GPUs & CPUs & Enthusiast hardware: Questions, Discussion and fanboy slap-fights - Nvidia & AMD & Intel - Separe but Equal. Intel rides in the back of the bus.

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.

SPR HBM, which is what I think he is holding up, was ultimately unable to actually ingest data at the rated 1 TB/s. It was still a lot faster than a regular SPR, but for the applications that could actually use the bandwidth, Genoa-X CPUs were as good or better. Plus a lot of applications that theoretically should be able to benefit from that kind of bandwidth don't use AVX instructions at all (the reasons are fucking stupid), so didn't benefit from the HBM anyway.

Here's a very good rundown of the problems:

TL;DR - the cores and the internal mesh couldn't sustain 1 TB/s.

A lot of these lessons were incorporated into Granite Rapids, which can use MCR-DIMMs to sustain very high bandwidth. Overall, Granite Rapids has been a very good CPU, although it is overall a wash with the latest EPYCs, application-dependent as to which is better.
 
  • Thunk-Provoking
Reactions: Vecr

From the pinned comment from an hour ago, they aren't interested in deleting the video, but I tossed it into Preservetube anyway:
Мы предлагаем вам не доверять нам или кому-либо еще только потому, что у них большая аудитория и красивые графики. Не верьте нам, но верьте тому, что мы показываем: подробные настройки, игровой процесс и подробный мониторинг. У нас такой подход к тестированию. Максимальная прозрачность. Вы можете свободно задавать вопросы и воспроизводить наши тесты. Это легко и очевидно.

После просмотра видео Hardware Unboxed некоторые люди, которые его смотрели, в конечном итоге пришли к стойкому мнению, что видеокарту ARC B580 не следует покупать ни при каких обстоятельствах, если у вас старый процессор, даже если автор не сказал или не имел этого в виду явно. У нас были вопросы, и мы попросили Hardware Unboxed показать, где и как они тестируют свои игры. Мы считаем приемлемым, когда блогеры перепроверяют результаты для себя или других блогеров. В конце концов, наша работа — искать правду и информировать людей, а не быть слишком милыми или нравиться всем.

Кроме того, мы никогда не утверждали, что у Hardware Unboxed поддельные тесты. На самом деле, мы часто ссылались на их тесты до этой ситуации, когда мы не могли воспроизвести их данные, как не смог сделать это наш коллега. Мы спрашиваем в заголовке, почему результат отличается, но мы не обвиняем, это не оскорбление. Мы пытаемся расследовать не только их тесты по результатам другого канала «Мой компьютер», который Стив также упомянул в своем видео. Так что если вы считаете это обвинением, то большой вопрос в том, кого обвиняют — HU или их критиков? Здесь надо сделать поправку, что у нас немного разная культура общения, она отличается от англоязычной, что мы не всегда учитываем, так как заходная картинка и заголовок создавались для русскоязычной аудитории с другой культурой общения, но возможно на английском языке при прямом переводе это звучало не слишком хорошо.

Не зная, где именно команда HU проводила тестирование, мы попытались воспроизвести результаты в местах, показанных на видео. Мы не знали, что место в левом верхнем углу было точным местом тестирования, а не случайным местом игрового процесса. Если это указано где-то в методологии тестирования HU или написано в видео, пожалуйста, укажите, где именно. С этого момента, основываясь на том, что говорит Стив, мы будем предполагать, что каждое тестирование HU проводится в показанном месте. Однако в некоторых видео, как в этом видео, HU не показывают никакого геймплея, поэтому мы не можем догадаться, где они проводили свои тесты https://www.youtube.com/watch?v=00GmwHIJuJY, но в других видео это показано, как здесь https://www.youtube.com/watch?v=CYOj-r_-3mA

Теперь у нас достаточно данных, чтобы двигаться дальше и перепроверить данные Стива. Если мы ошиблись, как было сказано в первом видео, мы извинимся. Однако говорить, что наши тесты «плохие», «мы не знаем, что тестировать или как тестировать» и предлагать нам «удалить наше первоначальное видео» до того, как все перепроверить, было довольно непрофессионально, по нашему мнению. По крайней мере, мы сказали в видео, что МЫ МОЖЕМ ОШИБАТЬСЯ, ПОЭТОМУ НАМ НУЖНО БОЛЬШЕ ДАННЫХ, мы не говорили, что мы правы, а HU не правы. И Стив, похоже, уверен, что он прав. Может быть, так и есть. Или, может быть... В любом случае, мы узнаем. Наши тесты не плохие. По крайней мере, если «плохо» не в том, что наши тесты отличаются от тестов HU из-за разных локаций и разного подхода, что не делает их «плохими». Мы уверены, что наши тесты верны в определенных локациях, пока не доказано обратное, а прозрачность нашего подхода позволяет всем перепроверить наши данные, поэтому, если мы сделаем ошибку, это сразу будет видно в обсуждениях. Когда блогер показывает только графики и предоставляет ограниченные настройки без игрового процесса и мониторинга, воспроизвести его тесты становится сложно. В других локациях результаты могут быть другими.
 
Stop dropping them on the floor I guess? Seriously. Unless you guys are buying the cheapest shit possible I really don't know what you guys are doing to these things.
"Gaming laptops" tend to run extremely hot, which isn't exactly great for their structural integrity. I know someone who melted the pads on the bottom of his.

What I want to know is if it's a generic CPU overhead issue, or if it's an issue specific to Ryzen 5000 generation CPUs and their PCIe bus implementation.
It's not a PCIe bus issue, it's specific to the rBAR feature. Arc GPUs are known to have issues with Resizable BAR on Ryzen 5000s, and without Resizable BAR, they're essentially unusable. igorslab mentioned this in their initial review (e: of older Arcs, but the problem is the same):

The fact is that the Arc A380 does not harmonize equally well with all systems and on the Ryzen systems with the Ryzen 7 5800 X3D, for example, Resizeable Bar (rBAR) also does not work cleanly. The result was a glaring performance crash, even in applications that don’t actually need rBAR officially. A rogue… But let that be. AV1 support for YouTube didn’t really run on the AMD board either, even if AV1 is unique as an encoder and decoder in this performance class. In addition, some games like World War Z do not run or crash with Vulkan. Unfortunately, this also includes standard software such as Audodesk Maya 2017 or Catia, where partial refusal to work occurred. But as I said, of course I do Intel the favor and benchmark everything with the rBAR feature enabled.
 
It's not a PCIe bus issue, it's specific to the rBAR feature. Arc GPUs are known to have issues with Resizable BAR on Ryzen 5000s, and without Resizable BAR, they're essentially unusable. igorslab mentioned this in their initial review (e: of older Arcs, but the problem is the same):
The driver overhead problem should be a separate issue from ReBAR, since the 5600 has ReBAR for example. Intel's GPUs have more than one glaring flaw.
 
It's not a PCIe bus issue, it's specific to the rBAR feature. Arc GPUs are known to have issues with Resizable BAR on Ryzen 5000s, and without Resizable BAR, they're essentially unusable. igorslab mentioned this in their initial review (e: of older Arcs, but the problem is the same):

Resizable BAR is part of the PCIe interface. Maybe "bus" was the wrong term to use, maybe I should have said "controller." What matters is whether it's an implementation-specific issue with Ryzen 5000 or a generic issue. I managed to dig up some other performance issues with AMD's PCIe implementation a while back but it was nothing conclusive, other than to suggest that AMD in general has some issues. However, there is more evidence this is an AMD-specific issue.

The nearest recent Intel CPU I could find to the Ryzen 5600 is an i5-12400F, which is 6 P-Cores and no E-Cores. CPU benchmarks come out pretty close. However, while Spider-Man 2 at 1080p High can only sustain around 60 fps on the Ryzen 5600, it sustains an average around 100 fps on 1080p Very High using an i5-12400F:


It gets 128 fps when using XeSS, so it's really not CPU-choked at all. This suggests it's not a generic CPU overhead issue, but an issue specific to Zen 3's PCIe interface and reBAR implementation.
 
Last edited:
The driver overhead problem should be a separate issue from ReBAR, since the 5600 has ReBAR for example. Intel's GPUs have more than one glaring flaw.
Chips of that generation have ReBAR support on paper, but the implementation was very bad, and it was notorious for performance issues or not working outright on certain motherboards or BIOS versions.

As far as the original review goes, good luck reproducing AMD benchmarks off information in videos. Even for RAM performance, you'll see significant differences unless you control for the exact AGESA version you tested with.

I managed to dig up some other performance issues with AMD's PCIe implementation a while back but it was nothing conclusive, other than to suggest that AMD in general has some issues. However, there is more evidence this is an AMD-specific issue.
There's no shortage of AMD issues with PCIe, but the only way I know to directly cause noticeable PCIe overhead is to set FCLK too high on a silicon lottery loser. (It's worth noting that some motherboards will raise FCLK by 100MHz on default settings now, so it's possible you have this issue out-of-the-box if you're very unlucky.)

There is an issue I know about where a lot of X870 boards have problems with lanes being automatically downgraded due to signal quality issues. It's a fairly common problem with 5080s and M.2 drives on MSI AM5 boards, which is almost immediately noticeable since 5080s don't run well under PCIe 1.
 
I recall him having a bit of his own drama where he got overly defensive and avoidant when confronted with criticism but I didn't do any bigger research into what it was about. He is however a bit of an oddball. Technically yes, he has the know-how and is the loudmouth that's needed to tell it how it is, but at the same time his company, Threat Interactive, hasn't made anything worth of note, and he's essentially the only employee there, while constantly saying "we at Threat Interactive".

So it's a bit hard to take him seriously when he's constantly authoritative about his expertise while having nothing to show for it and exaggerating the size of his one man operation to do so.
Very late Seems like he will send a DMCA to anyone that disagrees with him.
 
I don't trust gaming laptops. Every single one I've seen has been on the verge of falling apart within a year.
Even the Lenovo Legions or ThinkPad Extremes? I've seen many many gaming laptops that always end up breaking some way but Lenovo seems to be a little bit better on their premium brands.
 
Stop dropping them on the floor I guess? Seriously. Unless you guys are buying the cheapest shit possible I really don't know what you guys are doing to these things.
I've dropped my macbooks on the floor several times and they're all still fine. A little bit dented, but everything works perfectly well. Maybe Windows laptops are just low build quality in general because the only thing they can compete on is price?
 
All I can say is the Acer Predator Helios has been through 2 toddlers for over 2 years and constant trips to work, slipping off the passenger car seat onto the floor, being sat down on countertops a tad too hard, kids opening and closing the screen constantly or dropping it on the floor.

Still like new. Again, I really don't know what you guys get up to or what cheap shit you buy. Maybe the 18" ones are just built more sturdy for the extra weight.
 
I've dropped my macbooks on the floor several times and they're all still fine. A little bit dented, but everything works perfectly well. Maybe Windows laptops are just low build quality in general because the only thing they can compete on is price?

As mentioned in the other thread, we've have Macbooks go for years in backpacks, suffering all the abuse implied by that, and be perfectly fine. The ASUS Tuf is 3 years old, and the bottom half of the clamshell splits open now because there are internal screw posts that have snapped off, and other screws that have fallen out. What apiece of shit.

Acer Predator Helios

Well, maybe I'll give that a try once I'm done with the ASUS. Either that or a Thinkpad. Really sick of dogshit laptops.
 
  • Horrifying
Reactions: Vecr and Susanna
As mentioned in the other thread, we've have Macbooks go for years in backpacks, suffering all the abuse implied by that, and be perfectly fine. The ASUS Tuf is 3 years old, and the bottom half of the clamshell splits open now because there are internal screw posts that have snapped off, and other screws that have fallen out. What apiece of shit.



Well, maybe I'll give that a try once I'm done with the ASUS. Either that or a Thinkpad. Really sick of dogshit laptops.
I cannot vouch for smaller ones. I think the cheaper models are plastic.
 
All I can say is the Acer Predator Helios has been through 2 toddlers for over 2 years and constant trips to work, slipping off the passenger car seat onto the floor, being sat down on countertops a tad too hard, kids opening and closing the screen constantly or dropping it on the floor.

Still like new. Again, I really don't know what you guys get up to or what cheap shit you buy. Maybe the 18" ones are just built more sturdy for the extra weight.
Not all Acers are equal. Coming from an Acer Aspire 5, plasticky as shit and extremely fragile.
 
Even the Lenovo Legions or ThinkPad Extremes? I've seen many many gaming laptops that always end up breaking some way but Lenovo seems to be a little bit better on their premium brands.
I'm using a Lenovo Legion 5 Pro 16ACH6H with the 32 GB RAM and 3070 config. I've had it for about 3.5 years and it is perfectly fine, lots of usual wear and tear but never had any actual hardware issues with it. I've traveled to Europe and the US with it, taken it to some very far flung places and on lots of road trips, plane flights, etc.

That being said I rarely have it off an AC adapter so the battery hasn't been cycled much, and it's still holding up okay in spite of age. I've also basically always had it on dedicated GPU mode instead of hybrid with the mux as this gave better compatibility with Linux. The RAM is original but I've swapped the WiFi card for an Intel one for better Linux support and expanded the storage.

Repasted it recently and I suggest anyone undertaking this be very careful with the CPU/GPU fan connectors. They're unbelievably fragile and in the end I don't think it's worth even unplugging them as you can awkwardly repaste the whole lot while they're still plugged in. Repasting it has helped a little, seems the fan doesn't have to run as hard as it used to when idling.

I'd still like a newer laptop but there really hasn't been anything compelling. The newer AMD chips aren't enough of a leap forward, NVIDIA mobile chips are a joke and the 2560x1600 165 Hz display is still perfectly fine.
 
I’ve recently been testing my phone out with steam link and I’m pretty convinced now that it’s superior to owning a a steam deck or rog ally. Just get one of those backbone controllers and you’ll have a superior screen to any hand held as well as good battery life due to cloud rendering. As long as you have a decent home internet connection and your client side internet isn’t shit it’s been pretty impressive imo. It should be even better once I set up Apollo and Artemis based on other people’s experiences as well.
 
Back