Intel vs amd build - stability argument

TechManCan · Jan 15, 2021

Here is one video that I came across that made me ask the queetion. He isnt the only reviewer that has said this.

JayzTwoCents

o1dschoo1 · Jan 15, 2021

TechManCan said:
Here is one video that I came across that made me ask the queetion. He isnt the only reviewer that has said this.

JayzTwoCents

cause i dont want a system that bsods with my memory pushed to the limit. Thats one thing i enjoy. I actually like ram tweaking more than cpu overclocking

Blameless · Jan 15, 2021

cstkl1 said:
as far as i know its amd issue not board vendors
its the delivery of agesa which controls hardware

AMD is certainly responsible for AGESA, but problems that can be traced to firmware aren't automatically due to AGESA. Plenty of boards with different behavior and different options despite using identical AGESA revisions.

AGESA itself rarely seems to be the issue, except for very early releases for a given processor generation.

cstkl1 said:
its like marketing told software what it wants without checking hardware (board design valid)

until today theres zero data-sheet to know what is the operating electrical condition or ram timings on imc.

amd argument is p&c . hence bsod.
a simple thing as default vrm loadline for cpu. does anybody know them??

AMD hasn't publicly released datasheets for their processors in years--which is unfortunate and of some concern to enthusiasts--but they're certainly providing this information to board partners. Lack of public datasheets isn't causing their platforms to be unstable, it's just making them slightly harder to tweak.

Something like the default loadline shouldn't be difficult to calculate, if one desires to do so.

And I'm not having any BSOD issues on any of my AMD setups.

cstkl1 said:
he bought 3800 ram. 3800 ram should work. year 2020 where ddr4 gonna be eol thats a waste in time to tweak.

No AMD processor officially supports DDR4-3800 (the best AM4 parts max out at 3200MTs, which is also the top JEDEC spec for DDR4). Expecting memory that is 300MHz beyond spec to work without tweaking, then being disappointed if it doesn't (or that XMP, an Intel specification, doesn't deliver stable settings), is not something one can reasonably blame AMD for.

Anything beyond spec is on the system builder to achieve.

o1dschoo1 said:
cause i dont want a system that bsods with my memory pushed to the limit. Thats one thing i enjoy. I actually like ram tweaking more than cpu overclocking

I enjoy tuning memory as well, and there is little doubt that Intel platforms have generally been able to take DDR4 further. However, the tuning friendliness/overclocking arguments are quite distinct from the 'build stability argument'. AMD isn't responsible for instability when memory or fabric clocks are pushed beyond what they claim to support, any more than Intel is responsible when someone's LGA-2011v3/X99 setup is unstable when trying to push DDR4-3600.

Blameless · Mar 9, 2021

cstkl1 said:
a simple thing as default vrm loadline for cpu. does anybody know them??

I recently measured this on a pair of my B550 boards.

Default load line seems to be in the ballpark of 0.7 to 0.75 milliohms.

I tested this with a 3900X in an ASRock B550 Phantom Gaming ITX/ax and with a 3700X in an MSI B550 Gaming Edge Wifi.

First I verified that the vcore and CPU VRM input current readings given in HWiNFO were close enough to my DMM measurements to be useful, then I ensured the "Power Reporting Deviation" figure for the default LLC setting never dipped significantly below 100%, applied a manual OC/voltage, and ran Prime95 in-place AVX 240k FFTs and looked at the droop vs. current. I was seeing almost exactly 100mV of droop at a 140A CCD load on the 3900X and a proportional drop at the lower load on the 3700X.

There is certainly some margin for error here, but I'm confident it's close enough for most purposes.

cstkl1 · Mar 9, 2021

Blameless said:
I recently measured this on a pair of my B550 boards.

Default load line seems to be in the ballpark of 0.7 to 0.75 milliohms.

I tested this with a 3900X in an ASRock B550 Phantom Gaming ITX/ax and with a 3700X in an MSI B550 Gaming Edge Wifi.

First I verified that the vcore and CPU VRM input current readings given in HWiNFO were close enough to my DMM measurements to be useful, then I ensured the "Power Reporting Deviation" figure for the default LLC setting never dipped significantly below 100%, applied a manual OC/voltage, and ran Prime95 in-place AVX 240k FFTs and looked at the droop vs. current. I was seeing almost exactly 100mV of droop at a 140A CCD load on the 3900X and a proportional drop at the lower load on the 3700X.

There is certainly some margin for error here, but I'm confident it's close enough for most purposes.

that is pretty high for idle to load state... but is that the default loadline done by the mobo
but is it confirmed the spec for the cpu..
i just wonder.. cause if ppl can measure this than they can check all mobos to find whats the exact " default"

i am finding the same problem sometimes with intel on lower/mid end mobo.. half the time its just wrong loadline..

Blameless · Mar 10, 2021

cstkl1 said:
that is pretty high for idle to load state... but is that the default loadline done by the mobo
but is it confirmed the spec for the cpu..
i just wonder.. cause if ppl can measure this than they can check all mobos to find whats the exact " default"

i am finding the same problem sometimes with intel on lower/mid end mobo.. half the time its just wrong loadline..

Intel's specified load lines are muck steeper (more vdroop) than this. From 1.0 milliohm on the Kaby Lake-X to 2.4 milliohms on the Coffee Lake-U parts, with 1.1-2.1 milliohms being typical on most mainstream S parts for the last several generations. The 10900K (or any other 125w Comet Lake-S part), for example, has a 1.1 milliohm load line spec: https://cdrdv2.intel.com/v1/dl/getContent/615211 (page 121).

AMD hasn't public published their electrical specifications, but I'd find it unusual for two boards of entirely different brands to default to the same incorrect load line...it's certainly not impossible though.

It should also be noted that changing the loadline on my AM4 setups results in "Power Reporting Deviation" figures that are significantly high or low at peak loads and setting a steeper LLC slope on the boards that allow it can be used to cheat PBO into giving higher multi-core boost clocks.

cstkl1 · Mar 10, 2021

Blameless said:
Intel's specified load lines are muck steeper (more vdroop) than this. From 1.0 milliohm on the Kaby Lake-X to 2.4 milliohms on the Coffee Lake-U parts, with 1.1-2.1 milliohms being typical on most mainstream S parts for the last several generations. The 10900K (or any other 125w Comet Lake-S part), for example, has a 1.1 milliohm load line spec: https://cdrdv2.intel.com/v1/dl/getContent/615211 (page 121).

AMD hasn't public published their electrical specifications, but I'd find it unusual for two boards of entirely different brands to default to the same incorrect load line...it's certainly not impossible though.

It should also be noted that changing the loadline on my AM4 setups results in "Power Reporting Deviation" figures that are significantly high or low at peak loads and setting a steeper LLC slope on the boards that allow it can be used to cheat PBO into giving higher multi-core boost clocks.

thats current reporting. no idea y they have these on z590 also.. is it to force false reporting. its like u want to lie for secreenshot lol.. really funny right.

i know the numbers
for skylake type cpu its 1.6mohm for 6 and below, 8-10 is 1.1

why these are important. as u can use trained loadlines so the ac is the vrm loadline and dc the cpu. asus boards allows this. intel boards even allows u to make your own loadline if u think the board default not good enough. but before that its all about vid overclocking.

my opinion all boards from both camp, instead of these current craze of marketing phases.. there should be transparency in loadlines or calibrated loadlines on everyboard via values. ...other one is a onboard switch to measure transient high and low.
boards should he measured via accuracy of their delivery of voltage based a vid request

o1dschoo1 · Mar 10, 2021

cstkl1 said:
thats current reporting. no idea y they have these on z590 also.. is it to force false reporting. its like u want to lie for secreenshot lol.. really funny right.

i know the numbers
for skylake type cpu its 1.6mohm for 6 and below, 8-10 is 1.1

why these are important. as u can use trained loadlines so the ac is the vrm loadline and dc the cpu. asus boards allows this. intel boards even allows u to make your own loadline if u think the board default not good enough. but before that its all about vid overclocking.

my opinion all boards from both camp, instead of these current craze of marketing phases.. there should be transparency in loadlines or calibrated loadlines on everyboard via values. ...other one is a onboard switch to measure transient high and low.
boards should he measured via accuracy of their delivery of voltage based a vid request

Theres a reason they market phases. 16 core overclocks can suck huge amounts of power.

Blameless · Mar 10, 2021

cstkl1 said:
my opinion all boards from both camp, instead of these current craze of marketing phases.. there should be transparency in loadlines or calibrated loadlines on everyboard via values. ...other one is a onboard switch to measure transient high and low.
boards should he measured via accuracy of their delivery of voltage based a vid request

Marketing on transients and ripple vs. load would certainly tell the enthusiast more about the actual properties of the VRM, but it probably wouldn't sell more boards. Phases are easier.

Would be nice if some reviewers would attach a scope to some of these boards though.

cstkl1 · Mar 10, 2021

Blameless said:
Marketing on transients and ripple vs. load would certainly tell the enthusiast more about the actual properties of the VRM, but it probably wouldn't sell more boards. Phases are easier.

Would be nice if some reviewers would attach a scope to some of these boards though.

maybe theres is a board that shows transient.. .. 😱

Chiraq · Mar 14, 2021

I went full Amd this time and I can happily report I have no stability issues apart from the ones I create myself. I just don't understand the reason for staying loyal to a brand when there's no real non fixable issues. FOr sure, there are quirks about the AMD platform that needs to be tweaked differently than Intel BUT that goes both ways!

I like 'em both.

Waiting for Cyrix to launch a third option.

ttang · Apr 14, 2021

My Ryzen 7 5800x experience, so 5 months after purchase it's still not stable. AMD platform is plagued with RAM and USB issues.

For most consumers the USB disconnects and a few reboots per day is acceptable but not for me as I use my PC for work.

If you need multi core unfortunately theres not much choice it's AMD 5900x or 5950x. For 8 cores are less then Intel is a option. You could wait for AMD to fix the issues (I have been waiting 6 months already) it will probably need a new chipset, BIOS updates haven't resolved theses issues currently.

Am looking for go Team Blue now that's 11th are released.

Chiraq · Apr 14, 2021

Which USB connections drop out? All of them?

TwoCables · Apr 14, 2021

ttang said:
My Ryzen 7 5800x experience, so 5 months after purchase it's still not stable. AMD platform is plagued with RAM and USB issues.

For most consumers the USB disconnects and a few reboots per day is acceptable but not for me as I use my PC for work.

If you need multi core unfortunately theres not much choice it's AMD 5900x or 5950x. For 8 cores are less then Intel is a option. You could wait for AMD to fix the issues (I have been waiting 6 months already) it will probably need a new chipset, BIOS updates haven't resolved theses issues currently.

Am looking for go Team Blue now that's 11th are released.

Which motherboard do you have?

What PSU do you have? Tell me the brand and the model. I know the PSU seems unimportant, but it's critical. Unexpected restarts shouldn't happen.

ttang · May 12, 2021

Asus C8I
Its mainly the USB 3 dropping out, not sure on USB-C as I haven't used it for anything

PSUs
EVGA SN G3 1200w
Silverstone SLT 800w

The only fix that works is to disable PCI-E 4.0, this kinda of defeats the point of going X570.

I have pretty much given up on this build now as work is v. busy at the mo, so I don't have time to investigate any further.

TwoCables · May 12, 2021

ttang said:
Asus C8I
Its mainly the USB 3 dropping out, not sure on USB-C as I haven't used it for anything

PSUs
EVGA SN G3 1200w
Silverstone SLT 800w

The only fix that works is to disable PCI-E 4.0, this kinda of defeats the point of going X570.

I have pretty much given up on this build now as work is v. busy at the mo, so I don't have time to investigate any further.

What about updating to the most current BIOS?

Elrick · May 14, 2021

TwoCables said:
What about updating to the most current BIOS?

It's never enough when it comes to AMD hardware. You will always be bios updating in that never ending pursuit of RELIABILITY, that shall never arrive.

That was why I went back to Intel. The stability and reliability is real nice....... using a six-month-old bios update.

TwoCables · May 14, 2021

Elrick said:
It's never enough when it comes to AMD hardware. You will always be bios updating in that never ending pursuit of RELIABILITY, that shall never arrive.

That was why I went back to Intel. The stability and reliability is real nice....... using a six-month-old bios update.

That hasn't been my experience. I've had my first AMD system up and running since April 11th, and I've had stability and reliability since the first day. I simply unboxed the motherboard, used BIOS flashback to get a newer BIOS to support the CPU, and it's been fine ever since. So, don't be so negative. Let's see how it goes for them with a newer BIOS if they haven't done so already instead of just assuming it's hopeless.

The red spirit · May 14, 2021

I would say that Intel generally had an edge in that department simply due to better enforcing specifications and dealing better with BIOS stuff. Historically Intel made better platforms from day one and had less disaster. But today? I don't know. On Intel side you have bullshit with TDP manipulation by all board makers, on AMD side you have USB issues and perhaps a ****load of BIOS updates. Rocket Lake was a flop and requires some good cooling even with proper PL values. Zen 3 is fine, but is overpriced and has occasional stability gremlin. I have to admit that for stability there's almost no difference now and it's not really the CPUs that determine that, but basically how much bullshit board vendors do and how good Intel and AMD is at enforcing specs and properly informing OEMs of how to build boards. AMD has struggled in the past with that as well as with stock coolers. Intel generally had more solid board drivers, but that's it.

The red spirit · May 14, 2021

D-EJ915 said:
Intel motherboards and server boards (actually from intel) were so unreliable the company I work for switched off of them even though their OEM kickbacks and partner pricing was amazing, our customers just hated them because of how often they would just die. Intel actually stopped manufacturing consumer boards in part because of this lol.

That's really odd, because Intel branded boards were well known to be the most stable. The main reason why they don't exist anymore is that they were pretty bland and lacked some enthusiast options. Many of them looked bad in comparison to others. Their availability in some regions sucked. They just never truly got into market, didn't know what people wanted more and called it quits. Those boards were mostly Foxconns with some careful work from Intel with BIOS. And they didn't really disappear, Intel NUCs still have Intel boards and they are really good in terms of BIOS.

Intel vs amd build - stability argument

TechManCan

o1dschoo1

Blameless

Blameless

cstkl1

Blameless

cstkl1

o1dschoo1

Blameless

cstkl1

Chiraq

ttang

Chiraq

TwoCables

ttang

TwoCables

Elrick

TwoCables

The red spirit

The red spirit

Top Contributors this Month

Recommended Communities