PCPer Podcast 785: New AMD Ryzen 5 & 7, Intel Extended Warranty, Most Popular Steam GPU Discontinued

[@mercuryspark]

Just to let Jeremy know that somebody did get his Maus comic book reference :)

[@ButtonPusher]

Josh you know your wings man! I concur, especially about the blue cheese!

[@CaptainShiny5000]

I think AMD's marketing and sales team dragged down a good product again. Also, props to Josh for delivering an excellent explanation of what the architecture is all about and who the product is for.

[@B16B0SS]

The steam surveys are random samples, so the numbers go up and down depending on who was selected to be asked to submit their hardware information (its opt-in)

[@NickVu1]

Concerning stock price. On one hand there has been a fall in all tech, on the other, usually when a company announces personnel cut, stock prices usually go up. So there are many forces but the results are bad.

[@delatroy]

Gonna be a great time to buy 2nd hand 7800x3d once the 9800x3d launches 😨 7 series wont bottleneck any gpu

[@Eternalduoae]

The thing that's really kinda crappy about Puget's "failure" rates is the way l they define failures. Their BIOS settings may not be good for AMD's chips, or various other factors.

[@POVwithRC]

Puget man would have done well to just not made any sort of statement

[@magottyk]

I don't see the X parts as a high performance label, but rather that they are a higher bin vs the non X parts. We saw in previous generations 65W X parts notably the 5700X and 3700X, both of which had a x800X counterpart. All of which could be PBO'd to similar performance. It became a common opinion why buy an 800X when a 700X SKU could be made to perform as well. People who wanted the highest binned part available would pay the 800X premium, or those who wanted a simple drop in defaulting to the higher all core performance. Just look at how eco mode generated a lot of interest particularly with 12 and 16 core SKUs. Ryzen 7900 with boxed wraith prism is a direct result of that interest. I'm pretty tired of reviewers who only want and demand improved performance at any cost. This release has shown them as one dimensional, their own testing in games with PBO (unshaped nor optimised) to mimick higher TDPs showed no substantial gaming improvements outside of frequency gains. The wider paths aren't being utilised by their benchmarks for the most part, so it's pointless releasing a high performance part as it'll be doubly disappointing as there'd be no efficiency gains to talk about.

[@RobBCactive]

It is one dimensional, but when they're reviewing for gamers they are mostly only interested in performance gains, much less power consumption on desktop (many bought Intel 12th-14th) and then want AMD to give the stuff away. I am not sure where AMD go with the 9700x, are lower binned parts failing on power efficiency, max frequency or features? If the worse bins run at 105W TDP 142W PPT won't they perform better? Is a 9700 that's a bit slower and less efficient not going to undermine the price of the 9700x by being similar in performance? They could come out with a 9800 non-vcache but that confuses the branding they created for 9800x3D.

[@magottyk]

@@RobBCactive Performance gains are a factor of IPC and frequency improvements for a given architecture. Zen 5s redesign with the wider front end won't be automatically leveraged by software and thus not everything will benefit from the IPC improvements that brings, especially on release day reviews. Given the performance difference between the 13900K and 14900K, who'd have thunk it that AMD would also hit the frequency wall given Zen 4s frequencies. You can't drop the 14900Ks power draw by 1/3 and expect it to keep up with the 13900K. If you drop the 7700X to 65W, it'll only lose about 5% (if that) as seen with the 7700. It looks like a deliberate strategy of obfuscation when they released the 7700X at 105W knowing that they were going to release the 9700X as a 65W part. Then they have efficiency on the table for reviewers. The true benefits of the wider front end will take time to realise. I'm still bemused that no gaming review (that I've seen) thought to compare the 9700X with its last gen 65W counterpart because X and price. The point of getting better binned silicon is as much about stability as it is performance gains. Not strictly required, but if you want to run 6400MHz memory 1:1 or have a crack at higher frequency 8000 memory, better binned silicon gives you a better chance of achieving that. I'm doubtful that AMD will release a 105W 9800X given the higher binned chips will go to the 9800X3D, but it's quite feasible for market segmentation that they'll release a 65W 9700 later rather than price cutting the 9700X too much.

[@RobBCactive]

@@magottyk that's partly true, but you need power to achieve high frequency. TSMC N4P is a refinement of N5P giving more density. Software knows very little about how the processort interprets the instruction stream. It's very simple, if more data can get in, more code be processed by the front end and the back end can keep up.. it go faster. Zen5 looks good to me, higher efficiency allows more boosting, if the power budgtet allows it and 11W per core on the 9700x as max socket power is nothing like what the 7700x had on launch and caused ppl to loose their minds. Nevermind Intel used more power and run hotter.

[@magottyk]

@@RobBCactive _" if more data can get in, more code be processed by the front end "_ IF...... Games are still single thread reliant, you can't speed up waits if the software needs an output from a pipeline to proceed. Go faster needs more frequency. Most of the additional power @105W and beyond is utilised boosting all core frequencies with diminishing returns requiring 100% all core workloads to see performance benefits. Curve optimiser/shaper would suggest that more power isn't always needed to get more frequency. It's part of the equation but do you need double the power for 5-10% performance gains when there's more tools in the toolbox to optimise with.

[@RobBCactive]

@@magottyk your statements aren't backed by what those testing found. You're said programs have to account for the front end dispatch to benefit, but CPU design is about eliminating bottle necks and balance. Games have struggled to hit an IPC much above 1, but these days they are mostly LT not ST except in old ones. Any game benefitting from V-cache is having memory stalls eliminated, so just isn't code expected to benefit from wider dispatch. The fact is the 65W/88W 9700x has been panned, but those who chose to push the chips obtained much bigger uplifts on 7700x and even matching at 60% of the power of a previous gen shows a large architectural uplift.

[@VicariousAdventurer]

Karl Ein

[@Mr11ESSE111]

zen5% not even 5%

[@gabonesmith4750]

what a bunch of grown arse nerds