Why still so much Intel?

[Passing Bye]

CPU market right now is amazing, I don't know what I would even buy first, I have Intel build on the side, I have AMD build on the side, I'm lusting after M1 since it came out, I have decade old desktop Intel 3770 machine that does the job for me, I don't need new machine right now, but I'm ready to pull the trigger on best thing when that happens, keep my options open for anything, follow market and keep myself informed. I used both PC and Mac, I used both AMD and Intel before, I'm CPU agnostic, I don't care who wins, I want best thing I can get for my money when the time comes, I can't go wrong with either, but I would like to save up, I follow gear like that, trends, news, I'm every day in listings, I don't need all that, but one day I will need and I will again buy best thing for my money, I have zero remorse over all my purchases, I bought some amazing stuff and I'm the guy in my circle that my friend turn to if they need new computer or music gear, I hunt best deals for them, I go extra mile in research, that's my thing, I don't follow social media, I follow tech and audio trends, that's my internet hobby and this forum.

Nothing wrong with not following trends, one can't get much wrong these days, there's so much great options, we are living in amazing times for music making, gear is cheap and good, computers are cheap and powerful, even cheapest stuff.

[cron]

Experience. Intel worked and will work, AMD had big trouble processing plugins and DAWs on my end. Some dev explained to me the AMD CPU had to emulate certain instructions and vectorizations.

When was this and what software was it?

Intel is consistently ahead of AMD, when it comes to AVX and so on. That is a very valid point, for audio stuff. How many devs are actually on the bleeding edge, though?

I don’t think this is the case since the Ryzen 3000 series. The 1000 and 2000 series infamously executed AVX2 instructions by treating them as two 128-bit instructions rather than a single 256-bit instruction. There was indeed a heavy penalty on AVX2 as a result.

There really weren’t that many workloads where this translated to a performance disadvantage given how fast Ryzen was otherwise though. Video encoding and scientific computing seemed to be the main beneficiaries. A dev is better placed to confirm this, but these instructions aren’t necessarily a free ride because they’re power hungry and come with a clock speed penalty when used.

The latest Intel chips offer AVX-512 on the mainstream desktop for the first time, but I can only think of 2 audio devs who broadcast that they’ve done some optimisation for this. It really is almost solely hardcore scientific computing that’s seeing big benefits there.

Massive X is the only piece of software I’m aware of that demands AVX (not AVX2) as a minimum requirement. It’s a vanishingly rare demand. You shouldn’t ever be seeing compatibility issues with an AMD processor, or indeed any x86 processor made in the last 10 years when it comes to instruction support or anything else.

[sjm]

You are welcome if that really comes as a surprise. However, it shouldn't really, because that was the obvious premise of my counter-question.

Can't believe it took multiple explanations of something so blatantly obvious.

[sjm]

Nobody is buying something they don't need and whatever they choose will serve them good, both options are amazing, difference isn't in quality, it's mostly in price and availability, don't understand why are you constantly bringing that straw man

I was paraphrasing the counter-argument; the fact that it was absurd was exactly the point.

[mike_the_ranger]

Why still so much Intel? Why not? Don't see a reason to jump up, get rid of my running system and invest in a currently hyped AMD CPU. The differences are also not that impressive that I'd say "OMG, absolute gamechanger". Competition is great and I'm sure Intel will come up with new stuff too. Further more, though not being affected myself, a collegue had indeed also troubles with AMD in combination with music production. Similar for a friend of mine who tried the apple silicon stuff and reverted because it made more trouble than useful stuff.

[Passing Bye]

You are welcome if that really comes as a surprise. However, it shouldn't really, because that was the obvious premise of my counter-question.

Can't believe it took multiple explanations of something so blatantly obvious.

Thanks guys!

[Super Piano Hater 64]

Massive X is the only piece of software I’m aware of that demands AVX (not AVX2) as a minimum requirement. It’s a vanishingly rare demand. You shouldn’t ever be seeing compatibility issues with an AMD processor, or indeed any x86 processor made in the last 10 years when it comes to instruction support or anything else.

It's not far off. SSE4.2, which appeared in the generation before AVX, is now a minimum requirement for the latest versions of u-he plugins (and some other stuff I can't remember off the top of my head). There were a bunch of reports from people with computers from 2008 or so who couldn't use Hive 2.1.

For most lines of work, I agree that it's very weird for someone to be using a computer more than ten years old... Wait a second. No it's not. I've lost count of the (visual) artists I know who just moved on from their Windows 7 installs in the past year, mostly due to failing hardware. Maybe creative work just incentivizes people to never let go of the tools and techniques they already use.

[BONES]

I'm pretty sure Synapse now requires AVX, seem to remember it giving problems to one of the beta testers a year or so back.

Sorry, but that Dell is outperformed by M1 and comparable AMD offering (like ROG Zephyrus G14)

First of all, the M1 powered Air is not faster, simply because the G7 is a far more balanced system, with two SSDs and proper nVidia RTX graphics. I looked at the Zephyrus and couldn't see anything to make me spend an extra Au$1000 on it. And remember, you said "performance per dollar", you weren't talking in absolutes. The G14 would indeed outperform my G7 but not by the 60% more that it would have cost me. Therefore, in terms of "performance per dollar", the G7 comes out well on top. But if I was going to spend Zephyrus money, I'd have spent it on a Zenbook Duo because I'd happily compromise on power for the extra screen real estate. That said, I'm not sure I'd be willing to give up user upgradability for that extra screen. (I recently filled up the system drive on the G7 but for Au$178 I was able to get a newer, faster one with double the capacity that I installed this week. It will be a few more nights before I've got all the software back on it but it's a lot less effort than buying a whole new laptop.

[.jon]

I'm speccing a new PC build for August (moving to a new house and also hoping that the GPU market recovers after the summer), and after reading my eyes sore about the current options, I will go with an AMD Ryzen 5000 series CPU. Ryzen 9 5900x is the most likely choice, it's abotu 250e more than 5600x, but it seems worth the price difference for the number or cores.

As I'm upgrading from an old PC, all my choices are open but I really can't figure out any reason why choose Intel over AMD Zen 3 generation CPU at the moment.

[chk071]

I'm speccing a new PC build for August (moving to a new house and also hoping that the GPU market recovers after the summer), and after reading my eyes sore about the current options, I will go with an AMD Ryzen 5000 series CPU. Ryzen 9 5900x is the most likely choice, it's abotu 250e more than 5600x, but it seems worth the price difference for the number or cores.

As I'm upgrading from an old PC, all my choices are open but I really can't figure out any reason why choose Intel over AMD Zen 3 generation CPU at the moment.

I have my eyes on a new PC as well. Currently targetting this one: https://www.dell.com/de-de/shop/dell-de ... cdx89423co Seems to be good specs for what I want to do (1080p gaming and audio stuff).

Would be great to get it for 200 € less though... wonder if that's realistic though, I don't expect the prices to recover that quick, or, at all. Totally nuts at the moment.

I have all the time in the world though. If people want to buy for such nutty prices, I don't have to.

[.jon]

That Dell seems to come with the non-K version of I7-11700, with lower base clock than the K-version. Just something to keep in mind when looking at benchmarks, there aren't really any out for that specific CPU to see how it performs.

I can't comment on the pricing, never looked into pre-built PCs and VAT base is different for our countries (19 vs 24%).

[chk071]

Unfortunately, it's all at least 200 or 300 € more expensive than usual. Guess they have high demand, and low stock.

Regarding the K/non K CPU: I'm aware of that. Still plenty processing power with the non K though.

[codec_spurt]

yuu

[kingozrecords]

Because there's less telemetry built into the inner workings and because there's no such connections available for hardware to connect to statics and analytical data mines, there's an overall faster speed.

It's funny because even on Linus tech tips they seem to miss that, and instead of running many programs at the same time they only run one to benchmark. That really is an inferior test and would not be conclusive. Put any older intel chip capable of 3.8 ghz and it will always win in multi-tasking because there's less connections and controversial analytics being gathered in the background.

It's really best to rely on everything 8 years old for PC, should be a rule of thumb for anyone trying to multitask so to streamline workflow. If there was a good m.2 NvMe hack for win7 also; I'd say go that route. It would be preferred to be running samsung evo.

It's funny, because benchmarking tools are fixed these days to take into account the amount of overhead in win10, so any online benchmarking tool says you're slower in win7. When in actuality there's less happening in the background and your throttle is unhindered by dependencies. I mean, wow - a Lamborgini Diablo (V12) might beat a Ferrari F40 (V8) uphill but not on the straight away.

Ferrari F40: Regarded as the first 200-mph production car, the F40's 2.9L twin-turbo V8 generated 471 hp and 426 lb-ft of torque. These were the official specs reported by Ferrari, but many experts agreed that this was an underestimate, with claims that the Ferrari F40 easily broke 500 hp.

Lamborgini Diablo F40: The Lamborghini Diablo SV is a 2 door coupé-bodied motor vehicle designed by Marcello Gandini with a mid mounted engine driving through the rear wheels. Its engine is a naturally aspirated petrol, 5.7 litre, double overhead camshaft 60 degree V 12 cylinder with 4 valves per cylinder. It develops 510 bhp (517 PS/380 kW) of power at 7100 rpm, and maximum torque of 580 N·m (428 lb·ft/59.1 kgm) at 5900 rpm. A 5 speed manual transmission transmits the power to the wheels. The Lamborghini Diablo SV weighs a claimed 1576 kg at the kerb. Maximum speed stated is 301 km/h, which is 187 mph..

V8 Vs V12: https://www.caranddriver.com/reviews/co ... ison-test/
With the cars' quirks laid bare, Adam cracks off three fast laps in each. He averages 1:29.4 in the Ferrari, 1:34.8 in the Lamborghini. A difference of more than five seconds per lap. And the F40, he reckons, "has another couple of seconds under its belt, if we fiddle with tire pressures and start using all of the track."

Simpler and more compact is faster, the tests are bugged because they are being nice to Windows and accounting for the amount of telemetry and bloatware. They consider a "tolerance" along with the benchmark.

[whyterabbyt]

Because there's less telemetry built into the inner workings and because there's no such connections available for hardware to connect to statics and analytical data mines, there's an overall faster speed.

It's funny because even on Linus tech tips they seem to miss that, and instead of running many programs at the same time they only run one to benchmark. That really is an inferior test and would not be conclusive. Put any older intel chip capable of 3.8 ghz and it will always win in multi-tasking because there's less connections and controversial analytics being gathered in the background.

It's really best to rely on everything 8 years old for PC, should be a rule of thumb for anyone trying to multitask so to streamline workflow. If there was a good m.2 NvMe hack for win7 also; I'd say go that route. It would be preferred to be running samsung evo.

It's funny, because benchmarking tools are fixed these days to take into account the amount of overhead in win10, so any online benchmarking tool says you're slower in win7. When in actuality there's less happening in the background and your throttle is unhindered by dependencies. I mean, wow - a Lamborgini Diablo (V12) might beat a Ferrari F40 (V8) uphill but not on the straight away.

Ferrari F40: Regarded as the first 200-mph production car, the F40's 2.9L twin-turbo V8 generated 471 hp and 426 lb-ft of torque. These were the official specs reported by Ferrari, but many experts agreed that this was an underestimate, with claims that the Ferrari F40 easily broke 500 hp.

Lamborgini Diablo F40: The Lamborghini Diablo SV is a 2 door coupé-bodied motor vehicle designed by Marcello Gandini with a mid mounted engine driving through the rear wheels. Its engine is a naturally aspirated petrol, 5.7 litre, double overhead camshaft 60 degree V 12 cylinder with 4 valves per cylinder. It develops 510 bhp (517 PS/380 kW) of power at 7100 rpm, and maximum torque of 580 N·m (428 lb·ft/59.1 kgm) at 5900 rpm. A 5 speed manual transmission transmits the power to the wheels. The Lamborghini Diablo SV weighs a claimed 1576 kg at the kerb. Maximum speed stated is 301 km/h, which is 187 mph..

V8 Vs V12:https://www.caranddriver.com/reviews/co ... ison-test/
With the cars' quirks laid bare, Adam cracks off three fast laps in each. He averages 1:29.4 in the Ferrari, 1:34.8 in the Lamborghini. A difference of more than five seconds per lap. And the F40, he reckons, "has another couple of seconds under its belt, if we fiddle with tire pressures and start using all of the track."

Simpler and more compact is faster, the tests are bugged because they are being nice to Windows and accounting for the amount of telemetry and bloatware. They consider a "tolerance" along with the benchmark.