Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
videocardz.com
Well, no real surprise there.Raja Koduri is leaving Intel - VideoCardz.com
Raja Koduri leaves Intel Raja Koduri, the former head of the Intel’s AXG unit is leaving Intel by the end of March. As part of the company’s reorganization, Intel has decided to part ways with Raja Koduri who was Executive Vice President and Chief Architect. According to Intel CEO, Pat...
Raja will step down as Intel Chief Architect and leave Intel after 5 years at end of March 2023. Raja will create a new software company startup to generative AI for gaming, media & entertainment.
Had my 2700K running at 5Ghz on a D14 iirc, lovely cpu that.
In transistors the ACM-G10 seems halfway between Navi22 and Navi21,and I would argue drivers and support in games are what is not helping performance. Supposedly most of the driver development team being in Russia caused big problems too. Just look at some of the Chinese dGPUs too - the hardware looks capable enough,but it seems again drivers and support in games which are the problem.
This is a problem with Intel IGPs,even predating Raja being there. At least AMD has the excuse of being a relatively small company,what excuse did Intel have for over a decade?
AMD,despite being in consoles and having a long history of involvment with games devs,has the same issues compared to Nvidia too,so it will be even harder for Intel in this regard but they had the resources. Nvidia has always been pushing games to favour its own features and strengths or capitalising on AMD uarch weaknesses,ie,think tessellation or more use of RT reflection effects. Even further back,lack of DX10.1 support.
Just look at Vega after a few years:
AMD Radeon RX Vega 56 8 GB Review
Radeon RX Vega 56 is the second AMD Vega card launched today. It comes at an affordable $399 price point, with a slightly reduced shader count that actually improves things greatly. For example, power efficiency now trades blows with some Pascal cards, which means less heat and noise, too.www.techpowerup.com
NVIDIA GeForce RTX 2060 Founders Edition 6 GB Review
NVIDIA's GeForce RTX 2060 doesn't seem cheap at $349, but this new card has enough steam to beat AMD's RX Vega lineup at better pricing and with much better power/heat/noise. Actually, the RTX 2060 obsoletes much of NVIDIA's GeForce 10 Pascal stack, too.NVIDIA GeForce RTX 3070 Ti Founders Edition Review
NVIDIA's new GeForce RTX 3070 Ti is designed to bring the fight to Radeon RX 6800 non-XT. To achieve their goal, NVIDIA maxed out the GA104 GPU design and switched to faster GDDR6X memory. Unfortunately, this also resulted in an increase in power consumption and heat output.
Look where Vega56 was - barely beating a GTX1070 and GTX1080 and RTX2060 beating it. It pretty much is now matching a GTX1080/RTX2060. Sometimes it can match a GTX1080TI:
It was always a design designed for commerical usage - the first Vega cards were commercial cards. AMD repurposed it because they had no other dGPUs to use above the RX480 8GB. But it still lives on in cDNA which seems relatively successful. Vega also did pretty well in AMD IGPs until relatively recently.
Also,people forget it was Raja who was involved with Polaris,and Vega was developed during a time with very little R and D money at AMD. IIRC,over 70% of AMD R and D was being pushed towards designing Zen.When AMD nearly went bankrupt a few years ago(share price started tanking),they massively cut R and D spend due to Bulldozer being a disaster, and AMD having locked themselves into the Global Foundries contracts which meant they had to pay penalties for not hitting order requirements. IIRC,it was the period after the R9 290 series was launched too. RDNA/RDNA2 was started under his watch too.
RDNA3,OTH,might have been the first uarch which he had no involvement with.
Polaris was the final iteration on GCN.
Vega was all Raja and ARC has all the same odd behaviours of Vega, right down to what games it does well in and what games it does badly in, its a copy of... with a few tweaks i'm sure.
I am going to blow my own trumpet here, i was pretty much alone on this forum predicting a Raja Kuduri involved with ARC would result badly and end badly.
I was ridiculed, Raja is a genius, its all AMD's fault, they didn't give him enough money, he would be awesome at Intel, they will give him all the money he needs, RIP AMD. They said.... AMD were glad to get the _____ rid of him. I guarantee you that.
Raja at Intel
AND: you know those piles of cash you keep comparing to ours..... wave them bye bye.
I was bang on, and if i was to say "RIP Intel" now i would be very wrong, but still far more accurate than that ^^^^ was ever going to be. I was right about that too.... Jon Peddie? pft... these people couldn't build an accurate analysis if their lives depended on it, they don't have the instict for it and you can't teach that.
While i'm at it.... i was also right about Nvidia. I should do this sort of analysis for a living, many people paid 50X more than i do keep getting it wrong where i get it right.
Polaris was the final iteration on GCN.
Vega was all Raja and ARC has all the same odd behaviours of Vega, right down to what games it does well in and what games it does badly in, its a copy of... with a few tweaks i'm sure.
I am going to blow my own trumpet here, i was pretty much alone on this forum predicting a Raja Kuduri involved with ARC would result badly and end badly.
I was ridiculed, Raja is a genius, its all AMD's fault, they didn't give him enough money, he would be awesome at Intel, they will give him all the money he needs, RIP AMD. They said.... AMD were glad to get the _____ rid of him. I guarantee you that.
Raja at Intel
AND: you know those piles of cash you keep comparing to ours..... wave them bye bye.
I was bang on, and if i was to say "RIP Intel" now i would be very wrong, but still far more accurate than that ^^^^ was ever going to be. I was right about that too.... Jon Peddie? pft... these people couldn't build an accurate analysis if their lives depended on it, they don't have the instict for it and you can't teach that.
While i'm at it.... i was also right about Nvidia. I should do this sort of analysis for a living, many people paid 50X more than i do keep getting it wrong where i get it right.
www.oregonlive.com
He was a manager,with good experience of the software side,and some knowledge of the hardware side. But you need to have a clue about software to target where the hardware can be good at. He was at S3 Graphics,ATI,AMD,Intel and Apple over 30 years. He can't be that bad if so many leading companies hired him.
Raja Koduri was director of advanced technology development at ATI from 2001 to 2009. Many of us older folk know of him because of that. He joined ATI to work on the Radeon R300,aka,the 9700 series and the follow up X800 series. The reality is that when he was at ATI they made decent hardware most of the time(OK the X1800 had some issues and the R600 was a flop),but the 9000 series,X800 series,X1900 series,HD4000 series and HD5000 series were developed when he was there. The R300 was the DX9 graphics hardware standard for years.
He was at Apple,and was behind the push to Retina displays and probably their move to their own dGPU designs.
When he came back a second time to AMD you had Polaris,Vega and RDNA1/RDNA2. Considering Polaris and Vega had to be made on a subpar GF 14NM,because the people at AMD before Rory Read put AMD into a lose/lose scenario to stave off bankruptcy. Just look at the R and D budget for graphics at the time against Nvidia. 70% of AMD R and D was put towards CPUs. Their total R and D was still less than Nvidia. Vega was released firstly as compute card called Radeon Instinct. Only two Polaris designs were made.
Because of all the moaning AMD fans,they went re-released these cards to consumers,and had to make a loss on them. That is one of the biggest mistakes AMD made - they should have just not bothered,as it was a loss leader card.
Yet,Polaris wasn't a flop in the end? Same as Vega. It had relevance for 6 years as a uarch,especially since it was designed on a shoestring budget. Vega IGPs are still part of the Ryzen 5000 series APUs - they are still beating what Intel has. CDNA1 and CDNA2 are Vega derivatives and RDNA1 was developed when he was there,and by extension that means RDNA2.
It takes years to develop new designs,and more importantly get the design teams together.
The guy has worked at S3 Graphics,ATI,AMD and Intel since 1996. Have people not noticed he likes jumping into challenging scenarios? S3 was already having issues back then when he joined,and ATI when he joined in 2001 was a massive underdog compared to Nvidia at the time. The R300 was the design which made ATI's name in the industry.
He joined Apple as "Director, Graphics Architecture" meaning he had a hand in that too.So like Keller he stepped into something different. Rejoins AMD,literally as they are trying to starve off bankruptcy. Goes to Intel to try and push into a mature market with two incumbents. Now is going to found an AI gaming software startup.
If he was so bad,then he would have been out of the industry a long time ago,not nearly 30 years. Even Jim Keller only stays for a few years at many places,yet gets all the credit for those companies achievements years later. Jim Keller had good things to say about him IIRC.
Raja Koduri gets dunked on,even when a number of the designs he was involved with have stood the test of time.
Plus Jim Keller joined Intel only in 2018 and left two years later.Who do you think got Jim Keller onboard at Intel? Raja Koduri. Are people blaming him for the current Intel design problems? Why did he leave so quickly? He spend more time at AMD.Intel is having problems,left,right and centre and almost all their units.They let go of 1000s of older,more experience workers:
Intel layoffs skew older, spotlighting plight of aging workers
People over 40 were two-and-a-half times more likely to lose their jobs in this spring's layoffs than Intel employees under 40.
CPU,NAND,GPU,etc. The company has had too many people butting heads inside for the last 5~8 years. To try and blame Raja Koduri for all of them,when so many people have left comes across as people getting annoyed AMD couldn't beat Pascal,on a fraction of the budget.It's not only the hardware but the whole software stack for 100s if not 1000s of games. If it was so easy,then look at the numerous Chinese dGPUs,some made on 7NM which have horrible performance due to poor drivers and lack of gaming support. Apple has taken years to move to its own GPU uarch(they started moving towards it in 2014),and even then look at gaming performance? This is with a company that controls their whole ecosystem.I remember back in the late 90s and early 2000s,when we had more dGPU companies out there,how software could make or break things.
Lisa Su had made the correct decision to push more budget to CPU R and D. When AMD was competitive people just bought more Nvdia graphics cards anyway,irrespective of drivers,power consumption,etc.
Yet,where is all the moaning,that AMD still lost to Nvidia with RDNA2 on a far better node,or Nvidia is now ahead again with Ada Lovelace? This time R and D is much more,and AMD went back onto TSMC full time. They had access to TSMC 4NM(their new APUs are made on them). Why not make their new dGPUs on TSMC 4NM,like Nvidia? They clearly think APUs are more important. The reality is dGPUs are just a secondline priority for AMD,and Nvidia throws money at them.
The second reality is Intel,is cancelling projects everywhere and there is massive infighting. They have no clear vision of anything. Just look at their advanced packaging technologies,stuff such as L4 cache,etc. Intel had all the technology to out-AMD,AMD years ago. They had a process node advantage over AMD until 2019. It only took until Zen3 for AMD to finally field a faster core. They had decades to improve the software side on the IGPs.
semiaccurate.com
It takes 3~5 years to design something new. Even when Jim Keller was at AMD the first time he was there until late 1999. He was lead architect on the K8/Athlon 64 which came out over 3 years later. Raja Koduri left in late 2017. Navi samples were already floating about in 2018. So he was involved. Vega was a compute uarch made for commercial usage. Those cards were only released as loss leaders,because AMD didn't want to develop a bigger Polaris design. Even the smaller Polaris was probably funded by Apple(they were talking about die thinning used to make it fit in MacBooks). Vega as a design seems to have lasted well - cDNA is an evolution of it.
The big problem here,is AMD is a CPU company which dabbles in GPUs. So a lot of the development of new consumer orientated uarchs is to help their IGPs,laptops and semi-custom partners. This is what is funding their dGPU designs. So this means area optimised and lower power designs. Not the larger high performance dGPUs we all want. A lot of the talk about RDNA2 was about it maximising throughput for consoles. Nvidia,goes for the best performance even if it means larger chips and more power,or having to use the best nodes they can get access to.
Its all about budget and what is the priority. CPUs are the bread and butter of AMD.You can see that with RDNA3. AMD has TSMC 4NM volume. They could have made the RX7900XTX compute die on TSMC 4NM,and most likely would be somewhat faster or more efficient. But instead they think their Zen4 APUs are a more worthy use of that volume. Then going back to RDNA2 - AMD pushed most of its 7NM volume to CPUs and consoles.
With Intel,the big problem is software and a history of not really being good at it. But that will take a lot of money to overtake the decades of investment Nvidia,let alone,AMD have already done.Also the fact they used a lot of Russian engineers as part of the driver development team:
Why is Intel’s GPU program having problems?
What is going on with Intel’s GPU program? The chatter is negative but what is really happening?
It's why they have big problems with older games because they need to do all of it from scratch,as Intel didn't care for decades about games. Another instance of sheer stupidity as they have the most IGPs out there,and it would have cost them relatively little to do. But with all the infighting at Intel the last 5~8 years,instead of sticking to a plan,they seem to assigning blame to each other.
With Intel,the big problem is software and a history of not really being good at it. But that will take a lot of money to overtake the decades of investment Nvidia,let alone,AMD have already done.Also the fact they used a lot of Russian engineers as part of the driver development team:
Why is Intel’s GPU program having problems?
What is going on with Intel’s GPU program? The chatter is negative but what is really happening?
