AMD AGESA 1.0.0.3ABA Buggy, Company Pulls it from Motherboard Vendors

Auer

Auer

[H]ard|Gawd
Joined
Nov 2, 2018
Messages
1,972
The latest version of AGESA ComboAM4 microcode that enables 3rd generation Ryzen support on AMD 400-series chipset motherboards has been deemed buggy and pulled from motherboard vendors.

AGESA ComboAM4 1.0.0.3ABA (not to be confused with 1.0.0.3AB that's being widely distributed), was originally released to fix an application crash noticed with "Destiny 2."
The microcode inadvertantly destabilizes PCI-Express on motherboards, with users of ASUS motherboards complaining of stability issues with the latest BIOS updates that include 1.0.0.3ABA.

https://www.techpowerup.com/257514/...ggy-company-pulls-it-from-motherboard-vendors
 
wyqtor said:
And this is, sadly, why Intel remains the only viable upgrade option for me. Time is the most valuable commodity for me and I don't want to waste it fixing AMD's screw-ups.
Click to expand...


Give it time, and like everything else (including Intel) it will be stable. I mean come on, they're making sure it works on a 1-2 year old socket/chipset.
The only thing that doesn't work so far is Destiny 2.....That and you might not get max boost clocks, but thats hardly a stability issue.

Enjoy your security holes which are a much bigger issue...(maybe not for you though, but for Intel)
 
Right from the article: Peter "Shamino" Tan from ASUS commented that the company was under a tight schedule to push 1.0.0.3ABA out as BIOS updates, and didn't have the time to properly validate it.

Gee, yeah, must be AMD's fault for Asus rushing things out. Things happen and this is why validation and testing is important, making excuses like he did is not smart.
 
ManofGod said:
Right from the article: Peter "Shamino" Tan from ASUS commented that the company was under a tight schedule to push 1.0.0.3ABA out as BIOS updates, and didn't have the time to properly validate it.

Gee, yeah, must be AMD's fault for Asus rushing things out. Things happen and this is why validation and testing is important, making excuses like he did is not smart.
Click to expand...

What boggles my mind is why is ASUS screwing up so much, maybe its just me or is ASUS really sucking at times with Ryzen support in general starting at Ryzen 1000 with the Crosshair MB.

I have an Asrock board I got at Ryzen 1000 launch, and apart from having to use 2666Mhz instead of 2933 on my RAM at first, its been a freaking awesome board(Stability, regular BIOS updates, relatively cheap). Asrock's got me sold lol.
 
wyqtor said:
And this is, sadly, why Intel remains the only viable upgrade option for me. Time is the most valuable commodity for me and I don't want to waste it fixing AMD's screw-ups.
Click to expand...

This so much this


Rockenrooster said:
Give it time, and like everything else .
The only thing that doesn't work so far is Destiny 2.....That and you might not get max boost clocks, but thats hardly a stability issue.
Click to expand...

And Linux, and 4 dimms, and latest bioses, and Nvidia cards (reliably) and .(as I said in another thread)..

  • Destiny 2
  • Linux
  • Memory card reader compatibility issues
  • Nvidia whea
  • Ryzen’s vr incompatibilities (vive)
  • Scheduler issues
  • Ram compatibility (including latency). Write latency on single die 3xxx models.
  • Issues with the clock of the infinity fabric
  • Issues with PBO not working properly
  • Sata issues
  • Incompatibilities with soundcards and some other peripheral cards
  • Incomplete ECC implementation
  • Linux compilation issues
  • Revision issues (some things were fixed and no one spoke about them) - consumers didn’t know which to get
  • Random software incompatibilities (Adobe during the last year)
  • Temperature reporting problems
  • Over reliance on AGESA to fix things (which means putting a lot of pressure on motherboard manufacturers to release bios updates)

It will end, this stuff will get better, it just takes time
 
Master_shake_ said:
because of intels stellar release of PCI-E 4.0... 2 years from now.
Click to expand...

nah must mean digs the numerous issues over the years with Intel based solutions, such as the rushed first i9 parts and the motherboards meant to support them, or is it buying that fancy expensive Intel based system to find out that there are loads of silicon level "bugs"

I rather deal with some minor issues and get a very well made system that really holds nothing back because I did not pay X more for the CPU to support it AND X more for the motherboard the cpu selection "allows" sort of speak.

Time is valuable as you state, so is the security of the system as well as the overall quality (how much $$$ they put towards making the part as best it could be vs minimum uplift for major increase in cost..

to each their own, support who you will, but just know supporting the gorilla's (Intel, Nvidia, Apple etc) only really gives them more $$$$$$$$$ in their pocket and ever more asinine way of acting like they are only ones who make something "viable"
 
It amazes me that AMD can still be so lack luster when it comes to supporting the motherboard and their OEM partners. I'm still a bit salty with how much of a cluster it was with the 3 series release, its good to see they haven't changed much.
 
When it is time to upgrade, I replace the mobo/cpu/heatsink/ram all together. Why the hell would I want to build a system, and replace the cpu every year? Sounds too expensive, it's not like cpu's are $49.99...

I think trying to support new cpu's on old shit mobos has bit them in the ass (by shit I mean not engineered for unknown future cpu's, because, you can't always predict the future, plus this is all left to the mobo manufacturers, and the quality definitely varies...).

I personally don't see the point. The majority of their customers are not replacing the cpu all the time. (Nor are Intels customers). Change the socket, annoy a tenth of 1 percent of whiny fanboys, but improve stability for everyone. This is a no brainer.
 
Keljian said:
  • Destiny 2 - Valid
  • Linux - Valid (though only present on 5.0 kernel or higher)
  • Memory card reader compatibility issues - Really?
  • Nvidia whea - Last I heard Nvidia is working on the problem not AMD
  • Ryzen’s vr incompatibilities (vive) - https://www.androidheadlines.com/20...-series-works-perfectly-with-vr-headsets.html
  • Scheduler issues - Not with 1903
  • Ram compatibility (including latency). Write latency on single die 3xxx models. - RAM compatibility? Bro Ryzen 3000 works with 4000 > RAM. I don't see a stability issue here
  • Issues with the clock of the infinity fabric - Running 2:1 over 3773MHz isn't an issue. It is meant to run that way.
  • Issues with PBO not working properly - Valid but not a stability issue.
  • Sata issues - Not sure if serious...
  • Incompatibilities with soundcards and some other peripheral cards - Who uses a sound card? Who is having these issues???
  • Incomplete ECC implementation - Better than NO ECC implementation lol.
  • Linux compilation issues - While there are issues with 5.0 kernels, you can still compile linux just fine. https://techgage.com/article/amd-ryzen-7-3700x-ryzen-9-3900x-performance-in-linux/
  • Revision issues (some things were fixed and no one spoke about them) - consumers didn’t know which to get - What issues???
  • Random software incompatibilities (Adobe during the last year) - Good old Adobe.
  • Temperature reporting problems - Haven't seen. Link?
  • Over reliance on AGESA to fix things (which means putting a lot of pressure on motherboard manufacturers to release bios updates) - Aren't BIOS updates a good thing???
Click to expand...


I see only 3 stability issues in your list.
A lot of those are non-issues or not AMD's problem.
 
Last edited:
GoodBoy said:
When it is time to upgrade, I replace the mobo/cpu/heatsink/ram all together. Why the hell would I want to build a system, and replace the cpu every year? Sounds too expensive, it's not like cpu's are $49.99...
Click to expand...

DUDE WHAT THE HECK?!!!
Do you hear yourself???
You talk about replacing mobo/cpu/heatsink/ram all together then call upgrading ONLY the CPU "sounds to expensive".
Dude...............
I have never seen this level of irrationality in hardforum before.

At least rest of your post thankfully makes a little more sense...

I bought a decent X370 board in 2017 and I expect it to last 5 more years at least. I have the option to upgrade to the 3950x down the line too... And if I get that, then I'll be using it for 5 more lol, not bad with a $150 mobo. People that want their PC to last will buy quality parts that will last a while. My board already has Ryzen 3000 support, i'm just waiting for prices to drop once Intel releases something to compete with 3950x.
 
Last edited:
Rockenrooster said:
What boggles my mind is why is ASUS screwing up so much, maybe its just me or is ASUS really sucking at times with Ryzen support in general starting at Ryzen 1000 with the Crosshair MB.

I have an Asrock board I got at Ryzen 1000 launch, and apart from having to use 2666Mhz instead of 2933 on my RAM at first, its been a freaking awesome board(Stability, regular BIOS updates, relatively cheap). Asrock's got me sold lol.
Click to expand...

I had an asus board, bought the cheapest asrock board.
Granted the asrock bios menu suck.. big time, but my god it's been good to me!

Asus board is now as the server, trying now desperately to get my 1700 + 64gb ram to work.

Asrock AB350M pro4, reset bios, insert the 1700 + 4 sticks 16gbX4.
Yes it works even at 2666mhz.

Asus board, set 1333 mhz, CL16 boots with two sticks (I have issues with them even in pair as dualchannel).
Enough tinkering of memory swapping and getting it eventually trained and it works.
Dimm A1\A2 is always fine, A1\B1,A2\B2 is hard to get working, hmm, trying another asus board, X370.. same issue.. hmm, trying an Msi Tomahawk, boots right up.

My mfr being sad? trying 64gb of hynix AFR.. no go.. what about cjr's.. no go.. my cjr's did 3466cl16 on my ab350 on 1700 on the asus board 2400 is even a struggle.
My experience is also.. Asus have gone to shit, and it's not just on amd side, the intel boards seems to have horrible memory training issues as well.. compared to Gigabyte.

My top pick must be Gigabyte in recent times, if only they made a Matx X570..
 
ole-m said:
I had an asus board, bought the cheapest asrock board.
Granted the asrock bios menu suck.. big time, but my god it's been good to me!

Asus board is now as the server, trying now desperately to get my 1700 + 64gb ram to work.

Asrock AB350M pro4, reset bios, insert the 1700 + 4 sticks 16gbX4.
Yes it works even at 2666mhz.

Asus board, set 1333 mhz, CL16 boots with two sticks (I have issues with them even in pair as dualchannel).
Enough tinkering of memory swapping and getting it eventually trained and it works.
Dimm A1\A2 is always fine, A1\B1,A2\B2 is hard to get working, hmm, trying another asus board, X370.. same issue.. hmm, trying an Msi Tomahawk, boots right up.

My mfr being sad? trying 64gb of hynix AFR.. no go.. what about cjr's.. no go.. my cjr's did 3466cl16 on my ab350 on 1700 on the asus board 2400 is even a struggle.
My experience is also.. Asus have gone to shit, and it's not just on amd side, the intel boards seems to have horrible memory training issues as well.. compared to Gigabyte.

My top pick must be Gigabyte in recent times, if only they made a Matx X570..
Click to expand...

Dang bro.
I have 2x16GB of Hynix RAM too, Glad I didn't get an Asus board lol.
 
Rockenrooster said:
DUDE WHAT THE HECK?!!!
Do you hear yourself???
You talk about replacing mobo/cpu/heatsink/ram all together then call upgrading ONLY the CPU "sounds to expensive".
Dude...............
I have never seen this level of irrationality in hardforum before..
Click to expand...

Well, my i7-920 system lasted 7.5 years (and still works), and the i7-6850k I am on is less than three years old. Buy a good complete setup, and it's a worthwhile investment. I do upgrade the GPU about every 2 years (but currently past that on the 1080Ti).

So its more upfront, but lasts longer. Stable. And once I upgrade, I got a near complete spare PC... 2 pc's for 1 upgrade price. And I just trickle those down to my second pc each upgrade cycle.

Pretty rational sounding to me...
 
Last edited:
wyqtor said:
And this is, sadly, why Intel remains the only viable upgrade option for me. Time is the most valuable commodity for me and I don't want to waste it fixing AMD's screw-ups.
Click to expand...

It is why you wait 3-6 months minimum on any new line if reliability is your highest priority. Or in this case, wait 1-2 years for AMD to get things right with ram and mobos.
 
As long as you can roll back to your last known good version without having to do a new system install, don't really see the issue. How many times has Microsoft had to do something similar with a Win 10 update that killed things? Or your favorite video card company with a driver release?
 
Alright, grab a sandwich boys.

Auer said:
The latest version of AGESA ComboAM4 microcode that enables 3rd generation Ryzen support on AMD 400-series chipset motherboards has been deemed buggy and pulled from motherboard vendors.

AGESA ComboAM4 1.0.0.3ABA (not to be confused with 1.0.0.3AB that's being widely distributed), was originally released to fix an application crash noticed with "Destiny 2."
The microcode inadvertantly destabilizes PCI-Express on motherboards, with users of ASUS motherboards complaining of stability issues with the latest BIOS updates that include 1.0.0.3ABA.

https://www.techpowerup.com/257514/...ggy-company-pulls-it-from-motherboard-vendors
Click to expand...

Nice one ASUS. I won't totally blame AMD for this. While they certainly made a mistake from the looks of it, the OEMs should test these things before release. However, the motherboard vendors and AMD are under incredible pressure to get issues sorted out. This includes the Destiny 2 crowd who want to play their game on their fancy new CPU's. It's a problem that really shouldn't exist anyway.

wyqtor said:
And this is, sadly, why Intel remains the only viable upgrade option for me. Time is the most valuable commodity for me and I don't want to waste it fixing AMD's screw-ups.
Click to expand...

Fair enough. This perspective isn't really all that uncommon and I want to point out that despite performance advantages AMD sometimes has, this is precisely why AMD has so much trouble in the server market competing with Intel. Its this attitude which is extremely prevalent there. I can count the number of physical AMD based servers I've encountered in a 22 year IT career on both hands with plenty of left over fingers. I've worked at a dozen different companies of varying sizes. The largest of which had tens of thousands of servers.

Rockenrooster said:
Give it time, and like everything else (including Intel) it will be stable. I mean come on, they're making sure it works on a 1-2 year old socket/chipset.
The only thing that doesn't work so far is Destiny 2.....That and you might not get max boost clocks, but thats hardly a stability issue.

Enjoy your security holes which are a much bigger issue...(maybe not for you though, but for Intel)
Click to expand...

You have some good points, but you aren't entirely correct. Destiny 2 isn't the only problem. There are Linux kernel issues as well. Furthermore, there are some issues that won't be fixed due to intrinsic problems with the design that cannot be resolved after the fact. Memory compatibility is one example with this. AMD isn't going to suddenly support faster speeds using four DIMMs as an example. Lastly, the boost clock issues have been resolved with AMD AGESA Combo PI 1.0.0.3 patch AB.

ManofGod said:
Right from the article: Peter "Shamino" Tan from ASUS commented that the company was under a tight schedule to push 1.0.0.3ABA out as BIOS updates, and didn't have the time to properly validate it.

Gee, yeah, must be AMD's fault for Asus rushing things out. Things happen and this is why validation and testing is important, making excuses like he did is not smart.
Click to expand...

This is a screw up on ASUS' part. While AMD may have screwed up the code, the motherboard vendors should catch this. MSI hasn't released a BIOS update using that AGESA code. Neither has GIGABYTE as far as I know. Interesting side note that sort of ties in here, is the fact that MSI pulled the MEG X570 GODLIKE v1.2 BIOS from its website which uses AGESA code 1.0.0.3 patch AB. This is the one I'm using on the test bench right now that resolved the boost clock issues on the GODLIKE board.

Rockenrooster said:
What boggles my mind is why is ASUS screwing up so much, maybe its just me or is ASUS really sucking at times with Ryzen support in general starting at Ryzen 1000 with the Crosshair MB.

I have an Asrock board I got at Ryzen 1000 launch, and apart from having to use 2666Mhz instead of 2933 on my RAM at first, its been a freaking awesome board(Stability, regular BIOS updates, relatively cheap). Asrock's got me sold lol.
Click to expand...

ASUS' problem is and always has been its desire for speed. It wants to be first with everything. They have a "If your not first your last" kind of mentality. To their credit, this served them well for the most part but there are times where this blows up in their faces. The Crosshair VI Example you brought up is a good one. ASUS never acknowledged this, but given we ran through three of those boards and plenty of other people had issues, it seems that design was flawed in a way that couldn't be fixed with BIOS updates. The 680i SLI based boards were atrocious as well. ASUS may have been one of the first out of the gate with these, but that wasn't a good thing. While better than the reference designs, they were still trash. In fairness that wasn't all ASUS' fault. The chipset was flawed as was NVIDIA's BIOS code. Now, that said I haven't experienced any major issues with any ASUS AM4 or TR4 motherboard between its horrendous X370 based Crosshair VI Hero and the release of X570. I don't have any of their X570 offerings on the bench, so I can't speak to those beyond what I've read here.

Master_shake_ said:
because of intels stellar release of PCI-E 4.0... 2 years from now.
Click to expand...

And this matters because? It probably won't take that long but even if it does, I doubt it will mean anything. Intel is always slow to adopt certain standards and it has its reasons, even if we don't know what they are. Even so, most people tend to agree that PCIe 4.0 isn't going to be a big deal for a very long time. The only immediate benefits we will see are in the realm of storage. Graphics cards, sound cards, and other devices won't benefit from this change. Right now, all its doing is driving the cost of motherboards through the roof.

Keljian said:
This so much this




And Linux, and 4 dimms, and latest bioses, and Nvidia cards (reliably) and .(as I said in another thread)..

  • Destiny 2
  • Linux
  • Memory card reader compatibility issues
  • Nvidia whea
  • Ryzen’s vr incompatibilities (vive)
  • Scheduler issues
  • Ram compatibility (including latency). Write latency on single die 3xxx models.
  • Issues with the clock of the infinity fabric
  • Issues with PBO not working properly
  • Sata issues
  • Incompatibilities with soundcards and some other peripheral cards
  • Incomplete ECC implementation
  • Linux compilation issues
  • Revision issues (some things were fixed and no one spoke about them) - consumers didn’t know which to get
  • Random software incompatibilities (Adobe during the last year)
  • Temperature reporting problems
  • Over reliance on AGESA to fix things (which means putting a lot of pressure on motherboard manufacturers to release bios updates)

It will end, this stuff will get better, it just takes time
Click to expand...

Let's go over the list.
  • Destiny 2 - Fix was inbound, not sure how long this will take.
  • Linux - See above.
  • Memory card reader compatibility issues - I haven't heard about this. My guess is that this is probably a USB issue, which wouldn't surprise me. I'd like to know more about this before speculating or commenting further.
  • Nvidia whea - NVIDIA needs to fix this, not AMD.
  • Ryzen’s vr incompatibilities (vive) - I don't know anything about this.
  • Scheduler issues - Windows 10 Build 1903 optimizes the scheduler for all Ryzen based CPU's. You really should be running 1903 on modern AMD CPU's, this is especially true of the Ryzen 3000 series.
  • Ram compatibility (including latency). Write latency on single die 3xxx models. - RAM compatibility is something I would expect to improve, but like previous Ryzen's, don't get your hopes up too much. This platform is fickle when it comes to RAM and always will be. Latency issues will probably always exist in models with multiple CCDs.
  • Issues with the clock of the infinity fabric - AGESA code can resolve this, although I haven't heard of specific problems with this.
  • Issues with PBO not working properly - PBO not working properly? Try: "PBO doesn't really do anything at all. We don't know that it ever will."
  • Sata issues - I haven't heard this, and even if true, I'd need to know more about what people are experiencing to comment further.
  • Incompatibilities with soundcards and some other peripheral cards - To be frank, this has been a problem with AMD going back to the Super 7 days. That said, this could be UEFI / AGESA code again.
  • Incomplete ECC implementation - Incomplete how? AMD includes support on the Ryzen 3000 series processors and the chipset doesn't preclude support for it. However, its up to motherboard manufacturers to implement the feature.
  • Linux compilation issues - See Linux above.
  • Revision issues (some things were fixed and no one spoke about them) - consumers didn’t know which to get - Not sure what the hell your talking about.
  • Random software incompatibilities (Adobe during the last year) - Fair.
  • Temperature reporting problems - Fair if true, but I haven't heard this. Largely, this may have nothing to do with CPU's nor be AMD's fault. Temperature reporting is often up to the motherboard.
  • Over reliance on AGESA to fix things (which means putting a lot of pressure on motherboard manufacturers to release bios updates) - I think this is a side effect of how the platform works. So I'll agree with this for the most part.

Additionally, you can already run 4x DIMMs on AM4 motherboards. You just lose speed doing it.

dragonstongue said:
nah must mean digs the numerous issues over the years with Intel based solutions, such as the rushed first i9 parts and the motherboards meant to support them, or is it buying that fancy expensive Intel based system to find out that there are loads of silicon level "bugs"

I rather deal with some minor issues and get a very well made system that really holds nothing back because I did not pay X more for the CPU to support it AND X more for the motherboard the cpu selection "allows" sort of speak.

Time is valuable as you state, so is the security of the system as well as the overall quality (how much $$$ they put towards making the part as best it could be vs minimum uplift for major increase in cost..

to each their own, support who you will, but just know supporting the gorilla's (Intel, Nvidia, Apple etc) only really gives them more $$$$$$$$$ in their pocket and ever more asinine way of acting like they are only ones who make something "viable"
Click to expand...

AMD's CPU's are designed to match Intel's pricing structure right now. AMD is pushing it's mainstream market options into a higher price bracket than Intel with it's upcoming Ryzen 9 3950X. That's $750 where Intel tops out at under $500 now. And X570 motherboards are more expensive than Z390 options if you want to look at the fact that AMD has a higher price ceiling than Intel. So, knocking on Intel for that now is an argument that no longer holds water. This is simply because AMD wants to be a premium option and not a second rate, loss leader of an option. It's simply because they want to be more expensive, not because it has to be. Let that sink in while you trash Intel on pricing.

Look, I've been among the first to call Intel out on pricing bullshit I don't like. In the HEDT market they went up $500 in one generation from $1,099 (Core i7 5960X) to $1,599.99 (Core i7 6950X) and again to nearly $2,000 the very next generation with it's higher end HEDT parts. Those kinds of increases year over year are absolutely insane. I've also called them out on their vROC bullshit as your paying an additional licensing fee to use a feature that's locked on the motherboard because......................reasons. I've said this before, and the proof is in today's online and printed ads. AMD is not some champion of the people who wants to look out for the little guy and set a shining example of corporate responsibility by being less greedy. They raised their prices because they wanted to. Period. AMD also charges as much or more than Intel when it thinks it can get away with it. They did it in the Athlon 64/FX era, and they are doing it now.

Let's keep things in perspective. AMD has generally not been good at the CPU market. While it has made some out standing products here and there, out of the last 25 years, it's only held the performance crown definitively, once. For about a five year period, it was dominant in that area. The rest of the time, its cannibalized itself by selling off profitable businesses and divisions that actually made money to support its CPU division which for most of the last three decades hasn't been all that profitable. AMD has largely mismanaged itself for decades which has led to a bad financial situation for years. It wasn't until their most recent CEO change that the ship started heading in the right direction. But some things never change. AMD charges as much as it can when it can. They are no different than Intel on that front.

IdiotInCharge said:
...that will still be meaningless, two years from now?
Click to expand...

More than likely.

Trimlock said:
It amazes me that AMD can still be so lack luster when it comes to supporting the motherboard and their OEM partners. I'm still a bit salty with how much of a cluster it was with the 3 series release, its good to see they haven't changed much.
Click to expand...

That's not actually it. I've covered this before, but it comes down to a ton of issues many of which aren't necessarily AMD's fault. The X370 launch was screwed up by Intel who moved up the Z270 launch to cause problems for AMD because they have the OEM's balls in a vice. Some of the issues are the motherboard vendor's fault. You need to understand that. AMD doesn't have the control over its board partners that Intel does. In essence, they do what they want. That's changing as we've seen with the X570 series, but back then it was very much the case.

Rockenrooster said:
Did you not see the post above where ASUS didn't test the updated AGESA?
QA 101...
Click to expand...

Exactly. Again, while AMD may have screwed up the AGESA code, the motherboard vendor should have caught it.

GoodBoy said:
When it is time to upgrade, I replace the mobo/cpu/heatsink/ram all together. Why the hell would I want to build a system, and replace the cpu every year? Sounds too expensive, it's not like cpu's are $49.99...

I think trying to support new cpu's on old shit mobos has bit them in the ass (by shit I mean not engineered for unknown future cpu's, because, you can't always predict the future, plus this is all left to the mobo manufacturers, and the quality definitely varies...).

I personally don't see the point. The majority of their customers are not replacing the cpu all the time. (Nor are Intels customers). Change the socket, annoy a tenth of 1 percent of whiny fanboys, but improve stability for everyone. This is a no brainer.
Click to expand...

As I've said many times, people give Intel flak for its socket changes which I agree are sometimes forced without a good reason that we are aware of or even out of sheer greed. Fine. However, we don't have the same problems on the Intel side we have on the AMD side. That's the benefit of socket / chipset changes as you pointed out. I don't think most people buy CPU's to upgrade existing machines all that often, but they are in love with the idea of doing it. Many people have this fantasy of having a machine for two or three years and then doing a round of upgrades to get another year or two out of it before moving on. Even though, they probably are better off building a new system at that time rather than spending 50% of the cost of a new one on upgrades and still getting left behind as it were.

The idea of upgrading your CPU was more attractive when CPU sockets didn't change quite as often and CPU's would practically double in clock speed (or more) during their life span. The Pentium chips went from 75MHz to 200MHz. Depending on when you bought your motherboard, you had options in that entire range. The Pentium more than doubled in speed over the time they were around. So a socket 7 motherboard could go as high as a Pentium MMX 233 or as low as a Pentium 75. This wasn't the greatest idea necessarily, because factors effecting performance were part of the design. Cache was on the motherboard itself and this varied greatly, impacting performance to a significant extent.
 
GoodBoy said:
Well, my i7-920 system lasted 7.5 years (and still works), and the i7-6850k I am on it less than three years old. Buy a good complete setup, and it's a worthwhile investment. I do upgrade the GPU about every 2 years (but currently past that on the 1080Ti).

So its more upfront, but lasts longer. Stable. And once I upgrade, I got a near complete spare PC... 2 pc's for 1 upgrade price. And I just trickle those down to my second pc each upgrade cycle.

Pretty rational sounding to me...
Click to expand...

How's that Broadwell E overclocking working out for ya? :LOL:
Running 1809 or newer? Checked what your actual clock is versus what you set in BIOS?
BTW, that bug can possibly kill a Broadwell E IMO (setting overclock including voltage bump, but current microcode FROM INTEL disabling anything higher than stock-I don't think they like cranked voltage at stock clocks)
 
Rockenrooster said:
DUDE WHAT THE HECK?!!!
Do you hear yourself???
You talk about replacing mobo/cpu/heatsink/ram all together then call upgrading ONLY the CPU "sounds to expensive".
Dude...............
I have never seen this level of irrationality in hardforum before.

At least rest of your post thankfully makes a little more sense...

I bought a decent X370 board in 2017 and I expect it to last 5 more years at least. I have the option to upgrade to the 3950x down the line too... And if I get that, then I'll be using it for 5 more lol, not bad with a $150 mobo. People that want their PC to last will buy quality parts that will last a while. My board already has Ryzen 3000 support, i'm just waiting for prices to drop once Intel releases something to compete with 3950x.
Click to expand...

Whoa there. It's more common than you think. You are over reacting without understanding one key point. How often does the person your responding to actually buy new CPUs? If the OP buys new CPU's every three to five years, then its not unreasonable to replace the motherboard. In fact, you almost certainly have to at that point. Even AMD generally changes sockets by then. Most people never upgrade the CPU in a given system. I worked in PC repair shops and a parts counter slinging PC upgrades. CPU purchases for upgrading existing systems wasn't all that common back when it was probably more common than it is now. This was back when five to seven year cycles were normal. When people would look at the cost of a CPU and memory upgrade to extend the life of their PC, they often saw they were half way or better to the cost of a new one and went that route instead. Or they realized that they'd spend hundreds of dollars on a machine that wasn't worth all that much and buy a new one. Only enthusiasts buy CPU's more frequently and that's the crowd that's also most likely to buy new motherboards more frequently as well.

There have been times where I bought CPU's frequently because there was something gained by doing it. Going from a Core 2 Duo to Core 2 Quad was an example of this. I kept the board during that transition. However, when I got my Core i7 5960X, there wasn't a compelling reason to go with a Broadwell-E based CPU. None of them would have been an upgrade in any way except the Core i7 6950X. That gave you two more cores. However, the minimal IPC gains were a wash given that you lost 200MHz of overclocking headroom and those two cores wouldn't have done anything for me in games. Add to that the price hike from $1,050 (Microcenter) I paid for my 5960X to $1,550 (Microcenter) for the Broadwell-E and it just wasn't worth it. Kyle and I each killed one of those CPU's overclocking them so I really didn't want that. By the time I felt a need to upgrade, I was basically at the end of the road for the socket and chipset. So a new motherboard was necessary anyway.
 
Last edited:
GoodBoy said:
When it is time to upgrade, I replace the mobo/cpu/heatsink/ram all together. Why the hell would I want to build a system, and replace the cpu every year? Sounds too expensive, it's not like cpu's are $49.99...

I think trying to support new cpu's on old shit mobos has bit them in the ass (by shit I mean not engineered for unknown future cpu's, because, you can't always predict the future, plus this is all left to the mobo manufacturers, and the quality definitely varies...).

I personally don't see the point. The majority of their customers are not replacing the cpu all the time. (Nor are Intels customers). Change the socket, annoy a tenth of 1 percent of whiny fanboys, but improve stability for everyone. This is a no brainer.
Click to expand...
If you think they change the socket to improve stability... I have bridge to sell you. Adding or reducing one pin has more to do with profit than stability.
 
jmilcher said:
If you think they change the socket to improve stability... I have bridge to sell you. Adding or reducing one pin has more to do with profit than stability.
Click to expand...

It doesn't improve stability by itself, but Intel locks the platform down to a handful of CPU models. This makes microcode easier to manage and you end up with far fewer BIOS issues or other issues that come from broader reaching CPU compatibility. I've written about the problems that go with longevity in sockets numerous times. Basically, Intel's platform is generally less problematic because of how its locked down and AMD's is more versatile for the end user, but there are a ton of pitfalls that come with that.
 
Seem's to me asus and amd do not get along.this is not the first time for this problem.
 
Not sure why people are freaking out. This was not widely distributed yet. It was the fix for destiny 2 and Linux systemd boot issues. The destiny 2 stuff only being thrown in so OEMS would push the non windows fix. lol

These things happen it's not like AMD just pulled 3000 support or something. The Linux systemd issues have workarounds already pushed for distros like Suse and Manjaro. Perhaps AMD needs to take another week or two with it. No big deal. The workaround for the randd issue will do until then. I imagine they where embarrassed a bit with the 3000 launch not working on newer Linux distros day one and rushed this fix a bit too much. Live and learn. I'm sure the next update will sort it.
 
ChadD said:
Not sure why people are freaking out. This was not widely distributed yet.
Click to expand...

Team Intel is desperate for anything they can get.

wyqtor said:
And this is, sadly, why Intel remains the only viable upgrade option for me...
Click to expand...

You are aware that Intel is still selling chips with massive security flaws, right?
 
"most people don't do it so no one should be able to do it"

Am2 -> am3+, I must've used 10 different chips on the same board, from 1st gen dual core athlons to 6 core phenoms. I would never throw that flexibility away for a handful of issues that come essentially with every new platform
 
Keljian said:
This so much this




And Linux, and 4 dimms, and latest bioses, and Nvidia cards (reliably) and .(as I said in another thread)..

  • Destiny 2
  • Linux
  • Memory card reader compatibility issues
  • Nvidia whea
  • Ryzen’s vr incompatibilities (vive)
  • Scheduler issues
  • Ram compatibility (including latency). Write latency on single die 3xxx models.
  • Issues with the clock of the infinity fabric
  • Issues with PBO not working properly
  • Sata issues
  • Incompatibilities with soundcards and some other peripheral cards
  • Incomplete ECC implementation
  • Linux compilation issues
  • Revision issues (some things were fixed and no one spoke about them) - consumers didn’t know which to get
  • Random software incompatibilities (Adobe during the last year)
  • Temperature reporting problems
  • Over reliance on AGESA to fix things (which means putting a lot of pressure on motherboard manufacturers to release bios updates)

It will end, this stuff will get better, it just takes time
Click to expand...

This is borderline trolling, the issues manifest with Ryzen's initial release which saw same board vendor complain about time constraints, those problems were fixed about 4-6 weeks after release with constant updates thereafter to ensure optimal performance. Despite all these "so terrible" problems AMD systems are flying off the shelves, i guess if AMD had 4 generations on the same socket they would also have things working to a tee as seen with Ryzen + X400 releases and no issues.
 
Dan_D said:
It doesn't improve stability by itself, but Intel locks the platform down to a handful of CPU models. This makes microcode easier to manage and you end up with far fewer BIOS issues or other issues that come from broader reaching CPU compatibility. I've written about the problems that go with longevity in sockets numerous times. Basically, Intel's platform is generally less problematic because of how its locked down and AMD's is more versatile for the end user, but there are a ton of pitfalls that come with that.
Click to expand...
My response was directed at the post above mine originally. Which basically said the socket change was for stability issues. I understand the concept. Look how apple controls their hardware and QA.
 
jmilcher said:
My response was directed at the post above mine originally. Which basically said the socket change was for stability issues. I understand the concept. Look how apple controls their hardware and QA.
Click to expand...

Fair enough. One point about that example, even Apple screws up from time to time. I remember a rash of iMacs with bad capacitors years after the capacitor issues had been corrected by everyone else in the industry.
 
GoodBoy said:
Well, my i7-920 system lasted 7.5 years (and still works), and the i7-6850k I am on is less than three years old. Buy a good complete setup, and it's a worthwhile investment. I do upgrade the GPU about every 2 years (but currently past that on the 1080Ti).

So its more upfront, but lasts longer. Stable. And once I upgrade, I got a near complete spare PC... 2 pc's for 1 upgrade price. And I just trickle those down to my second pc each upgrade cycle.

Pretty rational sounding to me...
Click to expand...

That sounds rational yes...
You saying upgrading only the CPU was too expensive was irrational.
You can do the same with AMD if you want.

X58 was a awesome platform, you can drop in some xeons for cheap, albeit old xeons.

Upgrading ONLY the CPU is ALWAYS going to be cheaper than building a whole new system of the same perf.

Example:
User1 buys Ryzen 1600 in 2017
User2 buys a 7700k in 2017.
New CPUs come out in 2019 and 1600/7700k is no longer adequate.
Both users have $500 to work with.
User1 can buy a 3900x for $500 and sell 1600 (not for much but still).
User2 cannot build a z370 PC for $500, User would have to sell 7700k system to recoup costs but sitll probably wouldn't be able to afford 9900k or 9700k system.

Everything else you say makes sense, I was just pointing out how dumb what you said was: "When it is time to upgrade, I replace the mobo/cpu/heatsink/ram all together. Why the hell would I want to build a system, and replace the cpu every year? Sounds too expensive,"
 
Last edited:
GoodBoy said:
When it is time to upgrade, I replace the mobo/cpu/heatsink/ram all together. Why the hell would I want to build a system, and replace the cpu every year? Sounds too expensive, it's not like cpu's are $49.99...

I think trying to support new cpu's on old shit mobos has bit them in the ass (by shit I mean not engineered for unknown future cpu's, because, you can't always predict the future, plus this is all left to the mobo manufacturers, and the quality definitely varies...).

I personally don't see the point. The majority of their customers are not replacing the cpu all the time. (Nor are Intels customers). Change the socket, annoy a tenth of 1 percent of whiny fanboys, but improve stability for everyone. This is a no brainer.
Click to expand...

I see a lot of people in other forums who only build systems from scratch. They use them until they feel it isnt fast enough, then build another system from scratch, not reusing any components.

I can't imagine doing things this way. I see my system as comprising of multiple independent parts, all of which I can either choose to upgrade at any point.

8 years ago when I picked up my current motherboard and CPU (Asus P9x79 & i7-3930k) I replaced a 990FX and Phenom II 1090T I had bought as a "hold me over for Bulldozer" purchase. The plan was to drop in a Bulldozer at launch, but since Bulldozer was the biggest disappointment in computing history, I never did. When I did that upgrade, I kept everything else in the system, case, PSU, cooler, drives, GPU, you name it.

Since that purchase, I have swapped out my GPU six times, replaced my 8GB of slower older RAM with 64GB of newer faster RAM (over time, in increments) and swapped out drives more times than I care to count.

This is the way I have done it ever since I built my first system in 1991. In a way, my current desktop is a spiritual successor to my first build, upgraded one or two components at a time ever since. The oldest part in my system today is probably my Creative X-Fi Titanium HD.

That said,I haven't done an in place CPU upgrade on my desktop in almost 20 years. IN that tie I've done only one in-place CPU upgrade, and that was on my Server. While Bulldozer sucked the donkey balls for the desktop, it actually wasn't a bad server CPU, and I used an FX-8120 as my server CPU for many years. At one point, Newegg had a great sale on FX-8350's, so I replaced the CPU. I don't use Consumer hardware in my servers anymore, but at the time, it was a nice and cheap little upgrade.

Now , if we go back to the Socket A days I used to do this all the time. My first Socket A system was a Duron 650 (which I was able to overclock to 950, which was pretty awesome) At some point it got in-place upgraded to an Athlon 1200. Then when AMD increased their FSB from 100Mhz to 133Mhz, I needed a new motherboard, but that replacement lasted me through a few more Athlon upgrades. That was pretty nice, but back then CPU's were much cheaper, AND there was actually a pretty huge difference when you upgraded them.

Today, socket longevity matters less. CPU advancement is so slow (evidenced by how my 8 year old CPU is still relevant and usable today, even though I am planning its replacement) I have had no need to replace my CPU in those 8 years, but many times I wished I had more modern features (NVMe slots and booting, DDDR4. etc.) and these require new motherboards.

Because of this I feel like socket longevity really doesn't matter as much anymore. You just don't need to upgrade the CPU that often, and when you do it's more for the features and those require new motherboards anyway.

Other things have an impact here too. For instance, the strain it puts on motherboard manufacturers, to have to do BIOS updates for new CPU's for years old discontinued boards. I also wonder how much more awesome Ryzen 3000 could have been if they weren't working under the design constraints of AM4, and instead could make it as great as possible on a new socket.

AMD would probably be well served to make AM4 the last long life socket, and move towards Intel's model.

As I see it the best approach is probably this:

New Arch and New Socket (example: Zen) -> Arch refresh + same socket (Example Zen+) -> New Arch + New Socket (Example Zen2) -> Arch refresh + same socket (Zen2+?) -> etc

All that said, I built a desktop for my Stepson in December of 2017 with a Ryzen 5 1700x and an MSI B350 Tomahawk. At some point, if there is a good sale I could see dropping a Ryzen 3000 in there as an upgrade. Why not?
 
Last edited:
Between this and MSI cutting corners trying to shoehorn bios updates into their mobos... B/c they only used 16-megabyte EEPROMs, I'm thinking my next choice will be ASRock
 
You must log in or register to reply here.
Back
Top