AMD Zen 3 (5000 Series), rumored 17% IPC gain.

[LePhuronn]

I didn't suggest it at all. That is YOUR summation of the solution.

Um OK, so care to clarify what you actually mean then?

AMD's biggest mistake here was making a family of CPUs (they all have the same name, socket, pincount, RAM compatibility) so mutually exclusive they do not even post on a board without a specific AGESA. Amateur hour execution.

"So what you're saying is..." has every place in a debate, because if you state something yet have objection to another making their own interpretation, rather than what you actually meant, then I'm going to request clarification.

So what is it you're saying?

 

[ubersonic]

Intel has a 30 year history of needing high frequencies to make up for their low IPC.

Apart ofc from that period between 2011 and 2017 where AMDs IPC was in the toilet and even their 5GHz inferno chips struggled to match Intel's 4GHz ones.

Not an Intel fan just thought I would point out that not insignificant detail.

 

[Martini1991]

Apart ofc from that period between 2011 and 2017 where AMDs IPC was in the toilet and even their 5GHz inferno chips struggled to match Intel's 4GHz ones.

Not an Intel fan just thought I would point out that not insignificant detail.

And before that AMD were behind on IPC
AMD have been behind on IPC since as long as I've been in PC's until Zen 3XXX.

The first system I built was an Athlon X2 and 45NM Intel CPU's were better IPC

 

[CAT-THE-FIFTH]

Main part i took away from the GN opinion statements is that AMD is so focussed on the lowest common denominator for sales and the whole industry is in a race to the bottom. Hence all solutions to issues have to be 100% effective and can't cost any money to the consumer. Like that method has ever worked.

The 'solution' AMD came out with to 'solve' the bios compatibility issue (sending out compatible CPUs) AMD created was just another level of hassle even though it addressed the problem and of course it was probably unsustainable from a cost perspective nevermind a customer satisfaction one. GN seems to suggest these support CPUs now finding their way onto the Chinese retail market is a good reason why old CPU support can't be deprecated

AMD's biggest mistake here was making a family of CPUs (they all have the same name, socket, pincount, RAM compatibility) so mutually exclusive they do not even post on a board without a specific AGESA. Amateur hour execution.

I actually looked back at some of the old AM3+ motherboards I used in builds,and they had 32MB BIOS chips,and costs would have been higher back then! Even socketed BIOSes and dual BIOSes back then!

But realistically,they still made the platform reference specifications,so if there is a lowest common denominator it's that. Certainly by the 400 series chipset launch,they should have got over the initial problems they had. There was nothing stopping at least the X470 motherboards being specified with one 32MB BIOS chip or a dual 16MB BIOS chips,so you could have a dual BIOS setup.

The fanboys would argue over the colour of *****, i love this mentality that when you buy a component you're somehow inducted into some window licking cult always on the lookout for negative comments that need to be leaped on. Childish at best.

I would argue if you own a company's products we should be more critical of said company,as we gave them their custom and voted with our pockets. If we defend their behaviour,they will just say its validation of their actions and continue forwards doing what they want. Pretty much what has happened with Intel and Nvidia.

And before that AMD were behind on IPC
AMD have been behind on IPC since as long as I've been in PC's until Zen 3XXX.

The first system I built was an Athlon X2 and 45NM Intel CPU's were better IPC

I have been more seriously into PCs for 20 years,and once Intel Core was out in laptops,AMD started to loose their IPC advantage,as Intel pretty much had got to similar levels by then,and at lower power consumption. When the Core2 came out,that was when AMD fell behind. Zen2 is the first time they have been ahead since the Athlon 64,but Intel still wins in pure performance due to latency and higher clockspeeds. Also their latest mobile CPUs,are probably not worse in IPC anymore,from the reviews I have read.

 

[humbug]

And before that AMD were behind on IPC
AMD have been behind on IPC since as long as I've been in PC's until Zen 3XXX.

The first system I built was an Athlon X2 and 45NM Intel CPU's were better IPC

You're showing your young age there

Not sure about the K5 and K6 but the first Athlon's had higher IPC than Intel's Pentium 4, as far as i know that was the first case where Intel clocked the proverbials off them making them boiling hot to keep up with AMD's much lower clocked more efficient Athlon's. this was around the turn of the millennium.

The problem is AMD didn't sell many of those outside of self builders because Intel were paying OEM's not to use AMD's CPU's, not long after that AMD started to run into financial problems.

 

[Skynet5]

You're showing your young age there

Not sure about the K5 and K6 but the first Athlon's had higher IPC than Intel's Pentium 4, as far as i know that was the first case where Intel clocked the proverbials off them making them boiling hot to keep up with AMD's much lower clocked more efficient Athlon's. this was around the turn of the millennium.

The problem is AMD didn't sell many of those outside of self builders because Intel were paying OEM's not to use AMD's CPU's, not long after that AMD started to run into financial problems.

Ahh the K6. My first chip bought with my own money. Upgrading from my old 486dx4 my dad got me before going to uni.

The mobo had a lovely coil whine that went when you moved the mouse.

25years later and my new shiny build has almost the same problem lol. Although not really noticable and seems to be under certain types of load.

 

[CAT-THE-FIFTH]

You're showing your young age there

Not sure about the K5 and K6 but the first Athlon's had higher IPC than Intel's Pentium 4, as far as i know that was the first case where Intel clocked the proverbials off them making them boiling hot to keep up with AMD's much lower clocked more efficient Athlon's. this was around the turn of the millennium.

The problem is AMD didn't sell many of those outside of self builders because Intel were paying OEM's not to use AMD's CPU's, not long after that AMD started to run into financial problems.

The K5 was the first proper in-house AMD design and came out in the Pentium era in the 1990s,and even though it had better integer performance than comparable Intel CPUs,it had a worse FPU. Also there were different versions of the K6. The K6 wasn't really an in-house AMD design,it was designed by a company called NexGen as the original K6 was canned. The K6-III was the fastest socket 7 CPU. That was back in the day when AMD and Intel could share sockets.

The first Athlon came out around the time of the P3. It actually used aspects of the DEC Alpha CPU,as AMD had licensed or bought up DEC patents. That is when Intel started to sweat:
https://www.tomshardware.com/reviews/intel-admits-problems-pentium-iii-1,235.html

They ended up with releasing a barely stable P3 as the Athlon was so quick.

 

[CuriousTomCat]

Apart ofc from that period between 2011 and 2017 where AMDs IPC was in the toilet and even their 5GHz inferno chips struggled to match Intel's 4GHz ones.

Not an Intel fan just thought I would point out that not insignificant detail.

Oh you mean when Intel used bribery and coercion to prevent OEMs selling AMD which nearly bankrupt them? That period in history is null and void.

The 90s and early 2000s were the glory days.

Of course, Ryzen is on a whole new level

 

[Augustus]

"So what you're saying is..." has every place in a debate, because if you state something yet have objection to another making their own interpretation, rather than what you actually meant, then I'm going to request clarification.

No it doesn't have every place in a debate. It is a statement, not a request for clarification. It is also widely used to twist the stated words to fit the other side of the argument in an attempt to trap the opponent by their own words. In context, this is exactly how you used it.

The conversation ceases to be a debate at that point and there is nothing positive to be gained from it.

Had you wanted clarification you would have mentioned you didn't understand what i'd written and then:

Um OK, so care to clarify what you actually mean then?

So what is it you're saying?

I meant what i said in my last reply. AMD's execution of this socket compatibility has been poorly executed and that they should have thought about it a bit more to create or design in a less clunky way of implementing it.

The options for executing better than they have and i have subsequently mentioned, in this thread, an option i think would be the best one. I'm sure there are lots of other, easier, ones that give a less satisfactory result. Rather than trying to ensure seamless socket compatibility AMD has tried to walk away from a main plank of their AM4 marketing with a flimsy excuse.

Your point well made about Intel suggests my assertion is valid in general although Intel gets away with it because the lack of socket compatibility masks it.

 

[CAT-THE-FIFTH]

Oh you mean when Intel used bribery and coercion to prevent OEMs selling AMD which nearly bankrupt them? That period in history is null and void.

AMD had a huge IPC gain during the 90s, early 2000s and of course, the classic Athlon 64.

AMD did some stupid things though. You have to appreciate until 2002,their CEO was Jerry Saunders. He was really a character(apparently) but fearless in taking on Intel despite Intel having tried to screw over AMD in the 1990s too. You should see the whole second source fiasco during the late 1980s and early 1990s. But he had no problems realising when AMD had problems,and even canned the original K6,when he saw a startup have a better design(NexGen) and bought the company. When HP divested it's DEC holdings,he bought or licensed many DEC patents,and acquired former DEC engineers such as Jim Keller in 1999. The Athlon 64 has some of it's roots in the DEC Alpha. People talk about Lisa Su,but it was literally a David vs Goliath fight back then.

However,AMD CEOs after that,made some questionable design choices,despite Jerry Saunders having put the foundations in place,ie,the Athlon 64. One of them was overpaying for ATI,which meant AMD couldn't invest enough in it's own fabs,ie,what we know now as Global Foundries. It also left the coppers empty regarding funds of backup CPU designs I suspect. The CEO involved in the ATI deal was implicated in this too:
https://www.engadget.com/2009-10-27...uiz-caught-up-in-insider-trading-scandal.html
https://www.marketwatch.com/story/former-amd-chief-tells-a-bit-but-not-all-2013-02-21

Another was the decision to prioritise production of earlier Athlon X2s over the new Phenom which was pushed backwards,as the former was more profitable,and then they hadn't anticipated the Core2 for some reason. This is despite a year or so previous,Intel had made the Core for laptops which showed itself to be a very strong contender. Then after that the whole Bulldozer uarch. It was originally designed to use a more compact core with shared resources to partially get over AMD's fab disadvantage of the time. It was meant to be 45NM,but AMD just kept on with it,and we got what we got. If Jerry Saunders was probably in charge he would have canned it,but they kept on with it.

This is all on top of what Intel was doing with its bribes.

It was only really until Rory Read came along,we had someone who actually understood things a bit better. He was the one who was responsible for AMD getting consoles,and actually trying to form an ecosystem of sorts.

 

[Gerard]

Oh you mean when Intel used bribery and coercion to prevent OEMs selling AMD which nearly bankrupt them? That period in history is null and void.

The 90s and early 2000s were the glory days.

Of course, Ryzen is on a whole new level

I fondly remember my 1.33ghz Thunderbird and that cooler that was popular at the time which was infamous for cracking the core if put on incorrectly. Had to use a screwdriver for one side of it that had a slot so you could get the bracket pushed down over the socket clip. Can't recall the name of the cooler but seen a few people crack their cores when installing it.

ed:

actually think it was this, the delta fan on it sounded like a drill.

 

[CAT-THE-FIFTH]

I fondly remember my 1.33ghz Thunderbird and that cooler that was popular at the time which was infamous for cracking the core if put on incorrectly. Had to use a screwdriver for one side of it that had a slot so you could get the bracket pushed down over the socket clip. Can't recall the name of the cooler but seen a few people crack their cores when installing it.

Also if you didn't put the cooler on properly,you could have a smoky CPU!

I still remember,companies selling the copper shims,you could put around the CPU,to reduce the chance of a cracked die.

 

[Gerard]

Also if you didn't put the cooler on properly,you could have a smoky CPU!

I still remember,companies selling the copper shims,you could put around the CPU,to reduce the chance of a cracked die.

Yeah they had the shims then they started putting rubber feet on the cpu like you can see in that video. That cooler at 00:59, near sure i have mine still. Noise control silverado. Not very common but good coolers for the time

 

[CuriousTomCat]

I fondly remember my 1.33ghz Thunderbird and that cooler that was popular at the time which was infamous for cracking the core if put on incorrectly. Had to use a screwdriver for one side of it that had a slot so you could get the bracket pushed down over the socket clip. Can't recall the name of the cooler but seen a few people crack their cores when installing it.

Yes I remember those issues. A 3rd party company released a shim which fitted around the CPU core to stabilise the heatsink and prevent it from cracking.

I was still using a Duron 850MHz overclocked to 1050MHz during that time so it didn't affect me. Then I upgraded to an Athlon XP 2400 Thoroughbred-B which had rubber feet around the CPU core to stabilise the heatsink and prevent cracking.

Edit: You guys beat me to it lol

 

[CuriousTomCat]

AMD did some stupid things though. You have to appreciate until 2002,their CEO was Jerry Saunders. He was really a character(apparently) but fearless in taking on Intel despite Intel having tried to screw over AMD in the 1990s too. You should see the whole second source fiasco during the late 1980s and early 1990s. But he had no problems realising when AMD had problems,and even canned the original K6,when he saw a startup have a better design(NexGen) and bought the company. When HP divested it's DEC holdings,he bought or licensed many DEC patents,and acquired former DEC engineers such as Jim Keller in 1999. The Athlon 64 has some of it's roots in the DEC Alpha. People talk about Lisa Su,but it was literally a David vs Goliath fight back then.

However,AMD CEOs after that,made some questionable design choices,despite Jerry Saunders having put the foundations in place,ie,the Athlon 64. One of them was overpaying for ATI,which meant AMD couldn't invest enough in it's own fabs,ie,what we know now as Global Foundries. It also left the coppers empty regarding funds of backup CPU designs I suspect. The CEO involved in the ATI deal was implicated in this too:
https://www.engadget.com/2009-10-27...uiz-caught-up-in-insider-trading-scandal.html
https://www.marketwatch.com/story/former-amd-chief-tells-a-bit-but-not-all-2013-02-21

Another was the decision to prioritise production of earlier Athlon X2s over the new Phenom which was pushed backwards,as the former was more profitable,and then they hadn't anticipated the Core2 for some reason. This is despite a year or so previous,Intel had made the Core for laptops which showed itself to be a very strong contender. Then after that the whole Bulldozer uarch. It was originally designed to use a more compact core with shared resources to partially get over AMD's fab disadvantage of the time. It was meant to be 45NM,but AMD just kept on with it,and we got what we got. If Jerry Saunders was probably in charge he would have canned it,but they kept on with it.

This is all on top of what Intel was doing with its bribes.

It was only really until Rory Read came along,we had someone who actually understood things a bit better. He was the one who was responsible for AMD getting consoles,and actually trying to form an ecosystem of sorts.

I had no idea about any of that. I guess they're both in it to win at all costs

 

[Augustus]

I actually looked back at some of the old AM3+ motherboards I used in builds,and they had 32MB BIOS chips,and costs would have been higher back then! Even socketed BIOSes and dual BIOSes back then!

So GN's point about 16MB bios chips being ubiqutous and cheap is another opinion point not grounded in fact then.

16MB is a backward step.

I get the feeling that all of the 'tech press' are desperate not to take sides here. A lot of GN's points excusing AMD feel flimsy to me but they have to make then to appear neutral.

Your research on the bios chips on AM3+ supports my point that all the manufacturers are in a race to the bottom. In this scenario the quality of the product always suffers.

But realistically,they still made the platform reference specifications,so if there is a lowest common denominator it's that. Certainly by the 400 series chipset launch,they should have got over the initial problems they had. There was nothing stopping at least the X470 motherboards being specified with one 32MB BIOS chip or a dual 16MB BIOS chips,so you could have a dual BIOS setup.

I was referencing the opinion in the GN segment where it was mentioned the reason why having specific bios per CPU or bios branches cannot exist is because the employee in Best Buy can't retain all that info along with the info needed to sell mobile phones and televisions. Then it will be too complicated for users to look up different specific bios that matches their motherboard and CPU. This is the SALESPERSON & CONSUMER lowest common denominator and not be board design lower common denominator.

AMD is making decisions and creating solutions for all of us based on the lowest possible level of competency within a scenario where they have engineered in a level of complexity they think is above the LCD salesperson/consumer.

AMD created a paradox.

The only universal solution that works is that the CPUs 'operate' with any bios. What 'operate' means is completely up for debate. Any CPU operating at a sufficient level will remove all of the silly processes put in place up to now as well as stopping people immediately sending components back due to 'no worky'. Context of this solution is any AM4 CPU, on any AM4 chipset 'operates' with any bios for a given motherboard.

Motherboards can already access the internet from the bios so as long as a CPU is socket compatible it could even update itself to the correct bios!

I would argue if you own a company's products we should be more critical of said company,as we gave them their custom and voted with our pockets. If we defend their behaviour,they will just say its validation of their actions and continue forwards doing what they want. Pretty much what has happened with Intel and Nvidia

Agreed. It's one of the reasons i've piped up on this thread. It has got to the point where it feels like more and more liberties are being taken from all companies across all their products such that it's become a SOP and needs to be stopped.

 

[CAT-THE-FIFTH]

Yeah they had the shims then they started putting rubber feet on the cpu like you can see in that video. That cooler at 00:59, near sure i have mine still. Noise control silverado. Not very common but good coolers for the time

I was too scared I would crack the die,so used one even if it did make cooling potentially a tad worse. I can't remember the model of cooler I used,but IIRC it was copper one. Had my XP2800 in a MSI KT6 variant,and then decided a few years later,thought smaller PCs looked interesting so went and got a Shuttle XPC with an nForce2 chipset. That XP2800 lasted me a while!

I had no idea about any of that. I guess they're both in it to win at all costs

AMD gets it's underdog status,and slayer of giants status due Jerry Saunders as he was one of the founders. The guy had some real balls,and he was of the view AMD would win,despite whatever Intel would try and do:
https://en.wikipedia.org/wiki/Jerry_Sanders_(businessman)

He steered the company through hard times as well. In 1974, a particularly bad recession almost broke the company. Through many difficult recessions he refused to lay off employees, a reaction to the rampant layoffs that had occurred at Fairchild earlier. Instead of reducing employees, he asked them to work Saturdays to get more done and get new products out sooner. There were also good times for the company. Sanders gave each one of his employees $100 as they walked out of the door during AMD's first $1M quarter. AMD was also the first US company to implement a cash profit-sharing employee compensation program, where employees would regularly get profit checks of $1,000 or more.

His maxim was "People first, products and profit will follow!" This was given as a printout for each AMD worker who started a job at AMD in Dresden until Sanders's retirement

Some of the people after him,did make some questionable decisions,which were compounded by yet another round of Intel playing the douche.

 

[CuriousTomCat]

Your research on the bios chips on AM3+ supports my point that all the manufacturers are in a race to the bottom.

Yeah but their aim is the get the most sales as possible. If manufacturer A sells a motherboard with a 32Mb BIOS and manufacturer B sells a competing motherboard with a 16Mb BIOS, the later will get more sales because they undercut.

That's why some Windows 10 laptops were sold with a 32Gb SSD. They undercut competing products with a bigger SSD. Tech savvy people know better but most people aren't tech savvy.

Although surely most motherboard sales are by tech savvy people anyway. So I dont know.

 

[CuriousTomCat]

Motherboards can already access the internet from the bios so as long as a CPU is socket compatible it could even update itself to the correct bios!

This. That would be such an awesome feature. Surely it cant be too difficult.

Edit: Opps I should have combined those into one post.

 

[CAT-THE-FIFTH]

So GN's point about 16MB bios chips being ubiqutous and cheap is another opinion point not grounded in fact then.

16MB is a backward step.

I get the feeling that all of the 'tech press' are desperate not to take sides here. A lot of GN's points excusing AMD feel flimsy to me but they have to make then to appear neutral.

Your research on the bios chips on AM3+ supports my point that all the manufacturers are in a race to the bottom. In this scenario the quality of the product always suffers.

They saddle a line,otherwise they can cut off. 10~20 years ago the tech press could be far more savage about companies,ie,just look at people like Thomas Pabst who started Toms Hardware.

Yes,the motherboards are in a race to the bottom. Intel compounded this when they locked all overclocking out of H and B series chipsets,and started all this K series locked multipliers stuff.

To put it in context,even sub £150 AM3/AM3+ motherboards could have VRMs which could handle between 200W~300W with massive heatsinks,etc.

This is how VRM cooling looked on higher end AMD motherboards:
https://www.techpowerup.com/review/ecs-a990fxm-a/images/board_front.jpg
https://www.techpowerup.com/review/asus-m5a99x-evo/images/board_front.jpg

So $165 to $240 class motherboards. This is the more mainstream motherboards:
https://www.techpowerup.com/review/asus-m5a97-evo/images/board_front.jpg
https://static.gigabyte.com/Product/2/3907/4462_m.jpg
https://techreport.com/r.x/amd-890gx/asus-board.jpg
https://techreport.com/r.x/amd-890gx/gigabyte-board.jpg

Those were around the $100 to $140 mark.

Also the whole 32MB BIOS chip problem,could be solved as I mentioned before,by using a dual 16MB BIOS chip system,aka,as dual BIOS.

You can have one BIOS on one chip,and another BIOS on another. This is not a new thing.

Unfortunately to appeal to later newer generation PC builders,companies are more worried about cosmetics and RGB,then engineering first.

I was referencing the opinion in the GN segment where it was mentioned the reason why having specific bios per CPU or bios branches cannot exist is because the employee in Best Buy can't retain all that info along with the info needed to sell mobile phones and televisions. Then it will be too complicated for users to look up different specific bios that matches their motherboard and CPU. This is the SALESPERSON & CONSUMER lowest common denominator and not be board design lower common denominator.

AMD is making decisions and creating solutions for all of us based on the lowest possible level of competency within a scenario where they have engineered in a level of complexity they think is above the LCD salesperson/consumer.

AMD created a paradox.

The only universal solution that works is that the CPUs 'operate' with any bios. What 'operate' means is completely up for debate. Any CPU operating at a sufficient level will remove all of the silly processes put in place up to now as well as stopping people immediately sending components back due to 'no worky'. Context of this solution is any AM4 CPU, on any AM4 chipset 'operates' with any bios for a given motherboard.

Motherboards can already access the internet from the bios so as long as a CPU is socket compatible it could even update itself to the correct bios!

Some of these decisions still sound like we are stuck back in 2002!! Honestly,16MB BIOS chips wasn't even that big 10 years. Also these types of memory chips must have cost more 10 years ago,when compared to today. Also,even if adding dual BIOS added another $3 to the base price of an AMD motherboard,I don't think OEMs charging $5 more will be all of a sudden make such a big difference.

UEFI came out a decade ago!! The whole way BIOSes are handled are archaic and wouldn't look out of place a 20 years ago.

Agreed. It's one of the reasons i've piped up on this thread. It has got to the point where it feels like more and more liberties are being taken from all companies across all their products such that it's become a SOP and needs to be stopped.

Customers need to make their voices heard,as increasingly companies are less worried about the "customer is always right" mantra and more the "shareholders is always right" mantra,which often conflicts. This is what is happening with PC gaming right now - lots of microtransactions,etc is driven by it.

You have these large investment firms expecting more and more margins,profits,etc each year. You should see a few names crop up when it comes to Intel,AMD and Nvidia who are among the biggest shareholders in these companies.

So higher and higher prices,cheaper engineering and more built-in obsolescence. Companies can try and fight this but are at the mercy of these people who are all about the short term. What they don't understand,is that short-termism does not work longterm. They look at consumers as financial units to be milked. So more boom and bust.