I said a manycore CPU. It would have hundred of in-order cores optimized for throughput or, in its defect, out-of-order cores with big vector/SIMD units.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Sony Working on AMD Ryzen LLVM Compiler Improvements, Possibly for PlayStation 5
- Thread starter Megalith
- Start date
I said a manycore CPU. It would have hundred of in-order cores optimized for throughput or, in its defect, out-of-order cores with big vector/SIMD units.
Red Falcon
[H]ard DCOTM December 2023
- Joined
- May 7, 2007
- Messages
- 12,457
Don't ask me to "stop replying" - you don't own this forum nor this thread, nor are you Kyle or a mod.
I do understand what you are saying, but software has to be driven by the CPU, otherwise, how is the software processed?
Software (the game engine) isn't run on nothing, and when the CPU runs out of processing power and/or the program is unoptimized, it will lack the ability to properly keep data going to the GPU.
I explained this in great detail above, and even gave you a video example of a Pentium 4 unable to keep data going to a GTX 970 GPU properly, thus resulting in overly low frame rates.
I am not saying that the current-gen consoles are incapable of 1080p/4K @ 60fps, but I am saying that certain games require too much processing power for the low-power Jaguar CPUs to keep up with properly, despite being completely optimized, such as my example of the PS4 exclusive Bloodborne.
So hardware has no limitations?
If that's the case, then why are those games being played on a Pentium 4 and GTX 970 running at 5fps and not 60fps+, which the GTX 970 itself is more than capable of with those titles?
(I'll give you a hint: the Pentium 4 is limited with what it can process, thus the hardware itself is limited)
I'm not saying software can't dictate what hardware does what, but the software is driven and processed by the CPU, and when the CPU lacks the processing power necessary for the game engine to run properly and keep data going to the GPU, then the software can't be processed due to the CPU bottlenecking the system, aka, being CPU bound.
Yeah, I get it, now go watch a video of Bloodborne on that exact same hardware running at 1080p @ 30fps... what's your point?
Software is driven and processed by the CPU, aka, the hardware-level:
I don't see the game engine (software/application) at the bottom of that diagram, yet that is exactly what you are incorrectly touting.
The CPU is part of the hardware-level - this is mind blowing that you don't get this.
Your comments and repetitive questions, which have been explained in detail multiple times now, are becoming borderline troll level.
If you still don't get it, I would suggest doing some thread searches or doing a bit more research before touting further incorrect information, thanks.
Linus changes of opinion very fast
https://thenewstack.io/linus-torvalds-meltdown-spectre-perhaps-move-arm/
The x86 shill even starts praising the ARM servers now with a weak "ThunderX2 seems quite reasonable" after last benches show how the ARM CPU has IPC close to Skylake and beats the best Xeon CPU in several server/HPC benches.
Moreover regarding consoles I would trust more John Carmack than Linus Torvalds and Carmack was very receptive to the idea of an ARM console. He even gives hints to why Sony and Microsoft went the x86 route instead.
I would like to see PlayStation 5 being ARM-based, but I suspect it will be Zen-based.
Nobu
[H]F Junkie
- Joined
- Jun 7, 2007
- Messages
- 10,089
I take a lot of what Linus T. says (what we read in blogs) as much sarcasm/hyperbole with some sound reasoning behind it, especially since it's often his reactionary comments that are publicized. Usually I'll read further replies to find his reasoning, or if none I'll research more on my own.
I would welcome an arm ps5, except i can't even play most of my ps3 games on my ps4, nvm the older generations. And then the gpu becomes another question... We know nvidia can make a good arm+gpu chip (shit, AMD could too), we just don't know whether the combination would be enough of an improvement to warrant all of the costs (nvm the hardware...porting the os, game compatibility if wanted, writing new drivers. I'm sure there are more that would be unnecessary or much less involved if they stuck with x86). Once nvidia releases anything we can use with Xavier we might have a better idea, though I'd bet on sony using something based on the next iteration of tegra if they went that route (no earlier than 2h 2019).
I would welcome an arm ps5, except i can't even play most of my ps3 games on my ps4, nvm the older generations. And then the gpu becomes another question... We know nvidia can make a good arm+gpu chip (shit, AMD could too), we just don't know whether the combination would be enough of an improvement to warrant all of the costs (nvm the hardware...porting the os, game compatibility if wanted, writing new drivers. I'm sure there are more that would be unnecessary or much less involved if they stuck with x86). Once nvidia releases anything we can use with Xavier we might have a better idea, though I'd bet on sony using something based on the next iteration of tegra if they went that route (no earlier than 2h 2019).
Red Falcon
[H]ard DCOTM December 2023
- Joined
- May 7, 2007
- Messages
- 12,457
The speculation continues...
Pieter3dnow
Supreme [H]ardness
- Joined
- Jul 29, 2009
- Messages
- 6,784
You ignore my statements each and every time. And cling to your own.
You claim 30 fps I show you Battlefield 4 then you blab about Bloodborne. The hardware design is used as what you can do with Mantle you have cpu only to send data to the GPU the function in the design is to feed the beastly gigaflop gpu.
https://en.wikipedia.org/wiki/PlayStation_4_technical_specifications#Graphics_processing_unit
The processing power of the cpu is minimal because in this design it just functions as I told you before. You claimed multiple times that the hardware limitation in the design is the CPU where I will say that it is nonsense. You fail to show that the engine used in Destiny 2 is related to the cpu performance all you get is a statement from a developer that it is limited on all consoles to 30 fps.
Which means that it is engine related rather then hardware because if it was cpu it should be faster on a xbox 1 x or even optimized to use the faster cpu.
However you want to put this the design philosophy will not change because of Zen cores because the GPU is the star, people whom are saying that it will be much faster are wrong because the design is purely to feed the GPU, it means that some of the other tasks may end up running on the cpu but the cpu in this design will never amount to processing power of the gpu hence the argument that the increase from using Zen cores vs Jaguar cores is more then likely to end up as minimal.
You and many others fail to grasp how this and previous generation consoles are designed.
This again is engine design nothing else. If you want prove of this run Diablo 3 on any pc with a Jaguar core cpu it won't run. Console version does guess what changed?
Not if you used a model like Mantle/DX12/Vulkan where you would send data to the gpu and remember the early Mantle demo by Oxide games where it used some metrics to prove that scaling on multiple cores even when used slower cores (Bulldozer) that was unmatched by anything at the time. The performance metrics were far beyond what is possible with DX11 and that is how on a hardware level the consoles function. And at the same time it explains what is engine bound and what is hardware bound.
In a design where each core can talk dependently to the gpu does it matter how fast there going?
Then why would the design feature Jaguar cores to begin with that makes no sense.
Central processing units
The central processing unit (CPU) consists of two x86-64 quad-core modules for a total of eight cores,[39] which are based on the Jaguar CPU architecture from AMD.[24] Each core has 32 kB L1 instruction and data caches, with one shared 2 MB L2 cache per four core module.[40] The CPU's base clock speed is said to be 1.6 GHz. That produces a theoretical peak performance of 102.4 SP GFLOPS.
You can multiply any number you want but in a configuration where you still have the gpu doing the work the amount of processing power from the cpu will never reach the scales where it will be important then any other job to send data per core to the gpu and not use a lot of power.
Red Falcon
[H]ard DCOTM December 2023
- Joined
- May 7, 2007
- Messages
- 12,457
That right there says how little you know about anything.
At least get your nomenclature correct.
Um, have you actually looked a the system requirements for Diablo 3?
It's a game from 2012 (I played the original demo circa 2008 or 2009) and it will most certainly run on a system with an AMD Jaguar quad-core.
Here are the Diablo 3 system requirements:
- CPU: Intel Pentium D 2.8 GHz or AMD Athlon 64 X2 4400+
- RAM: 1 GB RAM (1.5 GB required for Windows Vista/Windows 7 users)
- OS: Windows XP/Windows Vista/Windows 7/Windows 8 (Latest Service Packs)
- VIDEO CARD: NVIDIA GeForce 7800 GT or ATI Radeon X1950 Pro or better
- FREE DISK SPACE: 12 GB
Those system requirements are basically for an x86-64 computer from late 2005 or early 2006 - how the hell would a Jaguar from 2013 not be able to match that???
This coming from me, a person who owns both a thin client (see sig) and PS4 Pro with Jaguar CPUs, and I know exactly what these CPUs are capable of.
Can't really say the same for you.
Yes, it does matter.
Because, as I have already explained to you in great detail, the Jaguar CPU was the most cost-effective solution at the time when this generation of consoles were being developed and released by Sony and Microsoft.
It is a low-power CPU/APU/SoC from 2013 designed for thin clients, embedded systems, and power-efficient systems - hardly performance computing, but it got the job done (barely) with the consoles at the time.
The only thing you have proven is that you know jack about anything on this subject matter and are touting total bullshit at this point to try and prove a point which logically makes no sense.
At least I have backed up my statements with proven facts, development history, statements from the devs themselves (which you flat out denied, insanely enough), and video evidence - I can't say the same for any of your "claims" at this point, other than that one gameplay video that doesn't really mean anything relating to this discussion.
Believe whatever you like - it doesn't make it true.
Last edited:
Not to wade into this epic forum fight... but....
Windows and Direct X. lol
Seriously the PS4 is a lean mean BSD running GNMXing machine.
I know this is very old now... but stories about PS4 developers aren't all that exciting anymore so you don't read to much new material.
https://www.eurogamer.net/articles/...its-really-like-to-make-a-multi-platform-game
"
Digital Foundry: DirectX 11 vs GNMX vs GNM - what's your take on the strengths and weakness of the APIs available to developers with Xbox One and PlayStation 4? Closer to launch there were some complaints about XO driver performance and CPU overhead on GNMX.
Oles Shishkovstov: Let's put it that way - we have seen scenarios where a single CPU core was fully loaded just by issuing draw-calls on Xbox One (and that's surely on the 'mono' driver with several fast-path calls utilised). Then, the same scenario on PS4, it was actually difficult to find those draw-calls in the profile graphs, because they are using almost no time and are barely visible as a result.
In general - I don't really get why they choose DX11 as a starting point for the console. It's a console! Why care about some legacy stuff at all? On PS4, most GPU commands are just a few DWORDs written into the command buffer, let's say just a few CPU clock cycles. On Xbox One it easily could be one million times slower because of all the bookkeeping the API does. "
He goes on to say MS was improving things... and I am sure they have for sure. I mean sounds like he caught himself bad mouthing a partner and back tracked.
Point is however. PCS have a ton of overhead that consoles just don't have.
Still as a great as a GPU with a Bagillion gillion zeta flops is... they are only as fast as the CPU feeding them the data to crunch. Also its not like a CPU doesn't have other things to do. GPUs are doing nothing much beyond pushing pixels and perhaps a bit of physics calculations. The more things an engine asks the CPU to do the less bandwidth it has to send data to the GPU to crunch. GPUs are not any good for NPC calculations, sound, networking, input control, prefetching preloading... and I'm sure tons of other stuff I'm not thinking about. The CPU is pinned with tons of queries and has to schedule how to deal with them all and still feed that GPU something to compute.
Ehhh screw x86 and ARM and Power... give us
RiscV
Seriously though Risc-V is going to start hitting storage and the like shortly... I can see a future where the idea of processor licences dies and even more companies build custom chips. (I know ARM is willing to licence to anyone) still free tech is an attractive offer. Western Digital has stated they plan to ship a billions of Risc-V chips in their drives. Nvidia is planning ot use RiscV inside all its GPUs in the future as well.
As soon as ARM takes over I'm going to start crystal balling the riscv take over. (perahps /s)
RISC-V is my favorite ISA. Second favorite is ARMv8.
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
I mean, I'm fine with that; main issues are desktop software, the compatibility of which is becoming less and less of an issue.
And of course these consoles where maintaining backward compatibility is a big advantage, there is very little reason for MS and Sony to stray from AMD's effective if not absolutely class-leading technology.
Red Falcon
[H]ard DCOTM December 2023
- Joined
- May 7, 2007
- Messages
- 12,457
DukenukemX
Supreme [H]ardness
- Joined
- Jan 30, 2005
- Messages
- 7,970
I agree, because ARM is inevitable. But right now nobody has an ARM CPU that can compete with X86 just yet. AMD and Nvidia will likely be the first to do so as both companies have worked with ARM. The big drive for ARM will be that its cheap and lots of competition. As much as people harp on Apple and Tegra chips, there's lots of good enouhg ARM chips like Allwinner, Rockchip, and Amiga. The way that ARM licenses their designs makes it so anyone could build an ARM CPU. Intel's x86 licensing will be the downfall of ARM.
But yea, PS5 will certainly be x86, just cause right now it makes so much sense. Nvidia doesn't have a product right now that would replace a Ryzen+Vega setup, though I wouldn't be shocked if they're working on one right now.
DukenukemX
Supreme [H]ardness
- Joined
- Jan 30, 2005
- Messages
- 7,970
Something else to think about is that if the PS5 does have a Ryzen 8 core CPU, then you have a lot of CPU power that could be used for Ray-Tracing. If Nvidia were to make specialized hardware to accelerate Ray-Tracing, that hardware would be only good for Ray-Tracing, where as AMD's GPU (assuming no RTU accelerator) would have the entire GPU dedicated to traditional rasterization. Cause either way no hardware can do 100% Ray-Tracing in games.
Nvidia might be playing the same game as they did with PhysX where they forced the industry to use their API when they threw money at companies.
I demonstrated you one ARM CPU that is beating both Intel and AMD. The 32-core CN9980 is ~15% faster than 32-core EPYC 7601 on SPECint. Using AT recent review I have made some computations and I got that the TX2 core gives 99% of the single thread IPC of the Skylake-X core, but beats it on multithread, because the ARM core can execute up to four threads simultaneously whereas Skylake core only two threads.
I will stop here, there are a thread devoted specifically to ARM vs x86.
Last edited:
Its possible in the next few years that we will see some mainstream ARM chips that put the boots to x86 from companies most people wouldn't expect. Marvell the network card, the chips in the chromecasts, and embedded printer chip guys... have bought Cavium and their Thunder chips. No doubt the x2 is going to compete very well against the xeons. Those 54 core beasts aren't going into a PC anytime soon, still its interesting how many patents Marvell and Cavium own between them that could be used in some interesting consumer devices down the road. Just in purchases they have networking, wireless, video compression, embedded stuff... what would be missing would be gaming GPU tech, but people forget that ARM also licences that. I know a mali chip is no geforce or radeon. Still a year or two out what ever follows the G72 might be more then fast enough for gaming chromebooks and the mid range laptop market.
If Apple really does go ARM for their macbooks, and they have a product good enough to sprinkle some Apple marketing on. I could see demand for a company like Marvell to pop out a Thunderx2 12-16 core ARM + Mali graphics for the mass market.
Actually Nvidia does not lead in that regard. The industry is no longer locking themselves into DirectX crap.
Sonys Playstations have always used their own API. GNM... and now GNMX.
The thing is GNM is actually built off OpenGL. GNMX has more in common with Vulcan and Mantel then DX.
AMD does real time raytracing through vulcan about as well as NV does... they simply use open APIs, and aim their marketing at developers not end users.
I'm sure if they wanted to liquid cool 4 super binned Vegas they could build some slick looking marketing crap to convince end users their next card in the pike was some ray tracing beast as well.
Last edited:
DukenukemX
Supreme [H]ardness
- Joined
- Jan 30, 2005
- Messages
- 7,970
My point is that nobody made an end user ARM CPU with a motherboard you can actually use. I think AMD said the AM4 socket would be compatible with their ARM CPUs, but I haven't heard anything for a few years about this. Nvidia would need to make a motherboard chipset, with a motherboard, that you can socket an Nvidia ARM CPU. Not that you need these things for PS5, but you would need to showcase the product you have before expecting anyone to buy it.
This is going to be sad but true... when ARM starts taking over the PC space. We won't be buying them in socket form. If your shipping a SOC there is no need for "chipsets"... PCI memory controllers are all on the chip. I would expect arm stuff will continue to ship soldered on for consumer devices. I believe Qualcomm and Cavium both offer server chips in a socket... they are still SOC designs though. We'll see high performance ARM is still pretty early days but I think a bunch of those companies are learning a lot pretty quickly.
As for Sony needing other things to be ARM first... I am not sure that matters at all. No one using a switch is really all that concerned what chip is inside it. Having said that I still think the chances are good Apple starts marketing ARM powered Macs sooner rather then later.
Red Falcon
[H]ard DCOTM December 2023
- Joined
- May 7, 2007
- Messages
- 12,457
Don't forget about the NVIDIA Jetson TX1 and TX2!
https://www.newegg.com/Product/Product.aspx?item=N82E16813190007
Granted, they are development kits, and they definitely aren't cheap, but at least they are a mini-ITX board instead of a SBC, and can actually use a single PCI-E expansion card.
I totally get what you are saying though, and I've been waiting for just what you are describing for years now.
Pieter3dnow
Supreme [H]ardness
- Joined
- Jul 29, 2009
- Messages
- 6,784
You prove that what you say is not true everyone that ran Diablo 3 can tell you it won't run on any Jaguar core based solution certainly not 1080p and unless you do very low settings hardly at 720p.
You also fail to understand that in your chain of though it is cpu bound guess what game is CPU bound Diablo 3 it will run like crap on Bulldozer and even Thuban it does not like anything else then raw single thread power.
You list theoretical specs but you never have an idea what is going on in the real world. Like your arguments there is something wrong with me and when I prove 30 fps is BS and show you a video on both older generation hardware you still stick with your own pie in the sky argument which Battlefield 4 proves you are wrong about on both consoles.
Jaguar was never a cost effective solution it was the only solution that did not use a lot of power and could feed the gpu as described here by Dice
https://www.tomshardware.com/reviews/johan-andersson-battlefield-4-interview,3688-5.html
That also means that the hardware solution by both Sony and MS was never about money but about how to feed the GPU and not run into a blazing mess that requires a lot of power and there by cooling of their solution.
Your assessment on money vs mine on performance will hold true, because they did the same thing with PS4 PRO and Xbox 1 X. You don't make the same mistake twice. The hardware sales by both companies were pretty good so why would it be about money ? It's about technical performance of the APU combo.
You can make a joke where I prove that something as Diablo 3 is inherently so single threaded that it will run like crap to prove my point about engines being the only way to force a game to be cpu or gpu limited. Hence console version uses a different engine.
So you fail to prove that you can get by hardware limitation by correctly using the hardware as intended or by stuck in the same cpu will be the main benefactor towards improving performance and it is not.
So why don't both of you guys get a job at Microsoft and Sony explain to them how wrong they are using the hardware solution they been choosing for the past years. I'm very sure they will be in awe of your astute knowledge.
Stop ignoring what I'm saying watch the Battlefield 4 video a couple of times then realize that developers will use the power of the console or just badly suck at it and port something that is close to another platform which does not require a lot of work.
DukenukemX
Supreme [H]ardness
- Joined
- Jan 30, 2005
- Messages
- 7,970
I doubt many consumers would want to buy a board with a soldered chip. Even if it's a SoC, the consumer wants a socketable chip. At least the initial customers who are dealing with this type of hardware.
Also look at the Switch with it's Tegra chip. That chip has been available for customers in other hardware. People already have an idea of how it works. But if you wanna compete with x86, you need a motherboard.
ccityinstaller
Supreme [H]ardness
- Joined
- Feb 23, 2007
- Messages
- 4,236
I think you are drastically overestimating what the average consumer cares about. The 1% of us that care pale in comparison to the tens of millions that do not care about a socketable chip. They simply care that they can turn it on, do their work, then spend the rest of their time on social media. The 2 year phone upgrade cycle driven into everyone's brain over the last 20 years will ensure that will come to pass. If you system is 1 second slower, then you buy the newest model every few years.
I would love to see a nice 8+core ARMV8 CPU to play around with today. That Cavium X2, with 8-12 cores clocked at 4+Ghz, would be a beast!
Nobu
[H]F Junkie
- Joined
- Jun 7, 2007
- Messages
- 10,089
Rather, they might care at some point, but it probably wouldn't stop them from buying a PC. If anything, they might be mildly irritated when they ask their friend to upgrade their PC and he tells them they can't.
DukenukemX
Supreme [H]ardness
- Joined
- Jan 30, 2005
- Messages
- 7,970
Thing is, Nvidia already tried this, and failed. Nobody else was adopting the Tegra chips at the volume that Nvidia had hoped, so they had to make the Shield devices that kinda showcase Tegra's abilities. Last I checked these devices didn't sell well, and they were geared for people who wanted to plug and play.
The smart thing to do would be to create a workstation motherboard with upgradable hardware for the 1% of us that would use it. Gear it for the nerds that run linux and can make use of the ARM hardware immediately, cause Linux already works well on ARM. Then some nerds port applications over to it, like QEMU which would allow to run x86 applications on it. Make it cheap, cause if it isn't then nobody would bat an eye at this hardware.
If the hardware was $200 total that performs like a Ryzen and uses less power, then people will start to pick up on it. This is basically what Raspberry Pi did and it owes its success to this. Though the RPi is a mess when it comes to working 3D accelerator, which Nvidia would likely have perfect working 3D hardware. But nobody is going to build a $300+ PC with a CPU that can't be upgraded, that also has a none existing software library.
Once the vocal minority priases the awesome that is ARM, it'll pick up. More software would get made and next thing you know is that the PS6 or Xbox Thrice will be using ARM. Assuming there's going to be another generation of hardware. More than likely by this time we'll have a lot of mini PCs that run on ARM.
Well the shield didn't not sell because it couldn't be upgraded/ Average users really really don;t care if their CPU is socketed or not.
We care sure... regular users nope they don't care.
Does anyone here actually know a REGULAR user that has ever upgraded a CPU. I mean come on no you don't
Put a better way how many of us [H] type users have honestly upgraded a CPU in the last 20 years. Intel releases a new board and often new socket with each new generation. AMD has been a bit better in that regard still unless your buying a new processor every year your going to need a new board anyway. I mean really when in the last ever has it been an option to just change a chip in a machine that's 3 years old. I am sure if we go through the history of the desktop we can find a few cases where if you bought the top of the line board the day that chipset launched perhaps a 3 year CPU only upgrade was an option... I don't believe it really works that way in practice. 3 years in your going to need a new board likely new ram.
Sockets are overrated even for power users. The only real reason sockets are still a thing is because Intel and AMD don't make mother boards. ARM hardware out of the gate isn't likely to have that large AIB Chip manufacturer ecosystem. So I would expect to see plenty of complete solutions board+chip.
I would hope we would get socketed chips at some point, I wouldn't hold my breath though. Good chance a move to ARM would greatly reduce the size of the enthusiast market. For average PC desktop users sockets just aren't important, and OEMs that sell the masses the majority of their machines would be fine selling soldered on chips. Cheaper to assemble, for likely to be treated like any other consumer device where customers just buy a new one. (again not that average PC users are upgrading cpus ever anyway)
We care sure... regular users nope they don't care.
Does anyone here actually know a REGULAR user that has ever upgraded a CPU. I mean come on no you don't
Put a better way how many of us [H] type users have honestly upgraded a CPU in the last 20 years. Intel releases a new board and often new socket with each new generation. AMD has been a bit better in that regard still unless your buying a new processor every year your going to need a new board anyway. I mean really when in the last ever has it been an option to just change a chip in a machine that's 3 years old. I am sure if we go through the history of the desktop we can find a few cases where if you bought the top of the line board the day that chipset launched perhaps a 3 year CPU only upgrade was an option... I don't believe it really works that way in practice. 3 years in your going to need a new board likely new ram.
Sockets are overrated even for power users. The only real reason sockets are still a thing is because Intel and AMD don't make mother boards. ARM hardware out of the gate isn't likely to have that large AIB Chip manufacturer ecosystem. So I would expect to see plenty of complete solutions board+chip.
I would hope we would get socketed chips at some point, I wouldn't hold my breath though. Good chance a move to ARM would greatly reduce the size of the enthusiast market. For average PC desktop users sockets just aren't important, and OEMs that sell the masses the majority of their machines would be fine selling soldered on chips. Cheaper to assemble, for likely to be treated like any other consumer device where customers just buy a new one. (again not that average PC users are upgrading cpus ever anyway)
Red Falcon
[H]ard DCOTM December 2023
- Joined
- May 7, 2007
- Messages
- 12,457
I never said the consoles were incapable of 1080p @ 60fps, in fact quite the opposite - you can read my earlier posts to confirm this.
My point was that, especially with a PS4 exclusive and well optimized game like Bloodborne, the CPU (even on the PS4 Pro) isn't powerful enough to keep the GPU fed data enough for the GPU (which again, is fully capable of 60fps, just like you stated) so the game ends up running at 30fps, which the devs have stated multiple times, on both the original PS4/Slim and Pro.
As for the the Diablo 3 system requirements, those weren't "theoretical" like you incorrectly stated, but I was a bit off - the ones I posted above were from another website and were the original system requirements from 2012, so I admit I was a bit off with them, but the actual system requirements aren't much different.
Here are the most recent system requirements from the game's website: https://us.battle.net/support/en/article/34784
Operating System Windows® 7 / Windows® 8 / Windows® 10 (latest service pack) Windows® 10 64-bit
Processor Intel® Core™ 2 Duo or AMD Athlon™ 64 X2
Video NVIDIA® GeForce® 8800GT or ATI Radeon™ HD 2900 XT or Intel® HD Graphics 4000
Memory 2 GB RAM
Storage 25 GB available HD space
Internet Broadband Internet connection
Media DVD-ROM drive (required for retail disc versions only)
Resolution 1024X768 minimum display resolution
Still, though, a Core 2 Duo and/or Athlon X2, both dual-core CPUs, depending on the generation and clock speed, aren't much faster than a Jaguar dual-core in single-thread or full-SMP, and in full-SMP are not as powerful overall as a quad-core Jaguar.
Also, I wanted ask, how do you KNOW that Diablo 3 won't play on a Jaguar-based system (other than the consoles) - have you actually tried it yourself?
DukenukemX
Supreme [H]ardness
- Joined
- Jan 30, 2005
- Messages
- 7,970
If you used Intel then yes you probably upgraded the board and the CPU. Me, I've stuck with AMD and pulled out many CPU's to upgrade them.
Also the motherboard chipset market sucks today. Looking at my Ryzen ASRock motherboard it has five USB 3 ports, but it also has two USB 2 ports. What reason is there to still have USB 2 when USB 3 is backwards compatible? Other than M.2 slots, I don't see any massive improvements on new motherboards. The only incentive to upgrade is because new CPU's aren't usually backwards compatible. Cause I would still be using my AM3 motherboard for the PCI slots and the VIa audio chip built into the board.
If AMD did make an ARM CPU for AM4 motherboards, then that would be extremely useful for users who want to switch between their ARM and x86 chips.
Ya I'm really not sure what he is on about... not that I would suggest it but I have run Diablo 3 on a Core Duo with no issues, and D3 was hardly the most demanding game I have ran on that old ass machine. I'm pretty sure an Athlon 5370 would be darn near identical performance to the old e6750 I still have kicking around.
Well they would be different chipsets... still. Ya if AMD ever released a socketable ARM chip likely wouldn't be that big a deal for them to push one chipset for both. Memory controllers are on CPUs, I can't imagine it would be a lot of overhead for PCI info ect.
Red Falcon
[H]ard DCOTM December 2023
- Joined
- May 7, 2007
- Messages
- 12,457
I'm sure it would be able to as well, considering even the *official* system requirements for Diablo 3 are basically a system from late 2007 or so.
I really don't understand why he would be saying that the Jaguar CPU couldn't run it, but considering he hasn't responded to either of us, maybe Pieter3dnow finally realized that his claims are basically nonsense and had no proof to back anything up.
Maybe I'm remembering him wrong, but I thought he was a PSU or technology reviewer back around 5-10 years ago, which is really why I expected more out of him.
If not though, then my mistake.
As a test for myself to see what the AMD Jaguar @ 2.0GHz is capable of with the system in my sig, I decided to load up The Talos Principle, mainly because I've heard that the game is a bit CPU intensive. (system requirements are on that page)
I ran it on Linux with both OpenGL and Vulcan APIs, of which Vulcan got about 5-10fps more, with more consistency, than OpenGL did on the same platform.
I kept a close eye on both the CPU, GPU, and PCI-E lanes usage through out the tests.
The Talos Principle was running on Medium & Ultra settings for the tests at 2560x1080.
From what I saw, it pretty consistently ran between 40-50fps with VSYNC disabled, sometimes reaching 60fps if I was close to a wall or looking at the ground, with VRAM usage around 1.2-1.5GB used.
The CPU used between 1-2 cores (game must not be optimized in Linux to use more than 2 cores in SMP?), so at most only 2 of the 2.0GHz Jaguar cores were used, but both were maxed.
With the GT 1030 2GB SFF GPU being used, the GPU itself only ever went at around 55-75% load, meaning the 2 Jaguar cores were effectively holding the GPU back by quite a wide margin.
The system it was running on did have a PCI-E 2.0 16x slot, but it was only 2.0 4x mechanical, and even then, only around 5-10% of the PCI-E bandwidth was being used, so the PCI-E lanes were for sure not the bottleneck.
Considering the GPU wasn't under full load, that meant that the Jagaur CPU was most definitely holding the GPU back, albeit with using only 2 of the 4 cores available, and I bet had all 4 cores been in use with full-SMP, it probably would have been more than enough to drive the GT 1030 properly and get 60fps+ consistently.
So I guess we could say that with the game, when running on only 2 Jaguar cores, means it was CPU bound and not GPU bound.
I will look further into why only two cores were being used, and maybe there was an -smp flag or option in the settings I missed - adjusting the CPU usage from lowest to ultra in the game settings didn't really affect things much.
Last edited:
Talos principle fun game. Decent Linux game, one of the first to add vulcan. Interesting that it only uses 2 cores. I wonder if perhaps it is only using logical cores ?
You should check and see that its set to use all cores in the OS... and then perhaps experiment assign it Cores 0 and 2 or something. Can't remember if the jaguar cores where all logical or if 2 shared interconnects or something like that its been awhile. Might be interesting to see if forcing it to different cores would make any real improvement.
If your not sure how to do that its pretty easy just fire up your terminal
$taskset -cp [pid]
This will return a current affinity list... I would assume it should say 0,1,2,3 but if it only says 0,1 or something you could try
$taskset -cp 0,1,2,3 [pid]
If its already set at the os level to use all 4 and its just the program isn't using them... perhaps try setting it to 0,2 and see if that runs faster on the jaguar. (its a nice trick to force programs to skip HT cores)
If your not sure of the processes PID you can run
$ps -aux
or use top or htop or whatever system monitor your like.
You should check and see that its set to use all cores in the OS... and then perhaps experiment assign it Cores 0 and 2 or something. Can't remember if the jaguar cores where all logical or if 2 shared interconnects or something like that its been awhile. Might be interesting to see if forcing it to different cores would make any real improvement.
If your not sure how to do that its pretty easy just fire up your terminal
$taskset -cp [pid]
This will return a current affinity list... I would assume it should say 0,1,2,3 but if it only says 0,1 or something you could try
$taskset -cp 0,1,2,3 [pid]
If its already set at the os level to use all 4 and its just the program isn't using them... perhaps try setting it to 0,2 and see if that runs faster on the jaguar. (its a nice trick to force programs to skip HT cores)
If your not sure of the processes PID you can run
$ps -aux
or use top or htop or whatever system monitor your like.
Last edited:
Rainbow 6 Siege is a great example of a CPU bound game for console. It has a Vsync toggle, and you can view your FPS in game!
By default, the game has a 30 fps cap in Terrorist Hunt because of the AI. Disabling Vsync will show massive FPS dips when the AI spawn.
Both the PS4 and Xbox One can get nearly 100 fps at times in matches without AI.
By default, the game has a 30 fps cap in Terrorist Hunt because of the AI. Disabling Vsync will show massive FPS dips when the AI spawn.
Both the PS4 and Xbox One can get nearly 100 fps at times in matches without AI.
Red Falcon
[H]ard DCOTM December 2023
- Joined
- May 7, 2007
- Messages
- 12,457
Yeah, I'm not sure what the deal is, though the Jaguar architecture is not like Bulldozer/Pildriver architectures, since Jaguar is not CMT (with no Hyper-Threading, so only 4 physical [and logical] cores) and not CMT like the others (no shared FPU per module).Talos principle fun game. Decent Linux game, one of the first to add vulcan. Interesting that it only uses 2 cores. I wonder if perhaps it is only using logical cores ?
You should check and see that its set to use all cores in the OS... and then perhaps experiment assign it Cores 0 and 2 or something. Can't remember if the jaguar cores where all logical or if 2 shared interconnects or something like that its been awhile. Might be interesting to see if forcing it to different cores would make any real improvement.
If your not sure how to do that its pretty easy just fire up your terminal
$taskset -cp [pid]
This will return a current affinity list... I would assume it should say 0,1,2,3 but if it only says 0,1 or something you could try
$taskset -cp 0,1,2,3 [pid]
If its already set at the os level to use all 4 and its just the program isn't using them... perhaps try setting it to 0,2 and see if that runs faster on the jaguar. (its a nice trick to force programs to skip HT cores)
If your not sure of the processes PID you can run
$ps -aux
or use top or htop or whatever system monitor your like.
I should try to force it using your method, though, I never thought to try that - good idea, thanks!
The OS and other applications like Folding@Home can definitely use all 4 cores in full-SMP, so it definitely is something up with the game itself, though from what the devs have stated, the Vulcan API is supposed to auto-detect usable cores (both physical and/or logical) but that sometimes it detects the true results incorrectly.
This might be why it is only using 1-2 cores instead of all 4 like it should, but again, I'm going to try your idea.
Last edited:
Red Falcon
[H]ard DCOTM December 2023
- Joined
- May 7, 2007
- Messages
- 12,457
That's actually pretty impressive and I didn't know that game even had those options.
That is a perfect example of being CPU bound, as when the AI starts to be used then the fps start to dip, probably since the CPU is too busy working the AI characters instead of being able to provide the GPU data properly enough to keep the higher fps up.
It's kind of funny, I remember back in 2015 playing Alien Isolation on an AMD FX-8150 clocked at only 2.0GHz (underclocked at the time for other testing purposes) paired with a GTX 770 2GB GPU.
The game almost always ran at 60fps+, except when the xenomorph appeared in the area (not even on screen yet) and the fps would take a noticeable dip since the CPU was now having to work harder in providing the AI for the xenomorph instead of being able to provide enough data to the GPU to get 60fps on the framerate.
In a way, it almost felt like cheating when I was playing since I would be running around in an area pretty fast, then when the fps would drop, I would know to crouch and slow my movements since that meant the xenomorph was in the vicinity due to the xenomorph's AI activating, even though the xenomorph itself was still physically a ways off and not even on screen yet.
Never thought being CPU bound in a game could actually be used to one's advantage!