw in 2 vs 4 sticks, 1R vs 2R, but as you can see, the number of tests grows rapidly and no reviewer has time for that. At least, not in one go. so they then have to decide what their target audience cares most about and hope for the best.

 

I think it was Hammer on Box (Hardware Unboxed) that addressed this in one of their QA videos.

Apparently, not many people care to watch DRAM deep-dive / tuning / overclocking / analysis stuff.

On the videos that they did, they only get a fraction of the views (less than 50K?) ...and it takes hours of testing, editing, etc... so they said they will stop making them.

 

 

[porina]

2 minutes ago, -rascal- said:

I think it was Hammer on Box (Hardware Unboxed) that addressed this in one of their QA videos.

Apparently, not many people care to watch DRAM deep-dive / overclocking / analysis stuff.

On the videos that they did, they only get a fraction of the views...and it takes hours of testing, editing, etc... so they said they will stop making them.

Not many of their audience maybe, doesn't mean people aren't out there seeking it. I dropped out of their audience years ago, as while I like their testing, their commentary on the wider industry was painfully bad. Similar to GN Steve and even Linus, it feels like stirring the pot for the views, as they can't be as stupid as their screen persona in reality.

 

Painful as it may be many people just want a single number to say why X is better than Y and not think about the details.

[-rascal-]

2 hours ago, GeoIV said:

They manage to get JEDEC guaranteed for all motherboards and sticks of RAM right?  Also not everyone can use XMP or EXPO (try running 4 sticks)

I'm just saying, they sell the product with the idea that EXPO/XMP are part of its 'regular' performance, more so than normal overclocking, but you are still playing a silicon lottery amongst other issues. 

The current issues with AMD has brought this to the front of my thoughts, considering I am currently now running my memory at JEDEC out of caution. But, it has always bothered me a bit that they could not guarantee performance, yet still get reviewed based off of it. 

 

JEDEC is a rule / spec CPU + Motherboard + Memory manufacturer has to guarantee.

It's not getting JEDEC to comply on the X-part, it's X-part must run at JEDEC spec.

It's like... a "car" must have 4 wheels + 4 tires, otherwise it is not a "car".

@MageTank already went into more detail.

 

EXPO / XMP has always been overclocking, but it's so common now, that it's just considered a "regular" thing by your day-to-day PC gaming demograph.

Those less / non-enthusiast people don't know that.

They are listening to people / videos to buy this kit of memory, and then go into BIOS to enable EXPO / XMP, save and exit.

They don't even necessarily know what EXPO / XMP means, or what it does ... other than enabling it and now memory is running at DDR4-3600.

I still remember when XMP didn't exist, and you just had to overclock the memory the old-fashioned way; grind it out, trial-and-error.

 

Absolutely correct.

But you have people on even the forums here will create a post, asking for help, Oh, I enable XMP, but the PC won't boot.

Is it DOA? Did I get ripped off / scammed?

Silicon lottery can be a bitch.

 

Just going into the CPU / Motherboard / Memory section of the forums...first page.

 

 

 

 

[Dabombinable]

27 minutes ago, -rascal- said:

 

JEDEC is a rule / spec CPU + Motherboard + Memory manufacturer has to guarantee.

It's not getting JEDEC to comply on the X-part, it's X-part must run at JEDEC spec.

It's like... a "car" must have 4 wheels + 4 tires, otherwise it is not a "car".

@MageTank already went into more detail.

 

EXPO / XMP has always been overclocking, but it's so common now, that it's just considered a "regular" thing by your day-to-day PC gaming demograph.

Those less / non-enthusiast people don't know that.

They are listening to people / videos to buy this kit of memory, and then go into BIOS to enable EXPO / XMP, save and exit.

They don't even necessarily know what EXPO / XMP means, or what it does ... other than enabling it and now memory is running at DDR4-3600.

I still remember when XMP didn't exist, and you just had to overclock the memory the old-fashioned way; grind it out, trial-and-error.

 

Absolutely correct.

But you have people on even the forums here will create a post, asking for help, Oh, I enable XMP, but the PC won't boot.

Is it DOA? Did I get ripped off / scammed?

Silicon lottery can be a bitch.

 

Just going into the CPU / Motherboard / Memory section of the forums...first page.

 

 

 

 

What about DDR4 3200 with XMP that is 2666MHz default, and primarily just uses XMP for tighter timings than JDEC spec DDR4 3200 (it is part of the spec at 20-20-20-12.5) at 1.35V instead of 1.2V?

[TetraSky]

"Hello, mr AMD? I think my CPU died. Overclock? I don't know what that is. Oh in the BIOS? I only ever opened that to change the boot order. I have not touched anything else in there. I just put the computer together and installed windows to play some games. I don't know anything about anything else"
 

And there's your warranty.

[GeoIV]

6 hours ago, MageTank said:

Reviewers were sent multiple kits of memory, including JEDEC kits. The reviewers guide also referenced performance data was captured at 5600Mhz. Reviewers using 6000Mhz memory chose to use it. When I received my samples, I tested at 5600.

 

As for why they don't warranty EXPO/XMP, that is very simple. They do not control what board partners and memory manufacturers do with their memory. If board partners like ASUS want to load more than just primary/tertiary timings with their profile and start pumping unstable SOC voltages to support 6000mhz+ memory, why should AMD (or Intel if we are talking VCCSA/VCCIO) be forced to eat the cost on that?

 

Their terms are very clear, they are not saying "trust us", they explicitly state XMP/EXPO voids their warranty. However, they have no means of proving XMP/EXPO was loaded either, so as long as you do not admit to it, you'll likely never have an issue redeeming it.

 

The JEDEC group is comprised of dozens of companies working together on the standards, including AMD and Intel. They are required to design their hardware around JEDEC compatibility which is why they work across the different IMC designs. XMP/EXPO is not subjected to the same standards. XMP/EXPO DIMMs are sorted and binned by third parties that advertise speeds beyond the design capabilities of the IMC, it's not on them if something out of spec doesn't work. No overclocks are guaranteed, period.

They cannot (and should not) have control over what third party reviewers are doing with their product. AMD made the IMC limitations very clear to reviewers and test labs prior to shipping them out and people chose to run them with higher speed memory. Your issue should be with the reviewers, not with AMD/Intel.

 

Unless you have evidence of AMD telling reviewers to use 6000Mhz memory with their platforms, I really don't understand the issue here.

Thanks for detailed explanation. I may have had some wrong understandings about EXPO/XMP. I thought AMD/Intel had input in their design/usage.

I have heard AMD say that DDR5 6000 was the "sweet spot" That seems like an endorsement to use it to me, but I suppose it doesn't have to be taken as such.

I suppose my problem is as much with the reviewers as it is with the CPU manufactures, or maybe more. Its pretty unlikely for a reviewer to recommend a CPU to the general public based on how well it could overclock. However, most of them do compare CPUs based on tests using XMP/EXPO. AMD and Intel are just happy to let them do that, especially when it makes their CPU look better. 

[MageTank]

9 hours ago, GeoIV said:

Thanks for detailed explanation. I may have had some wrong understandings about EXPO/XMP. I thought AMD/Intel had input in their design/usage.

I have heard AMD say that DDR5 6000 was the "sweet spot" That seems like an endorsement to use it to me, but I suppose it doesn't have to be taken as such.

6000 is the sweet spot in terms of performance scaling relative to the fabric. The infinity fabric is decoupled this generation and defaults to 2000 when XMP is loaded, so at 6000 you are running an IF ratio of 3:2 (3000Mhz memory to 2000Mhz fabric). Pushing beyond 2000 is typically very unstable.

 

It's also a limit for most boards, especially if pushing multi-rank DIMMs or 2DPC configurations where you will struggle to push that speed, let alone go beyond it.

 

9 hours ago, GeoIV said:

I suppose my problem is as much with the reviewers as it is with the CPU manufactures, or maybe more. Its pretty unlikely for a reviewer to recommend a CPU to the general public based on how well it could overclock. However, most of them do compare CPUs based on tests using XMP/EXPO. AMD and Intel are just happy to let them do that, especially when it makes their CPU look better. 

I don't doubt that AMD/Intel enjoy the free press of their products being shown in the best light by reviewers using high speed memory with them, however I personally do not see that as an endorsement of doing so, just them not outright banning it. After all, processor/memory overclocking are advertised features, they just come with inherent risk. I see this as being no different to buying a Dodge Challenger Hellcat with a top speed of 203MPH. Sure, the car can go that fast, but I guarantee you Dodge does not endorse people doing so on the freeway or residential neighborhoods.

 

At the end of the day, component manufacturers can only guarantee performance based on the advertised specifications of the product. Anything obtained beyond that by overclocking is simply down to luck of the draw. You can certainly argue that reviewing these products outside of their advertised specification is misrepresenting the average expected performance of the product, but I certainly wouldn't consider this false advertisement on behalf of the component manufacturers. If you want to see products reviewed differently, provide your feedback to reviewers, seek out other reviewers that test to your standards, or consider getting into the review industry yourself and apply those standards.

[GeoIV]

1 hour ago, MageTank said:

6000 is the sweet spot in terms of performance scaling relative to the fabric. The infinity fabric is decoupled this generation and defaults to 2000 when XMP is loaded, so at 6000 you are running an IF ratio of 3:2 (3000Mhz memory to 2000Mhz fabric). Pushing beyond 2000 is typically very unstable.

I vaguely understood the concept, but it seemed like AMD was specifically encouraging reviewers to use that speed, but I could be reading into it wrong. 

 

 

1 hour ago, MageTank said:

...I see this as being no different to buying a Dodge Challenger Hellcat with a top speed of 203MPH. Sure, the car can go that fast, but I guarantee you Dodge does not endorse people doing so on the freeway or residential neighborhoods.

 

Yeah I can see that. I also think there are some parallels with how some manufactures and reviewers use prepped surfaces for 0-60 times, or rolling starts. Not quite the performance the average user can expect.


I certainly wouldn't want them to lock down the ability to OC, and reviewers should review CPUs showing their full capabilities in whatever way they see fit. I do wish what the "base" expectation was VS their OC'd state was made more clear though.
 

[GeoIV]

Maybe my angst should be more directed at Asus?