[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: GbOJAhtXkAAcoqk.jpg (34 KB, 679x468)
34 KB
34 KB JPG
>>
>>103038735
the amd shills will ignore this thread or do damage control
side note, are you guys super tired of the amd shilling? its fucking annoying and anyone in pcbg that says they want an intel chip, they start attacking
>>
File: GbOMlloaQAAinky.jpg (76 KB, 1190x675)
76 KB
76 KB JPG
>>
>>103038735
>>103038808
>>103038833
Why would I whine about minimal 5 or 10% improvements ON A CPU? and I say this as someone who has bought 3 ryzens since 2016
>>
When will they admit that Moore’s law has expired? Even arm64 is having incremental gains once they ironed out a design.
>>
>>103038808
+8% performance increase bad but intel -20% performance decrease good
>>
>>103038735
>up to 26% faster
>1% faster in a single title
>total average is still double digits faster
>this is somehow bad
Keep on trying, Rakesh. I know you need the Rupees.
>>
there usually isnt that big emphasis on cpus when it comes to gaming.
At 1440p theres like 20 cpu from 150€ to 500€ that produce within 10% same fps.
>>
>>103038808
Yeah, I'd much rather AMD released chips that actually performed WORSE than the ones they're replacing. Fuck off, Pajeet.
>>
>>103038808
>anon you should buy the better product
>>nooooooo shut up shut up shut up you are shilling
>>
File: 1718247041911.jpg (32 KB, 797x500)
32 KB
32 KB JPG
>>103038735
LMAO OF COURSE
>>
>>103038920
because the price is 20% more
>>
>>103038971
Yeah I kinda hate how these CPU gaming benchmarks are done these days. They show significant differences because they use unreasonable setup and settings. Who plays 1080p with a 4090? It's very niche at best. Testers defend their decisions by saying that it's the only way to see significant differences. If you need an unreasonable setup to see significant differences, then it is unreasonable to claim significant differences!

CPU is not a top priority for most people who are building a gaming rig these days. Even with a 4090 essentially any high end CPU will do. You likely won't see much of difference between 5800x3D and 9800x3D unless you are CPU bound, in which case you're either doing it wrong, or you are playing something that isn't that graphically demanding (and in those games x3D CPUs aren't that great anyway)
>>
>>103039448
and if a game is cpu bound its probably shit like asstastic npc ai coding that does huge bottlenecks on 1 thread instead of 8 threads being fully utilized and you'd get more fps using a 12 core cpu
>>
its more like 11% faster but thats just avg fps. the real advanatage of x3d is the minimum lows. shader stutter is non existent on the 9800x3d unless the game engine by design stalls the render thread to compile.
>>
>>103038808
I like amd and all but the x3d cpus are extremely over shilled in pcbg for some reason, that and am5 boards with ddr5 ram. 'budget options? what are you, poor?'
>>
>>103039448
>in those games x3D CPUs aren't that great anyway
it’s been like a week since the factorio expansion launched and you already forgot?
>>
I'm starting to think am5 is a skip tier. My 5800x3d just werks..
>>
Is it that minimal of a difference that I'm fine with my 5900x + 4070S?
>>
>>103039385
Then buy the older one that's on sale and its 50% cheaper
>>
>>103039448
>eah I kinda hate how these CPU gaming benchmarks are done these days. They show significant differences because they use unreasonable setup and settings.
My nnnnguy. How are you actually this fucking stupid. Are you actually literally 12 years old?
This is a CPU benchmark. This is standard procedure for CPU benchmarking. The purpose here is not to show you 10 year old retards how well you can run the game, the purpose is to highlight differences in CPU performance because it can reveal intricacies about the CPU itself when compared to other competing arch implementations.
It has been done this way forever. This isn't new. You are new. You are gay. You are not white. You are retarded.
>>
>>103039448
A CPU matters for CPU-bound games.
>old source/source1 games
>factorio
>minecraft
>physics/simulation/factory type games
>sandbox games
Anything where the graphics take second seat to the game mechanics, and the game mechanics require or allow the existence of tens to hundreds of thousands of independent entities/machines/physical objects.

I fully agree that benchmarking them on the latest AAA graphics slop is worthless. Give me the FPS of firing a davy crockett at gmod prop build, or the FPS of an endgame factorio base.
>>
>>103039750
A bit of an unrelated question, but do any games like that do the game logic on the GPU? Would be a great fit
>>
>>103038833
lel
>>
>>103038833
Intel'aviv is finished
>>
>>103039448
clueless. imagine not knowing what latency is in 2024.
people with very high end gaming rigs still play at 1080p for extremely high frame rates, have a good think why.
>>
>>103040004
>still play at 1080p for extremely high frame rates
idk why you would over 170 begins to feel like you're travelling forward in time and the difference between 1440 and 1080 is not having your brain segmented into a bunch of scanlines
>>
>>103038735
8% faster after 2y. for +$200
meamwhile Tim Cock's Applel is adding 20% gen over gen and lowering the price
>>
>>103040085
>meamwhile Tim Cock's Applel is adding 20% gen over gen and lowering the price
>except you can only run logic
>and final cut still chokes
lol
lmao
>>
>>103039959
It depends. Dyson Sphere Program does a bunch of GPU compute, but it's mostly for kind of interpolation behaviour after the CPU has calculated the logic - e.g. a flying drone unit is transporting an item from container A to container B, internally the CPU doesn't update it until its pre-calculated arrival time, but the GPU interpolates its actual physical position between the two containers to render it in-world.
The thing is I don't know much about GPU compute but my very surface-level understanding is that it's not very well-suited to highly branching logic. So "recalculate an update for an in-progress process" is well suited, but e.g. "check if the source inventory has an item available, that item matches your internal whitelist, the destination inventory has space available, that inventory can accept said item, and if so trigger the action to pick up the item and move it" less so. Factorio for example is very efficient at putting inactive machines "to sleep" and only processing them when something's actually happening, so for the actually important interactions I don't know how much room there is for offloading.

Now physics engines do exist on the GPU (remember PhysX), so that's much more likely, though I don't know for sure since I don't play that many physics simulation games.
>>
>>103039722
>muh done for decades
Retards like this miss the point.
Hardly anyone gave a shit about CPU perf until "journalists" ran interference with intel with muh 640x480 benches against Ryzen. Before you just bought for fucking i5 unless you had a reason for hyperthreading.

According to that fucking chart lots of people should have been buying FX 8150s, the line of CPUs that was the ass end of jokes for years and almost sunk AMD. Or an HEDT platform that generally never sold well.

If you believe that it's somehow a valid exercise in looking at different microarchitectures that's fine, but the mainstream focus isn't on whatever pseud exercises you think they are. With GPUs being locked up by NVidia and GPUs being much harder to buy, people are going to latch on to whatever graphs tell them will give them a boost. Damn the fact that only have a 4070 and a 1440p monitor.
>>
>>103040093
Funny you should say that as I'm using FCP as we speak on an m2pro. Shit is smooth like butter with 1.6gbps c4k prores
>>
>>103040111
>latch on to whatever graphs tell them will give them a boost.
AMD can print modern smaller chips
Intel has shown they refuse to hire anyone to solve their issues and will become a zombie company that makes research chips for apple ipods.. the US government will prevent their pop because a boomer who bought computer = mone will lose money soon
>>
>>103040111
>newfaggot types up this cope post to try and brush off three decades of CPU benchmarking because he doesn't understand it
You're not white. You're gay. You're retarded.
>>
>>103040116
>Funny you should say that as I'm using FCP as we speak on an m2pro. Shit is smooth like butter with 1.6gbps c4k prores
>that'll be 1500 dollars goy
and I'm using an i7 3570k and 1070ti for a total of 300 dollars used to run davinci resolve and render 1440x2560 60fps lol kill yourself itoddler
>>
>>103040004
People literally only do this for extremely sweaty e-sports games. Out of all the ones on the benchmark above, that's literally pretty much just CSGO 2. Maybe CoD, though I don't know how big its actual pro scene is.
If latency was the reason to chase high FPS, then benchmarks would show off CSGO, Valorant, Apex Legends, Overwatch, etc. Even assfaggots like Dota would be better candidates than fucking starfield or cyberpunk.
>>
>>103039750
>the FPS of an endgame factorio base
check out factoriobox.1au.us
>>103040111
>Before you just bought for fucking i5 unless you had a reason for hyperthreading
underage b&
>>
>>103040181
oh shid they already have some 9800x3d results
>>
File: iuuaepfqkcmd1.jpg (61 KB, 500x572)
61 KB
61 KB JPG
>>103038833
>literal massive intel failure
big oof
israel lost
intel corporation is dead
>>
>>103040156
I'm sorry you're poor, Rakesh
>>
>>103040235
probably reviewers testing factorio and not realizing that the results are uploaded immediately lmao
>>
>>103040141
You lack nuance, you have no right to call anyone nonwhite.
IDK where this idea that some charts of a German site are the focus of just everyone building a gayming PC

>>103040181
>underage b&
Been here a hell of a long time. It's funny to see people try to rewrite history. HT didn't mean dick for games generally going back far enough.
>>
>>103040261
oh yeah fruitcake, davinci resolve is also FREE
>>
>>103040244
>intel corporation is dead
how to profit off this knowledge? the government will actively be stopping you from doing that because boomers bought tech stocks and if they get liquidated because of a bad apple..
>>
>>103038833
>not comparing against the 14900k
advanced marketing devices are at it again
>>
>>103040273
>bb-b-b-but da germans
It has bee an industry wide standard for 30 years. Longer than you've been alive, shit eating nonwhite retard.
The game is just a real world workload. Seeing the results across multiple architectures shows you limitations and advantages of each arch when you compare numerous benches together.
You not understanding this but continually bloviating here is ample proof that you have a two digit IQ.
>>
>>103040300
they should have compared it against both to embarrass intel further
they can still fuck with the RAM and power settings on the 14900K to guarantee a loss on all games
>>
>>103039448
The idea I guess is to avoid having the GPU be the bottleneck, but then you've got endless numbers of GPU generations running increasingly old games at a resolution that was maybe interesting 14 years ago. How do I show off the difference between the intel 285 486 777 9410 1776 k and the ryzen 69 420 52 AI PXQMD Shart X4D: liquid snake edition? Oh, I'll just use a 9090 Ti Titan Pro and run fuckin counter strike, shadow of the tomb raider, F1 2022, and apex legends at 800x600!
>>
>>103040304
Just because something is done doesn't mean it's something people pay attention to.
>oh yes, I bought an FX 8150 because of performance in crysis 3
said no one fucking ever (although I wouldn't discount one or two amd tards clinging to such a result)
>>
>>103039750
Yep, those are the sort of games I meant when I mentioned CPU-bound non graphically demanding games. Those generally don't benefit as much from extra cache as more "real time" games do, presumably as the computation is more predictable and thus lower number of instructions are likely to be needed at any given moment. They tend to benefit more from high clocks and core count


>>103039557
I didn't mean that they aren't still good, I meant that they don't reign supreme over CPUs with many more cores.

>>103040004
That's the 'very niche' I mentioned, so this: >>103040165

>>103039722
You are equally competent in understanding my arguments as you are estimating my age
>>
>>103040276
I already relayed my pity, Rakesh.
>>
>>103040344
Why is it so hard for you dumbfuck kids to understand the idea of aggregating data?
Eliminate as much GPU bottleneck as reasonably possible and it exposes the delta in CPU performance from one CPU to another. The vendors give you their media deck talking about a given arch, so you kno basic things like hitting L3 takes X cycles, 1 cycle slower/faster than the prior gen.
Using games as a bench provides you a window into how the CPU is functioning. You can look at any given CPU, all models that came before it, knowing what generational changes happened between, and point to real world performance uplifts that are directly caused by those architectural changes.
>>103040367
Literal mouth breathing 70 IQ shitskin
>>
>>103040389
>uhmmm sorry I just need an apple idrone because if I don't text my girlfriend in blue bubbles she'll take the children away and have me castrated
>>
My 5600x plays all the games at 1080p 144fps and the ones that don't are horribly poorly optimized!!
>>
>>103040374
>Those generally don't benefit as much from extra cache as more "real time" games do, presumably as the computation is more predictable and thus lower number of instructions are likely to be needed at any given moment. They tend to benefit more from high clocks and core count
The 7800x3d is still the king of factorio benchmarks, though. So I'm not sure that's really true.
>>
>>103040402
That wall of text can't disguise the fact you're getting excited over bar graphs lmao, same as any month breathing normie looking for a fix into their high
>>
>>103040451
>The 7800x3d is still the king of factorio benchmarks
only on small maps
4chan doesn't let me post the link but go to factoriobox.1au.us and select the 50k map
the current world record is held by a 14900KS (with a ridiculous B-die overclock)
>>
>>103040487
>70 IQ shitskin mounts the best defense of his ego that he possibly can
lol
>>
>>103040451
if you need a cpu that expensive for factorio, you have severe autism
>>
>>103040492
>literally just set 10 hours ago
zamn
>>
>>103040532
desu i think the 7800X3D could do at least 70 UPS if someone really tried
but the cache doesn't help
>>
>>103040413
You shitskins say the weirdest things I swear to God....
>>
>>103040085
actually its +$30, the 7800X3D released for $449
>>
>>103040402
>hurr durr butthurt and angery
How about you stop shitting your pants over a hyperbolic joke about how cpu gaming benchmarks are done? There are plenty of pure CPU workloads that could be used to benchmark anyway.
>>
>>103040664
How about you stop being a 10 year old mad about things you don't understand, little shitskin kid?
Competent review sites do a full suite of benches spanning office work, scientific workloads, rendering, and gaming, all to draw a more complete picture of the parts they're highlighting. You are whining like a little faggot baby because you didn't understand gaming benchmarks out of context.
>>
>>103040681
>whiny /pol/troon can't take a joke
>not even one aimed at him personally
>but at an inhuman third party
Go to the bathroom NOW!
>>
>>103040691
>more 70 IQ cope trying to save face after being a total fucking idiot for a handful of posts
Sad.
>>
>>103040451
Well perhaps I was overtly generalizing. I was thinking about games like Stellaris, Civilizations and Total War. I've seen benchmarks where 7800X3D got outperformed by 9700X in Stellaris, and by 7600X in Civ VI; presumably due to higher clocks
>>
>>103038971
Oh well. If you are coming from an old 4/6 core you might as well get the X3D. Not everyone has to justify to themselves upgrading from a good previous gen CPU.
>>
>>103040031
>over 170 begins to feel like you're traveling forward in time
Only during rendered 3D gaming? Or in general?
>the difference between 1440 and 1080 is not having your brain segmented into a bunch of scanlines
So the scanlines disappear at 1440? If so, does that apply to only rendered 3D gaming or in general?
>>
>>103038808
>anyone in pcbg that says they want an intel chip, they start attacking
gee I fucking wonder why
>>
>>103040096
Branching is indeed an issue, but you can mostly deal with it by grouping similar things together. Maybe the tradeoffs with complexity in development and so on are just not worth it.
>>
File: hmm coffee.png (226 KB, 645x521)
226 KB
226 KB PNG
>>103040973
>you might as well get the X3D
That's what the 1080 + 4090 benchmarks would imply, but does it really make sense? Instead of a 7800x3D you could get a 7700X for cheaper, and spend the extra money on a better GPU, which could have much larger impact on gaming performance. Or you could get 7900X for similar price. You'd lose a tiny bit of gaming performance, but 7900X would be faster on productivity, and anything that could leverage it's cores and clocks
>>
File: 1714252636302952.jpg (9 KB, 246x227)
9 KB
9 KB JPG
>>103040699
you are literally shitting your pants and crying about imaginary ethnic groups because you're too dense to comprehend a joke.
>>
>>103041178
It does make sense if you on an old 4/6 core. There's no need to justify to yourself a straight forward upgrade, any more than you did when you first bought the 4/6 core years ago.
>>
>>103041178
I think that the difference is that CPU performance provides a hard baseline, and there's usually almost no way to reduce CPU load. Whereas GPU performance, more than anything, provides headroom for fancier settings - but it's almost always possible to turn those settings down far enough to match the available performance.
1080p on a 4090 is an extreme case, used in data collection to ensure there is no chance of bottlenecking. You don't actually need a 4090 on 1080p to reach those framerates, obviously enough.

>7900X would be faster on productivity
Similarly, if you don't actually do typical CPU-heavy productivty tasks, then there's little point in that. If you know that you often compile code, or run simulations, or run multiple VMs, or whatever - then sure. But if you don't do anything of the kind, then you will lose a tiny bit of gaming performance for zero actual practical gain.
>>
>>103040004
>buying a high end rig to play your single player game in 200fps@lmao1080p when very diminishing returns start somewhere at 60-80fps
Some people are just retarded consumers. Inb4 competitive games, those can usually achieve 300+ fps easily, even on rigs from the lower end of the mid range.
>>
>>103038735
Okay but how does it perform in AI workloads??
>>
>>103041227
How does it make more sense than a non x3D CPU?

>>103041317
7700X would still save money with out really impacting gaming much. Unless you're CPU bottlenecked, in which case using that money on a better GPU would be more efficient way of out of the bottleneck than paying extra for the CPU
>>
Makes sense to go from a 7800x3d to a 9800x3d?
>>
>>103041668
maybe for the hardcore blender crowd
>>
>>103041668
Only if money is not at all a concern, or you have some hyper specific workload where you KNOW you are CPU limited and 5-15% uplift is worth it
>>
>>103041668
I was already thinking of 5900x to 9800x3d being a bit dumb of an upgrade to consider, why would someone do this
>>
I'm gonna buy a 9700x
>>
>>103041701
why
>>
>>103041658
>How does it make more sense than a non x3D CPU?
Most people prioritize gaming, otherwise they wouldn't still be on a 7700K or 8700K. If they have a 5950X then that probably says something different about them.
>>
>>103041797
While that's true in most cases, I just did it because it was cheaper in a moment of panic
In retrospect I should have gone for an X3D processor but now I'm stuck in an awkward place where it seems a bit dumb to upgrade but not really
>>
>>103040492
B-die is still top dog huh
Never expected it to have such strong staying power
>>
>>103041773
I need to upgrade from a 7700k, i don't play vidya and i'm not paying more than 350 for a fucking cpu.
>>
>>103041894
Understandable. When the 5950X first came out it was a great all around CPU. We didn't know there would be an X3D, or that it would mitigate the latency issue.
>>
>>103041989
The funny thing is that AMD presented a 3d vcached 5900x at 2020 CES, more than a whole year before the 5800x3d came to market
>>
>>103041979
Uh, 13400F it is then. Not like you need to do anything heavy.
>>
>>103041954
to be fair he's running 4533 G1
most chips struggle to get to 4400 G1
i honestly doubt his setup is daily stable
>>
>>103041979
the usual pick is 7600, or 7800x3d
the 9000 series are largely irrelevant due to pricing
>>
>>103040292
buy amd stock now when it's cheap and become a waitchad
>>
>>103040292
>how to profit off this knowledge?
Short INTC.
>>
Moore's Law is Dead.
>>
>>103042036
Right, AMD has had 3d v-cache tech for a while, it just took some time to get a viable product to market.
The 3000xt CPUs were originally intended to be x3d chips, but it didn't happen cause AMD & TMSC couldn't mass produce those yet.
>>
>>103042569
Electron microscopy shows no TSV bumps before the 5000 series.
>>
>>103038930
moores law was never true, it always was just something that was repeated that made people that dont know any better go
>huh yeah, i guess thats true
>>
>>103038808
>anyone in pcbg that says they want an intel chip, they start attacking
Would it have to do with all 13 and 14k processors being potential duds and RMA fodder? I don't do brandwarrior shit but Intel has been a constant fuckup for like a decade now.
>>
File: 1728413475355966.png (158 KB, 652x767)
158 KB
158 KB PNG
>>103038833



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.