Rx 580 8gb :: Hardware and Operating Systems

L37 2020년 4월 29일 오전 9시 16분

vadim님이 먼저 게시:
...
1. Can you provide benchmarks, please, to support your claim that CPUs with more cores provide better FPS than 4c/8t or 6c/6t?
...

Not trying to prove anyting here and this are definitely not controlled benchmarks (different driver versions etc), but this was actually the reason for me to swap CPU/MB (had this xeon lying around, got cheap chinese board and used the same RAM).
https://imgur.com/64ZczRE
https://imgur.com/8SgbdXH

#16

r.linder 2020년 4월 29일 오전 9시 18분

vadim님이 먼저 게시:
xSOSxHawkens님이 먼저 게시:
If you mean quad cores are enough as in the same way an FX8 is enough, then yeh, they are still 60FPS chips in most titles... But there are *plenty* of games right now that when pushed to max can and will saturate out a 4c/8t chip. Even (some) older titles.

Also... Did you not even read the post I made about VRAM above?...

Do you not see the *point* that a 4GB mid-range 670 is still usable in 2020 where a 2GB top of the line for the time 680 is no longer usable in 2020 due to VRAM limit?...

If its good enough to pull 60FPS, and if *low* settings push over 4GB, the card will be useless because of its 4GB limit. But if it can keep the GPU fed the user can keep gaming.

High VRAM low end cards are a joke (no one needs a 4GB GTS-250) but high VRAM mid-range and High-End cards *always* outlast their non-VRAM boosted counterparts in usable lifetime.
1. Can you provide benchmarks, please, to support your claim that CPUs with more cores provide better FPS than 4c/8t or 6c/6t?
2. As I can see, you don't deny that 8GB low end graphics is useless for now, but you expect that it can be used in the future. Sorry, but no-one knows the future requirements. Neither do you.
Exactly the same was told about RX 580, but this card already become obsolete and lowest level, and it still doesn't need 8 GB.
I use both GTX 780 Ti and GTX Titan (original) at my work and Titan doesn't have advantages in any titles despite it has twice more VRAM. 6 GB vs 3 GB.

1. Wholly depends on the game or workload. 4C/4T is low end now, and 4C/8T is pretty much dead minimum. It already shows its age in games where 6+ cores have more weight in determining the overall performance.
4C/8T is also the low end for both AMD and Intel now, it's not going to be more than a few years for 4C/8T to be less and less viable for latest releases since the standard has changed. Once proud beasts like the 7700K are now upper mid-range at best.

2.
a)8GB VRAM isn't useless on lower end cards. The 1650S is strong enough to handle higher graphics options, but it lacks the VRAM to actually run all of it.

b) If the price difference is small between a 4GB model and an 8GB model with basically the same performance, then it's a no-brainer to get the model with more VRAM.

c) You can predict how GPUs will progress and change per generation. The VRAM count has always been gradually increasing. 2GB is phased out entirely and 4GB is the dead minimum, with 6 being mid-range, and 8+ being the high end.

d) The RX 580 isn't obsolete, not sure what pile of dung you pulled that from. It's basically the same performance as the 5500XT and 1650 Super, can be found used at a steal, and still performs admirably at 1080p.

e) Well no, ♥♥♥♥, it's an original Titan, a 7 year old card. The RX 580 is around the same performance on average, and it's 3 years old. Expecting an older NVIDIA card to be anything special is laughable, because AMD cards age like fine wine while NVIDIA cards are left to rot.
The RX 580 was worse than the 1060 at launch, but then over time they became equal, and now the RX 580 is better on average by a few percent. The RX 5700XT was worse than the 2060-Super at launch, now it's just behind the 2070-Super for a much better price. AMD gets better with age.

r.linder 님이 마지막으로 수정; 2020년 4월 29일 오전 9시 19분

#17

xSOSxHawkens 2020년 4월 29일 오전 9시 20분

L37님이 먼저 게시:
vadim님이 먼저 게시:
...
1. Can you provide benchmarks, please, to support your claim that CPUs with more cores provide better FPS than 4c/8t or 6c/6t?
...
Not trying to prove anyting here and this are definitely not controlled benchmarks (different driver versions etc), but this was actually the reason for me to swap CPU/MB (had this xeon lying around, got cheap chinese board and used the same RAM).
https://imgur.com/64ZczRE
https://imgur.com/8SgbdXH

Exactly. Just look at how much better that GPU is being fed by the 6 core... Big gains there on the minimums too I am sure.

#18

xSOSxHawkens 2020년 4월 29일 오전 9시 28분

Escorve님이 먼저 게시:
AMD cards age like fine wine while NVIDIA cards are left to rot.
The RX 580 was worse than the 1060 at launch, but then over time they became equal, and now the RX 580 is better on average by a few percent. The RX 5700XT was worse than the 2060-Super at launch, now it's just behind the 2070-Super for a much better price. AMD gets better with age.

One of the best (modern) examples on AMD was the first gen Vegas. At launch the Vega 64 could barely match the 1070 stock and would even ocasionally fall to the 1060-6GB...

Now days it consistantly sits above 1080 non-ti's...

Thats like 3 product segmentations to jump through from drivers maturing.

I didnt like Vega at launch, specially with its price, but for 400 bucks in december 2018 it was a no brainer.

#19

vadim 2020년 4월 29일 오전 9시 40분

xSOSxHawkens님이 먼저 게시:
the 780ti and the titan is because you are choosing to use settings that do not effetively use the titan. FFS, toss on some Ultra textures and even in 1080P you will see most modern titles chew through 4.5GB+ on VRAM in a heartbeat, and at such a rez that titan will still sing the songs of our PC gaming people great.

Not a single person in their right mind would try to set ultra settings on ancient graphics. I always choose settings to get stable 60 FPS. On these settings Titan doesn't run better (except small difference because of more CUDA cores).
About allegedly insufficient number of CPU cores - can you, please, provide a link to the article on the reputable site. TBH, don't want to waste my time on noname video.

#20

L37 2020년 4월 29일 오전 9시 49분

xSOSxHawkens님이 먼저 게시:
Exactly. Just look at how much better that GPU is being fed by the 6 core... Big gains there on the minimums too I am sure.

It is v2, so it is actually 8C16T.
The main issue i had with i7 is that i was not obviously playing at 30FPS, i was lowering settings a bit to get to ~60 (a demonstration that 5700XT is not enough to run at 4k with max settings too), and with i7 i still had those 20-30FPS in certain conditions, with annoying stutters from time to time. Changing CPU fixed that completely.

L37 님이 마지막으로 수정; 2020년 4월 29일 오전 9시 50분

#21

xSOSxHawkens 2020년 4월 29일 오전 10시 11분

vadim님이 먼저 게시:
xSOSxHawkens님이 먼저 게시:
the 780ti and the titan is because you are choosing to use settings that do not effetively use the titan. FFS, toss on some Ultra textures and even in 1080P you will see most modern titles chew through 4.5GB+ on VRAM in a heartbeat, and at such a rez that titan will still sing the songs of our PC gaming people great.
Not a single person in their right mind would try to set ultra settings on ancient graphics. I always choose settings to get stable 60 FPS. On these settings Titan doesn't run better (except small difference because of more CUDA cores).
About allegedly insufficient number of CPU cores - can you, please, provide a link to the article on the reputable site. TBH, don't want to waste my time on noname video.

Not all setting are created equal, and some setting can be maxxed even on older hardware.

Running better textures will (in many cases) provide a noticable increase in quality, with an average of 10-15% adiional core loading in most gamer per level setting, but the setting EATS VRAM for breakfast.

Just because you have an older titan doesnt mean you need to run medium everything. Quit sticking to the pre-sets and start setting your settings youself.

That way you can max out the ones that you still have room to max out, and with 6GB of VRAM and an older but still usable GPU, you have no reason *not* to be running max textures in your games unless they are off the walls crazy like 8K textures which few devs even offer natively.

And at this point three seperate people have ancedotally told you you were wrong from their experiance *and* we have provided proof. Its not on us to provide the *type* of proof you want. If you are too lazy, and that is what it is, too lazy to watch a few quick vids, or even scan through them for avareages, then we are not able to help you and you are not worht helping. Perhaps look at what has been offered instead of insisting people jump thorugh hoops to please you when you are the one demanding proofs.

As the old adage goes, pics or it didnt happen (or vid in this case).

L37님이 먼저 게시:
xSOSxHawkens님이 먼저 게시:
Exactly. Just look at how much better that GPU is being fed by the 6 core... Big gains there on the minimums too I am sure.
It is v2, so it is actually 8C16T.
The main issue i had with i7 is that i was not obviously playing at 30FPS, i was lowering settings a bit to get to ~60 (a demonstration that 5700XT is not enough to run at 4k with max settings too), and with i7 i still had those 20-30FPS in certain conditions, with annoying stutters from time to time. Changing CPU fixed that completely.

Nice on the 8c!

Oh, I totally get that lagging bit. Part of the reason I decided to pull the trigger on mine (4790K @4.8 > 3900X) was for the same issues in GTA-V.

When runing at upper settings in GTA-Online I would consistently drop into the mid 40's when cruising through down town, and would see my GPU bottleneck down multiple times in a game session. Now my minimums are all *much* higher, my dips in the city are more or less gone, and my GPU will run 99-100% at all times as it should. Much better

#22

vadim 2020년 4월 29일 오전 10시 29분

GTA V cannot use more than 4 cores...

#23

xSOSxHawkens 2020년 4월 29일 오전 10시 54분

vadim님이 먼저 게시:
GTA V cannot use more than 4 cores...

Yes it can. It can easilly thread out to 8 or more thread when needed.

Dont think so?... Go boot it into MP using a 4c/8t chip at 2Ghz core speed or an FX8 chip and watch all 8 threads carry a load...

OR

If you have access to a more modern system boot GTA up, turn off pause/mute on focus loss and set it for windowed borderless, and run it on a primary screen with a secondary screen.

Now, for best results have soemthing with 12+ cores so you can isolate the game onto non-used cores *and* have room to scale out as you go... For me I start by forcing the exe to thread 24 which locks it to CCD2 core 12.

Now, run yourself through the city form one side to the other with the game locked via affinity settings to 1c1t, then scale out from there. Keep it OFF cores 0/1 and threas 0-3 so that it is not being interephed with by other things, and so you know that any loading on the extra cores is indeed the game.

Now watch as the game does in fact make use of more than 4 threads. Specifically there are some major gains in stability of frame rate and minimum dips with 6 cores, but they are not there as much with 3c/6t...

Or... Take techspots word for it...

techspot님이 먼저 게시:
...GTA V is one of the few games we have tested recently that does a good job utilizing more than four threads.

https://www.techspot.com/review/991-gta-5-pc-benchmarks/page6.html

Or DSOGaming who says they were abel to confirm usage up through 12 threads...

dsogaming님이 먼저 게시:
Unfortunately, the performance differences between our simulated quad-core and our hexa-core systems were minimal. However, GTA V scaled on all twelve threads, proving how efficient and flexible RAGE Engine has actually become.

https://www.dsogaming.com/pc-performance-analyses/grand-theft-auto-v-pc-performance-analysis/

So yeh... GTA absolutely will, you can test it personally, I can prove it in a vid, and there are two text sources professionally saying the same.

Not sure why you think its limited to 4c/4t but its flat out wrong.

xSOSxHawkens 님이 마지막으로 수정; 2020년 4월 29일 오전 10시 55분

#24

vadim 2020년 4월 29일 오전 11시 00분

xSOSxHawkens님이 먼저 게시:
Dont think so?... Go boot it into MP using a 4c/8t chip at 2Ghz core speed or an FX8 chip and watch all 8 threads carry a load...

OS scheduler works such way, that if you run it on 64 cores CPU, all 64 cores will be used.
There are a lot of benchmarks that clearly show that this game doesn't benefit from more than 4 cores.

#25

r.linder 2020년 4월 29일 오전 11시 04분

vadim님이 먼저 게시:
xSOSxHawkens님이 먼저 게시:
Dont think so?... Go boot it into MP using a 4c/8t chip at 2Ghz core speed or an FX8 chip and watch all 8 threads carry a load...
OS scheduler works such way, that if you run it on 64 cores CPU, all 64 cores will be used.
There are a lot of benchmarks that clearly show that this game doesn't benefit from more than 4 cores.

^

All that really matters is core performance. Even in games like AC:O where more cores can make a difference, there are still quad-cores that can hold their own because they have the core performance to make up for the lack of cores.

7700K has better core performance than the R5 3600, so it can keep up with the 3600 in those titles and beat it in titles where ~4 cores is plenty.

It's just pure core performance that has the most impact. Though how many cores you actually need depends on what you're doing.

r.linder 님이 마지막으로 수정; 2020년 4월 29일 오전 11시 04분

#26

xSOSxHawkens 2020년 4월 29일 오전 11시 15분

Escorve님이 먼저 게시:
vadim님이 먼저 게시:
OS scheduler works such way, that if you run it on 64 cores CPU, all 64 cores will be used.
There are a lot of benchmarks that clearly show that this game doesn't benefit from more than 4 cores.

^

All that really matters is core performance. Even in games like AC:O where more cores can make a difference, there are still quad-cores that can hold their own because they have the core performance to make up for the lack of cores.

7700K has better core performance than the R5 3600, so it can keep up with the 3600 in those titles and beat it in titles where ~4 cores is plenty.

It's just pure core performance that has the most impact. Though how many cores you actually need depends on what you're doing.

And yet... I linked back to two different valid sources showing it *does* use more than 4 thread and does so effectively.

Also I am well aware of the windows scheduler, but that is not the cause here.

The only benchmarks that outright claim no increase with more than 4 threads are at this point in history shown to be either fault of tester, or GPU limited.

Again, you can claim that it only uses 4 threads, but I can *show* in live video as well as in sourced professional edditorial that it *does* make use of more than 4 threads.

Period.

How about this, just go try it. Go boot the game up, and the sclae it out from 1t on up.

Forcibly lock it to specific threads and cores with third party utilities if you want.

You will see that it:

A) threads out

and

B) gets the biggest advatage from 1c/1t threading vs 1c/2t threading (ie runs better with 6c6t than it does with 3c6t, or runs better with 4c4t than 2c2t).

It even has command line options to disable HT/SMT support if you so wish, but *by default supports SMT/HT as enabled and used*.

#27

r.linder 2020년 4월 29일 오전 11시 17분

xSOSxHawkens님이 먼저 게시:
Escorve님이 먼저 게시:

^

All that really matters is core performance. Even in games like AC:O where more cores can make a difference, there are still quad-cores that can hold their own because they have the core performance to make up for the lack of cores.

7700K has better core performance than the R5 3600, so it can keep up with the 3600 in those titles and beat it in titles where ~4 cores is plenty.

It's just pure core performance that has the most impact. Though how many cores you actually need depends on what you're doing.

And yet... I linked back to two different valid sources showing it *does* use more than 4 thread and does so effectively.

Also I am well aware of the windows scheduler, but that is not the cause here.

The only benchmarks that outright claim no increase with more than 4 threads are at this point in history shown to be either fault of tester, or GPU limited.

Again, you can claim that it only uses 4 threads, but I can *show* in live video as well as in sourced professional edditorial that it *does* make use of more than 4 threads.

Period.

How about this, just go try it. Go boot the game up, and the sclae it out from 1t on up.

Forcibly lock it to specific threads and cores with third party utilities if you want.

You will see that it:

A) threads out

and

B) gets the biggest advatage from 1c/1t threading vs 1c/2t threading (ie runs better with 6c6t than it does with 3c6t, or runs better with 4c4t than 2c2t).

It even has command line options to disable HT/SMT support if you so wish, but *by default supports SMT/HT as enabled and used*.

But see, when I lock GTA 5 to use only 4 threads, it runs better than it does with more than that. There is no improvement with more than 4 threads, it's purely core performance that makes the difference.

#28

vadim 2020년 4월 29일 오전 11시 18분

https://www.tomshardware.com/reviews/multi-core-cpu-scaling-directx-11,4768-3.html
https://www.techspot.com/review/991-gta-5-pc-benchmarks/page6.html
https://www.extremetech.com/gaming/203800-rounding-up-gta-v-on-the-pc-how-do-amd-intel-and-nvidia-perform
...and so on.

#29

upcoast 2020년 4월 29일 오전 11시 48분

https://www.youtube.com/watch?v=o8l-kEs3XH8

The older i7 4c/8t had always beat out the i5 4c/4t in GTA V at the same clocks.

upcoast 님이 마지막으로 수정; 2020년 4월 29일 오전 11시 49분

#30

게시글 신고