• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

AMD "Ellesmere" ASIC Pictured Up Close in RX 480 PCB Picture Leak

cdawall

where the hell are my stars
Joined
Jul 23, 2006
Messages
27,670 (5.12/day)
Location
Houston
System Name All the cores || Into the Blue
Processor 2990WX || 5960X
Motherboard Asrock X399M || Asus X99M WS
Cooling CPU-XSPC RayStorm Neo, 2x240mm+360mm, D5PWM+140mL, GPU-2x360mm, 2xbyski, D4+D5+100mL || TBD
Memory 4x8GB G.Skill Trident Z 3200 CL16 || EVGA 3200 2x8GB
Video Card(s) (2) EVGA SC BLACK 1080Ti's || EVGA 1080Ti FE
Storage 2x Samsung SM951 512GB, Samsung PM961 512GB || OCZ Vector 180 480GB
Display(s) Dell UP2414Q 3840X2160@60hz || TBD
Case Caselabs Mercury S5+pedestal || Supermicro S5 (GS5A-753K)
Audio Device(s) Fischer HA-02->Fischer FA-002W High edition/FA-003/Jubilate/FA-011 depending on my mood
Power Supply Seasonic Prime 1200w || Seasonic Snow Silent 750w
Mouse Thermaltake Theron, Steam controller || TBD
Keyboard Thermaltake Poseidon ZX, Steam controller || Corsair K70
Software W10P || W10P
I said nothing at all about performance in my comments so it seems you are missing the point.

1. Go back and look at the post i was replying to, that post was talking about how amd has similar or even better oc headroom that nvidia, which is true because amd is clocking this thing very conservatively at stock, so i simply said this round nvidia got mad clocks and i wonder if it has anything to do with 14nm vs 16nm. Does that make more sence now? How about you guys contribute to a logical argument rather than being all offended. And secondly, any sane person will not compare gpu to cpu, in cpu single threaded performance is important so clock as well as ipc go hand in hand, then after than comes number of cores, gpus on the other hand arent that way since its all about fitting as many cores and clocking them as high as you possibly can before reaching thermal headroom. So while clock is not everything, it still works differently than with cpus

You missed the entire point of my post. Clockspeed doesn't matter. GPU's do not have cores you are correct, but do you know what an SP is, do you know how they work? Did you know they stuff as many of those into a GPU as is possible and those basically make the GPU. Something about number of transistors and those actually do the work.

So instead of arguing with the entire forum, all of whom has said you are wrong how about do some research.

Oh and just in case you missed it clockspeed doesn't matter. The GTX980 hit 2.3ghz in 2015, it also wasn't the first GPU to do so. http://www.kitguru.net/components/g...eforce-gtx-980-hits-2-30ghz-sets-new-records/
 
Joined
Jul 9, 2015
Messages
2,804 (1.32/day)
System Name My all round PC
Processor i5 750
Motherboard ASUS P7P55D-E
Memory 8GB
Video Card(s) Sapphire 380 OC... sold, waiting for Navi
Storage 256GB Samsung SSD + 2Tb + 1.5Tb
Display(s) Samsung 40" A650 TV
Case Thermaltake Chaser mk-I Tower
Power Supply 425w Enermax MODU 82+
Software Windows 10
"Naked" 480, not sure why it is a thing, but here we go:




@cdawall
Why do you think it would be "Fury replacement"? (at stock)
I could see 300$ (1500+) version going there, but don't 1266Mhz leaks show it somewhere at 980 (stock), and even slightly below it?
 
Joined
Jun 18, 2016
Messages
2 (0.00/day)
"Naked" 480, not sure why it is a thing, but here we go:




@cdawall
Why do you think it would be "Fury replacement"? (at stock)
I could see 300$ (1500+) version going there, but don't 1266Mhz leaks show it somewhere at 980 (stock), and even slightly below it?
36 CU's says it all.

This card is an R9 390 replacement at a better price.

Vega will bring the Fury replacement cards. Only these will be called the RX 490/x this time.

I don't expect much better performance than the R9 390 if both cards are clocked the same in DX11 titles. This card will sit firmly in the R9 390x performance range but it's clockspeeds will help it pull away in some games.
 

cdawall

where the hell are my stars
Joined
Jul 23, 2006
Messages
27,670 (5.12/day)
Location
Houston
System Name All the cores || Into the Blue
Processor 2990WX || 5960X
Motherboard Asrock X399M || Asus X99M WS
Cooling CPU-XSPC RayStorm Neo, 2x240mm+360mm, D5PWM+140mL, GPU-2x360mm, 2xbyski, D4+D5+100mL || TBD
Memory 4x8GB G.Skill Trident Z 3200 CL16 || EVGA 3200 2x8GB
Video Card(s) (2) EVGA SC BLACK 1080Ti's || EVGA 1080Ti FE
Storage 2x Samsung SM951 512GB, Samsung PM961 512GB || OCZ Vector 180 480GB
Display(s) Dell UP2414Q 3840X2160@60hz || TBD
Case Caselabs Mercury S5+pedestal || Supermicro S5 (GS5A-753K)
Audio Device(s) Fischer HA-02->Fischer FA-002W High edition/FA-003/Jubilate/FA-011 depending on my mood
Power Supply Seasonic Prime 1200w || Seasonic Snow Silent 750w
Mouse Thermaltake Theron, Steam controller || TBD
Keyboard Thermaltake Poseidon ZX, Steam controller || Corsair K70
Software W10P || W10P
@cdawall
Why do you think it would be "Fury replacement"? (at stock)
I could see 300$ (1500+) version going there, but don't 1266Mhz leaks show it somewhere at 980 (stock), and even slightly below it?

When did I say it replaced a fury? It should compete with a nano on the top end, but they will both be power limited minus the high end ones I feel.
 

Mussels

Moderprator
Staff member
Joined
Oct 6, 2004
Messages
49,776 (8.21/day)
Location
Australalalalalaia.
System Name Rainbow Sparkles
Processor Ryzen R7 5800X
Motherboard Asus x570 Gaming-F
Cooling EK 240mm RGB AIO | Custom 280mm EK loop
Memory 64GB DDR4 3600 Corsair Vengeance RGB @ 3800 C16
Video Card(s) Galax RTX 3090 SG 24GB (0.8v 1.8GHz) - EK ARGB block
Storage 1TB Sasmsung 970 Pro NVME + 500GB 850 Evo
Display(s) Gigabyte G32QC + Phillips 328m6fjrmb (32" 1440p 165Hz/144Hz curved )
Case Fractal Design R6
Audio Device(s) Razer Leviathan + Corsair Void pro RGB, Blue Yeti mic
Power Supply Corsair HX 750i (Platinum, fan off til 300W)
Mouse Logitech G Pro wireless + Steelseries Prisma XL
Keyboard Razer Huntsman TE
Software Windows 10 pro x64 (all systems)
Benchmark Scores Lots of RGB, so you know it's fast.
from everything i've seen its a 390(x) replacement at half the TDP.

odd/annoying that the cooler is longer than the PCB.
 
Joined
Aug 2, 2011
Messages
1,297 (0.36/day)
Processor Core i9 10940X @ 4.9GHz
Motherboard Asrock Taichi XE
Cooling Custom loop, 2x360mm radiator,Lian Li UNI, EK XRes140,EK Supremacy EVO
Memory 4x8GB G.Skill RGB 3600MHz C16
Video Card(s) Aorus RTX 2080 Ti Xtreme Waterforce WB 2050 core 8200 mem
Storage 960 EVO 1TB,XPG SX8200 Pro,Micron 1100 2TB SSD,OCZ Vector 512GB,750GB Samsung,1.5TB Caviar Green
Display(s) Acer X34P, Acer XB270HU
Case LianLi O11 Dynamic White
Audio Device(s) Corsair Void Pro RGB
Power Supply EVGA P2 850W
Mouse Logitech G502 Lightspeed
Keyboard Razer Black Widow Ultimate 2016 Edition
Software Win 10 x64
from everything i've seen its a 390(x) replacement at half the TDP.

odd/annoying that the cooler is longer than the PCB.

Fury is the exact same way though.
 
Joined
Feb 13, 2012
Messages
517 (0.15/day)
You missed the entire point of my post. Clockspeed doesn't matter. GPU's do not have cores you are correct, but do you know what an SP is, do you know how they work? Did you know they stuff as many of those into a GPU as is possible and those basically make the GPU. Something about number of transistors and those actually do the work.

So instead of arguing with the entire forum, all of whom has said you are wrong how about do some research.

Oh and just in case you missed it clockspeed doesn't matter. The GTX980 hit 2.3ghz in 2015, it also wasn't the first GPU to do so. http://www.kitguru.net/components/g...eforce-gtx-980-hits-2-30ghz-sets-new-records/
Lol ok I feel like you arent even reading my comments, or maybe you are slightly skipping through the lines. This time i will try to break things down as much as possible. There are 3 major factors in gpu production.
1- thermal headroom (dictates form factor)
2-die size (effects cost of production)
3- performance (number of cores/sps and clockspeed dictate this)

So in a gpu, engineers want to design power efficient sps to cram as many of them before reaching thermal headroom, but at the same time you want to clock them as high as you can as well in order to build reasonably priced/smaller chips that perform well. So there is always a trade off.

The question i presented here, is wether you believe clockspeed is a design decision by Amd, or is due to the variance between 14nm glofo and 16nm tsmc, id appreciate it if you stop refuting what you are assuming im trying to say(bash amd, which I'm not) and instead contribute to the conversation by answering my question.

Is amd prioratizing core count and bigger die with lower power over smaller chips with higher clocks and higher power?
And i also pointed out that I wonder if 14nm glofo having slightly higher density and possibly lower cost effected that decision from AMDs side.

Again sorry for the confusion, but you guys seriously didnt get what i was trying to say initially, but i hope you do now :)
 

cdawall

where the hell are my stars
Joined
Jul 23, 2006
Messages
27,670 (5.12/day)
Location
Houston
System Name All the cores || Into the Blue
Processor 2990WX || 5960X
Motherboard Asrock X399M || Asus X99M WS
Cooling CPU-XSPC RayStorm Neo, 2x240mm+360mm, D5PWM+140mL, GPU-2x360mm, 2xbyski, D4+D5+100mL || TBD
Memory 4x8GB G.Skill Trident Z 3200 CL16 || EVGA 3200 2x8GB
Video Card(s) (2) EVGA SC BLACK 1080Ti's || EVGA 1080Ti FE
Storage 2x Samsung SM951 512GB, Samsung PM961 512GB || OCZ Vector 180 480GB
Display(s) Dell UP2414Q 3840X2160@60hz || TBD
Case Caselabs Mercury S5+pedestal || Supermicro S5 (GS5A-753K)
Audio Device(s) Fischer HA-02->Fischer FA-002W High edition/FA-003/Jubilate/FA-011 depending on my mood
Power Supply Seasonic Prime 1200w || Seasonic Snow Silent 750w
Mouse Thermaltake Theron, Steam controller || TBD
Keyboard Thermaltake Poseidon ZX, Steam controller || Corsair K70
Software W10P || W10P
Lol ok I feel like you arent even reading my comments, or maybe you are slightly skipping through the lines. This time i will try to break things down as much as possible. There are 3 major factors in gpu production.
1- thermal headroom (dictates form factor)
2-die size (effects cost of production)
3- performance (number of cores/sps and clockspeed dictate this)

So in a gpu, engineers want to design power efficient sps to cram as many of them before reaching thermal headroom, but at the same time you want to clock them as high as you can as well in order to build reasonably priced/smaller chips that perform well. So there is always a trade off.

The question i presented here, is wether you believe clockspeed is a design decision by Amd, or is due to the variance between 14nm glofo and 16nm tsmc, id appreciate it if you stop refuting what you are assuming im trying to say(bash amd, which I'm not) and instead contribute to the conversation by answering my question.

Is amd prioratizing core count and bigger die with lower power over smaller chips with higher clocks and higher power?
And i also pointed out that I wonder if 14nm glofo having slightly higher density and possibly lower cost effected that decision from AMDs side.

Again sorry for the confusion, but you guys seriously didnt get what i was trying to say initially, but i hope you do now :)

What you have said is wrong. Claiming efficiency purely based off of clockspeed is ignorant. This tangent about sp's has never been in any of your comments and your understanding of it makes me think you watched one too many YouTube videos. Go back to Tom's hardware.
 

Mussels

Moderprator
Staff member
Joined
Oct 6, 2004
Messages
49,776 (8.21/day)
Location
Australalalalalaia.
System Name Rainbow Sparkles
Processor Ryzen R7 5800X
Motherboard Asus x570 Gaming-F
Cooling EK 240mm RGB AIO | Custom 280mm EK loop
Memory 64GB DDR4 3600 Corsair Vengeance RGB @ 3800 C16
Video Card(s) Galax RTX 3090 SG 24GB (0.8v 1.8GHz) - EK ARGB block
Storage 1TB Sasmsung 970 Pro NVME + 500GB 850 Evo
Display(s) Gigabyte G32QC + Phillips 328m6fjrmb (32" 1440p 165Hz/144Hz curved )
Case Fractal Design R6
Audio Device(s) Razer Leviathan + Corsair Void pro RGB, Blue Yeti mic
Power Supply Corsair HX 750i (Platinum, fan off til 300W)
Mouse Logitech G Pro wireless + Steelseries Prisma XL
Keyboard Razer Huntsman TE
Software Windows 10 pro x64 (all systems)
Benchmark Scores Lots of RGB, so you know it's fast.
sergio: you're comparing apples to oranges and arguing over which is better because it has more blue.

clockspeed alone means nothing, you cannot use it as a reference between different models of the same GPU, let alone different GPU designs.
 
Joined
Feb 13, 2012
Messages
517 (0.15/day)
What you have said is wrong. Claiming efficiency purely based off of clockspeed is ignorant. This tangent about sp's has never been in any of your comments and your understanding of it makes me think you watched one too many YouTube videos. Go back to Tom's hardware.
Because previously i used the generic term cores rather than sp's. i only pointed out clock speed and that nvidia went mad on the clocks, i assumed people here would get what i meant, instead it seems people including yourself are butthurt, and even after saying sorry for the misunderstanding and explaining in details what i was going for you still wanna go on and on about technicalities and what i said and what i didnt say rather than moving on and having a mature conversation.
 
Last edited:
Joined
Jul 9, 2015
Messages
2,804 (1.32/day)
System Name My all round PC
Processor i5 750
Motherboard ASUS P7P55D-E
Memory 8GB
Video Card(s) Sapphire 380 OC... sold, waiting for Navi
Storage 256GB Samsung SSD + 2Tb + 1.5Tb
Display(s) Samsung 40" A650 TV
Case Thermaltake Chaser mk-I Tower
Power Supply 425w Enermax MODU 82+
Software Windows 10
And i also pointed out that I wonder if 14nm glofo having slightly higher density and possibly lower cost effected that decision from AMDs side.
Note that AMD crammed in more transistors into the same die area than Nvidia even when on the same node.
More things at lower clock or less at higher clocks is a design decision, with neither way being clearly "superior".

If RX 480 OCing rumors are true, AMD has clearly "first chips released" round, with 15-20%-ish boost from OC, vs 10-14%-ish from nVidia.
 
Joined
Feb 13, 2012
Messages
517 (0.15/day)
Note that AMD crammed in more transistors into the same die area than Nvidia even when on the same node.
More things at lower clock or less at higher clocks is a design decision, with neither way being clearly "superior".

If RX 480 OCing rumors are true, AMD has clearly "first chips released" round, with 15-20%-ish boost from OC, vs 10-14%-ish from nVidia.
Yeah that makes sense, and having more oc headroom is good and was clear from your original post, but idk if its just me but i still feel like amd is clocking these chips conservatively even with the density vs clocks thing you mentioned, probably to make validation possible on a higher number of chips, but on later revisions snd probably rebrands im sure the clock will be much higher and somewhat closer to the clocks nvidia is reaching. We saw this with the hd7970 when it was released at 925mhz, then a ghz edition at 1050mhz was released on the next revision. One thing i know for sure though, i cant wait for the official reviews! If my educated guesses and assumptions are in their place, i suspect polaris 10 at its mid/end of its lifecycle and rebrands will be clocked and perform similar or close to gtx1070, but initially with current clocks and somewhat early silicon it wont come as close
 

cdawall

where the hell are my stars
Joined
Jul 23, 2006
Messages
27,670 (5.12/day)
Location
Houston
System Name All the cores || Into the Blue
Processor 2990WX || 5960X
Motherboard Asrock X399M || Asus X99M WS
Cooling CPU-XSPC RayStorm Neo, 2x240mm+360mm, D5PWM+140mL, GPU-2x360mm, 2xbyski, D4+D5+100mL || TBD
Memory 4x8GB G.Skill Trident Z 3200 CL16 || EVGA 3200 2x8GB
Video Card(s) (2) EVGA SC BLACK 1080Ti's || EVGA 1080Ti FE
Storage 2x Samsung SM951 512GB, Samsung PM961 512GB || OCZ Vector 180 480GB
Display(s) Dell UP2414Q 3840X2160@60hz || TBD
Case Caselabs Mercury S5+pedestal || Supermicro S5 (GS5A-753K)
Audio Device(s) Fischer HA-02->Fischer FA-002W High edition/FA-003/Jubilate/FA-011 depending on my mood
Power Supply Seasonic Prime 1200w || Seasonic Snow Silent 750w
Mouse Thermaltake Theron, Steam controller || TBD
Keyboard Thermaltake Poseidon ZX, Steam controller || Corsair K70
Software W10P || W10P
Because previously i used the generic term cores rather than sp's. i only pointed out clock speed and that nvidia went mad on the clocks, i assumed people here would get what i meant, instead it seems people including yourself are butthurt, and even after saying sorry for the misunderstanding and explaining in details what i was going for you still wanna go on and on about technicalities and what i said and what i didnt say rather than moving on and having a mature conversation.

No one is butt hurt everyone is scratching their collective heads trying to figure out why you think the ocean is full of soda.
 
Joined
Apr 16, 2015
Messages
305 (0.14/day)
System Name 2016
Processor intel i7-6700K
Motherboard ASUS Maximus VIII Impact
Cooling Noctua NH-D14
Memory G.Skill Ripjaws V DDR4 2x16GB 3200MHz
Video Card(s) Gigabyte GTX 1070 mini (3rd from RMA; 2 burned)
Storage Samsung EVO 850 + 24TB NAS
Display(s) AOC 31.5" ips freesync; Oculus Rift CV1
Case Raijintek Metis Classic Black
Power Supply Silverstone SX500-LG (with Scythe fan!!)
Mouse Aula Killing The Soul
Keyboard Logitech K290
Software Windows 10 Pro
Benchmark Scores upgrade: www.3dmark.com/compare/fs/9798108/fs/5598299 www.3dmark.com/compare/spy/285676/spy/64550
Why are you so illogical? He said top spot. With no reference to anything else. Top spot. Not middle top. Bottom top. Top.

Sooooooooooooooooooo yeah.

Maybe Radeon Pro would be Top Spot but with occasional profile absences, the average would possibly drag it below GTX 1080.

top spot at populatiry, like GTX 970 is right now (which isn't the king of performance either... just most popular; best seller). http://store.steampowered.com/hwsurvey
 

Aquinus

Resident Wat-man
Joined
Jan 28, 2012
Messages
12,093 (3.57/day)
Location
Concord, NH
System Name Apollo
Processor Intel Core i9 9880H
Memory 64GB DDR4-2667
Video Card(s) AMD Radeon Pro 5600M, 8GB HBM2
Storage 1TB Apple NVMe, 4TB External
Display(s) Laptop @ 3072x1920 + 2x LG 5k Ultrafine TB3 displays
Case MacBook Pro (16", 2019)
Audio Device(s) AirPods Pro, Sennheiser HD 380s w/ FIIO Alpen 2, or Logitech 2.1 Speakers
Power Supply 96w Power Adapter
Mouse Logitech MX Master 3
Keyboard Full Size Wireless Apple Magic Keyboard
Software MacOS 10.15.7
I think there is some confusion about how GPUs work, so maybe a little bit of explanation is in order because everyone is acting like the number of shaders are the things that can do independent computations and this is not a true statement. Both nVidia and AMD have terms for what constitutes a full "core". For nVidia that's a "Streaming Multiprocessor" or SM and for AMD it's a "Compute Unit" or CU. Each CU/SM has all of the inner workings to do independent calculations however, it's usually these units that dispatch more fine-grained computations. So, for regular CPUs, it's just having cores inside of a package. For GPUs, it's like having very simple cores which are put inside more CPU-like complex cores, which are then put into the package.

Just like regular CPUs, there is a limit to how much parallel work can be accomplished on a GPU. There is an argument that could be made that despite similar shader counts, the fact that nVidia used fewer SMs clocked higher yielded a better benefit than more CUs clocked lower (also with fewer shaders per CU.)

So with that said, I think there is a lot more that goes into this than clock speed, or even hardware for that matter. How the drivers and applications dispatch any instructions to the GPU has just as much to do with it as the way the hardware is setup... but simply put, the only time you can boil everything down to clock speed is on the same architecture.

tl;dr: GPUs do have "cores", they're just a little different than traditional CPU cores and they're not the shaders though, they're the CUs and SMs.
 
Joined
Oct 22, 2014
Messages
11,869 (4.96/day)
Location
Sunshine Coast
System Name Black Box
Processor Intel i5-9600KF
Motherboard NZXT N7 Z370 Black
Cooling Cooler Master 240 RGB AIO / Stock
Memory Thermaltake Toughram 16GB 4400MHz DDR4 or Gigabyte 16GB 3600MHz DDR4 or Adata 8GB 2133Mhz DDR4
Video Card(s) Asus Dual 1060 6GB
Storage Kingston A2000 512Gb NVME
Display(s) AOC 24" Freesync 1m.s. 75Hz
Case Corsair 450D High Air Flow.
Audio Device(s) No need.
Power Supply FSP Aurum 650W
Mouse Yes
Keyboard Of course
Software W10 Pro 64 bit
"Naked" 480, not sure why it is a thing, but here we go:

Just wondering, the picture shows the fins on the heatsink in a vertical alignment, but shouldn't they mount 90 degrees to that for air flow to have a clean path out the rear vents?
 

Mussels

Moderprator
Staff member
Joined
Oct 6, 2004
Messages
49,776 (8.21/day)
Location
Australalalalalaia.
System Name Rainbow Sparkles
Processor Ryzen R7 5800X
Motherboard Asus x570 Gaming-F
Cooling EK 240mm RGB AIO | Custom 280mm EK loop
Memory 64GB DDR4 3600 Corsair Vengeance RGB @ 3800 C16
Video Card(s) Galax RTX 3090 SG 24GB (0.8v 1.8GHz) - EK ARGB block
Storage 1TB Sasmsung 970 Pro NVME + 500GB 850 Evo
Display(s) Gigabyte G32QC + Phillips 328m6fjrmb (32" 1440p 165Hz/144Hz curved )
Case Fractal Design R6
Audio Device(s) Razer Leviathan + Corsair Void pro RGB, Blue Yeti mic
Power Supply Corsair HX 750i (Platinum, fan off til 300W)
Mouse Logitech G Pro wireless + Steelseries Prisma XL
Keyboard Razer Huntsman TE
Software Windows 10 pro x64 (all systems)
Benchmark Scores Lots of RGB, so you know it's fast.
to me, based on the base of that heatsink it needs to be rotated 90 degrees to the right - then the fins are horizontal, and it looks like it all lines up.
 
Joined
Oct 22, 2014
Messages
11,869 (4.96/day)
Location
Sunshine Coast
System Name Black Box
Processor Intel i5-9600KF
Motherboard NZXT N7 Z370 Black
Cooling Cooler Master 240 RGB AIO / Stock
Memory Thermaltake Toughram 16GB 4400MHz DDR4 or Gigabyte 16GB 3600MHz DDR4 or Adata 8GB 2133Mhz DDR4
Video Card(s) Asus Dual 1060 6GB
Storage Kingston A2000 512Gb NVME
Display(s) AOC 24" Freesync 1m.s. 75Hz
Case Corsair 450D High Air Flow.
Audio Device(s) No need.
Power Supply FSP Aurum 650W
Mouse Yes
Keyboard Of course
Software W10 Pro 64 bit
to me, based on the base of that heatsink it needs to be rotated 90 degrees to the right - then the fins are horizontal, and it looks like it all lines up.
That's what I thought, but the mounting holes look like they wont line up that way.
Maybe that is an illusion.
 
Joined
Feb 13, 2012
Messages
517 (0.15/day)
I think there is some confusion about how GPUs work, so maybe a little bit of explanation is in order because everyone is acting like the number of shaders are the things that can do independent computations and this is not a true statement. Both nVidia and AMD have terms for what constitutes a full "core". For nVidia that's a "Streaming Multiprocessor" or SM and for AMD it's a "Compute Unit" or CU. Each CU/SM has all of the inner workings to do independent calculations however, it's usually these units that dispatch more fine-grained computations. So, for regular CPUs, it's just having cores inside of a package. For GPUs, it's like having very simple cores which are put inside more CPU-like complex cores, which are then put into the package.

Just like regular CPUs, there is a limit to how much parallel work can be accomplished on a GPU. There is an argument that could be made that despite similar shader counts, the fact that nVidia used fewer SMs clocked higher yielded a better benefit than more CUs clocked lower (also with fewer shaders per CU.)

So with that said, I think there is a lot more that goes into this than clock speed, or even hardware for that matter. How the drivers and applications dispatch any instructions to the GPU has just as much to do with it as the way the hardware is setup... but simply put, the only time you can boil everything down to clock speed is on the same architecture.

tl;dr: GPUs do have "cores", they're just a little different than traditional CPU cores and they're not the shaders though, they're the CUs and SMs.
Well said, i personally didnt intend to get into sucj depth though and just used the term "cores" whereas in a cpu it was a handful while in a gpu its multi parallelism, and what i was saying the relationships between clockspeed to performance isnt the same as in cpus because in a cpu single core performance is key. In a gpu as i mentioned the big constrain is thermal headroom and die size, and you want to keep a good balance hence why i had all these question marks and everyone lost it. I said nvidia went mad on clocks and they surely did, yes you cant compare clockspeed directly between different architectures but i mean this is an over 30% clock difference between amd and nvidia, and this is kinda on similar nodes, so idk if im missing something but i dont remember the clockspeed gap ever being this wide to be honest, correct me if im wrong tho. I did point out 14nm vs 16nm so technicaly this is the first time amd and nvidia use different processes, still i though they should be similar for the most part
 

cdawall

where the hell are my stars
Joined
Jul 23, 2006
Messages
27,670 (5.12/day)
Location
Houston
System Name All the cores || Into the Blue
Processor 2990WX || 5960X
Motherboard Asrock X399M || Asus X99M WS
Cooling CPU-XSPC RayStorm Neo, 2x240mm+360mm, D5PWM+140mL, GPU-2x360mm, 2xbyski, D4+D5+100mL || TBD
Memory 4x8GB G.Skill Trident Z 3200 CL16 || EVGA 3200 2x8GB
Video Card(s) (2) EVGA SC BLACK 1080Ti's || EVGA 1080Ti FE
Storage 2x Samsung SM951 512GB, Samsung PM961 512GB || OCZ Vector 180 480GB
Display(s) Dell UP2414Q 3840X2160@60hz || TBD
Case Caselabs Mercury S5+pedestal || Supermicro S5 (GS5A-753K)
Audio Device(s) Fischer HA-02->Fischer FA-002W High edition/FA-003/Jubilate/FA-011 depending on my mood
Power Supply Seasonic Prime 1200w || Seasonic Snow Silent 750w
Mouse Thermaltake Theron, Steam controller || TBD
Keyboard Thermaltake Poseidon ZX, Steam controller || Corsair K70
Software W10P || W10P
Well said, i personally didnt intend to get into sucj depth though and just used the term "cores" whereas in a cpu it was a handful while in a gpu its multi parallelism, and what i was saying the relationships between clockspeed to performance isnt the same as in cpus because in a cpu single core performance is key. In a gpu as i mentioned the big constrain is thermal headroom and die size, and you want to keep a good balance hence why i had all these question marks and everyone lost it. I said nvidia went mad on clocks and they surely did, yes you cant compare clockspeed directly between different architectures but i mean this is an over 30% clock difference between amd and nvidia, and this is kinda on similar nodes, so idk if im missing something but i dont remember the clockspeed gap ever being this wide to be honest, correct me if im wrong tho. I did point out 14nm vs 16nm so technicaly this is the first time amd and nvidia use different processes, still i though they should be similar for the most part

They aren't similar in anyway. Just because two companies use 65nm or 14/16nm doesn't mean anything is the same internally....
 

Aquinus

Resident Wat-man
Joined
Jan 28, 2012
Messages
12,093 (3.57/day)
Location
Concord, NH
System Name Apollo
Processor Intel Core i9 9880H
Memory 64GB DDR4-2667
Video Card(s) AMD Radeon Pro 5600M, 8GB HBM2
Storage 1TB Apple NVMe, 4TB External
Display(s) Laptop @ 3072x1920 + 2x LG 5k Ultrafine TB3 displays
Case MacBook Pro (16", 2019)
Audio Device(s) AirPods Pro, Sennheiser HD 380s w/ FIIO Alpen 2, or Logitech 2.1 Speakers
Power Supply 96w Power Adapter
Mouse Logitech MX Master 3
Keyboard Full Size Wireless Apple Magic Keyboard
Software MacOS 10.15.7
They aren't similar in anyway. Just because two companies use 65nm or 14/16nm doesn't mean anything is the same internally....
They're a lot more similar than you think. Sure, there may be some implementation differences but when push comes to shove, they're doing it the same way. When push comes to shove, it really comes down to nVidia opting for lower latency and low parallel throughput compared to AMD which opts for higher overall latency but, more parallel throughput. The amount of performance increasing clocks for any given design will almost always result in performance improvements however, increasing the width of the engine (such as increasing the number of shaders/SMs/CUs,) has limited benefit depending on the workload and how that workload was dispatched.

Either way, all of this is besides the point. I don't think anyone really understands what @sergionography is trying to say and all I'm saying is this:
sergio: you're comparing apples to oranges and arguing over which is better because it has more blue.

clockspeed alone means nothing, you cannot use it as a reference between different models of the same GPU, let alone different GPU designs.
 

cdawall

where the hell are my stars
Joined
Jul 23, 2006
Messages
27,670 (5.12/day)
Location
Houston
System Name All the cores || Into the Blue
Processor 2990WX || 5960X
Motherboard Asrock X399M || Asus X99M WS
Cooling CPU-XSPC RayStorm Neo, 2x240mm+360mm, D5PWM+140mL, GPU-2x360mm, 2xbyski, D4+D5+100mL || TBD
Memory 4x8GB G.Skill Trident Z 3200 CL16 || EVGA 3200 2x8GB
Video Card(s) (2) EVGA SC BLACK 1080Ti's || EVGA 1080Ti FE
Storage 2x Samsung SM951 512GB, Samsung PM961 512GB || OCZ Vector 180 480GB
Display(s) Dell UP2414Q 3840X2160@60hz || TBD
Case Caselabs Mercury S5+pedestal || Supermicro S5 (GS5A-753K)
Audio Device(s) Fischer HA-02->Fischer FA-002W High edition/FA-003/Jubilate/FA-011 depending on my mood
Power Supply Seasonic Prime 1200w || Seasonic Snow Silent 750w
Mouse Thermaltake Theron, Steam controller || TBD
Keyboard Thermaltake Poseidon ZX, Steam controller || Corsair K70
Software W10P || W10P
They're a lot more similar than you think. Sure, there may be some implementation differences but when push comes to shove, they're doing it the same way. When push comes to shove, it really comes down to nVidia opting for lower latency and low parallel throughput compared to AMD which opts for higher overall latency but, more parallel throughput. The amount of performance increasing clocks for any given design will almost always result in performance improvements however, increasing the width of the engine (such as increasing the number of shaders/SMs/CUs,) has limited benefit depending on the workload and how that workload was dispatched.

Either way, all of this is besides the point. I don't think anyone really understands what @sergionography is trying to say and all I'm saying is this:

Same argument could be made with AMD vs intel CPU's...AMD uses a different everything for everything.

Oh well at the end of the day I just got a 980Ti for my PC at work so I am going to enjoy that



Lol because I so need a 980Ti to run web browsers.
 
Joined
Feb 8, 2012
Messages
2,978 (0.88/day)
Location
Zagreb, Croatia
System Name Windows 10 64-bit Core i7 6700
Processor Intel Core i7 6700
Motherboard Asus Z170M-PLUS
Cooling Corsair AIO
Memory 2 x 8 GB Kingston DDR4 2666
Video Card(s) Gigabyte NVIDIA GeForce GTX 1060 6GB
Storage Western Digital Caviar Blue 1 TB, Seagate Baracuda 1 TB
Display(s) Dell P2414H
Case Corsair Carbide Air 540
Audio Device(s) Realtek HD Audio
Power Supply Corsair TX v2 650W
Mouse Steelseries Sensei
Keyboard CM Storm Quickfire Pro, Cherry MX Reds
Software MS Windows 10 Pro 64-bit
The amount of performance increasing clocks for any given design will almost always result in performance improvements however, increasing the width of the engine (such as increasing the number of shaders/SMs/CUs,) has limited benefit depending on the workload and how that workload was dispatched.
You don't need that "almost" at all, unless you mean increasing the clocks beyond stability ... increasing parallelism in hardware must be harnessed by equally parallelized software, luckily all graphics processing is very SIMD parallelized by definition (same shader code being run for many pixels and vertices) ... so higher clocks and more throughput are both valid strategies and both are used - so far AMD and Nvidia are finding a sweet spot power efficiency wise for what they think gaming workload will consist of in next year or so (balancing fill rate vs. compute/shading and raw throughput vs. caching in their architectures)
 
Joined
Jun 15, 2006
Messages
129 (0.02/day)
I bet somebody will overclock/overvolt RX480 to near 50%, that will par with 1070 easily. Looking at the performance graph and clock speed, I'm sure it can be done.
 
Top