Kepler performance gimping myth.

cucker tarlson · Mar 17, 2019

AmioriK · Mar 17, 2019

I don't think it's a deliberate"gimping" thing. It's more to do with the Kepler SMX being underutilized due to uArch design choices affecting performance in some (not all) newer titles. And VRAM constraints.

GorbazTheDragon · Mar 17, 2019

Kinda feel like this is beating a dead horse....

Naito · Mar 17, 2019

This post really needs a quick summary of the article and some figures, not just links to off-site...

Shambles1980 · Mar 17, 2019

yeah i'm not clicking on a bunch of links to find out what you want to tell me.
If you want to use the links to back up what you want to tell me that's fine, but right now all i see is links.

Naito · Mar 17, 2019

Just to be more constructive; I think it is interesting to see the performance improvements gained from drivers overtime, but a small breakdown of some sort would be beneficial.

For example, pick one or two of the more interesting comparisons or games your prefer and present the data as such:

Crysis 3 @1080p SMAA x2
Latest (418.81) - Min: 50 Avg: 58.1
Earliest (320.18) - Min: 48 Avg: 55.4
Worse (353.62) - Min: 48 Avg: 55.3
Difference - Min: 2 Avg: 2.8

Or something akin to the above. Don't have to do it for every game and then refer everyone to the links for more information, if they would like to see more.

EarthDog · Mar 17, 2019

Right.. is there a Cliff's notes? Not clicking through a half dozen+ Polish links for the high level answer.

I thank you cucker, for beating the dead horse

, but maybe a summary and some links will get people discussing this subject (again).

Edit: geez... the last page even summarizes it... for the gtx 780 in the games tested, there was a 7% increase (average sans fc primal) over time for release to their last driver tested.
https://ithardware.pl/testyirecenzj...ych_kart_graficznych_odc_3_kepler-8591-8.html

Naito · Mar 17, 2019

EarthDog said:
for the gtx 780 in the games tested, there was a 7% increase (average sans fc primal) over time for release to their last driver tested.

Not bad at all, if you think about it. Add on top of that say a 10% performance gain from an OC and you have a card that's on average 17% faster than on the day it was purchased (assuming it was purchased at release).

EarthDog · Mar 17, 2019

Naito said:
Not bad at all, if you think about it. Add on top of that say a 10% performance gain from an OC and you have a card that's on average 17% faster than on the day it was purchased.

Sure.. but this is pretty OFN to be honest. Many had it in their heads that there was gimpage across the board, but, that wasn't true. Sure a title here or there may show something like that, but overall it stays the same or increases, typically.

This thread is just tossing updated fuel on some embers left over, lol

cucker tarlson · Mar 17, 2019

Naito said:
Not bad at all, if you think about it. Add on top of that say a 10% performance gain from an OC and you have a card that's on average 17% faster than on the day it was purchased.

lol,what ? wasn't oc available from day one ?

Naito · Mar 17, 2019

cucker tarlson said:
wasn't oc available from day one ?

Yes, but I'm talking free perfomance gains over vendor/factory specifications and release drivers here

64K · Mar 17, 2019

It's been five and a half years since the last Kepler GPU was released. You can't count the 750 and 750 Ti because they were the first Maxwells. Version 1 to be exact. The 9xx were version 2. It's just more of the confusion that Nvidia created with the Kepler series.

Anyway, the bulk of the improvements for Keplers was done long ago and there are people that claim that Nvidia is gimping the performance of the last generation each time a new generation is released. I doubt they are but for people to expect that Nvidia can keep improving the drivers for Kepler to keep up is not realistic. They've already dropped support entirely for Fermis and it won't be long before Keplers fall by the wayside too.

EarthDog · Mar 17, 2019

Naito said:
Yes, but I'm talking free perfomance gains over vendor/factory specifications and release drivers here

Overclocking has little to do with driver date. Some allow for slightly better or worse clocking, but are not associated otherwise. You seem to infer there is with your mention(s).

rtwjunkie · Mar 17, 2019

We had this also extensively tested 6 months ago on our own forums. The result was the gimping is a conspiracy-theory fanboy myth.

Any improvements by drivers already have eked out any performance gains long ago from Kepler. Indeed, it is likely so for Maxwell too. Of course it is going to seem as if performance falls behind when no improvements are being written for a specific GPU family any longer.

Naito · Mar 17, 2019

EarthDog said:
Overclocking has little to do with driver date. Some allow for slightly better or worse clocking, but are not associated otherwise. You seem to infer there is with your mention(s).

Perhaps the point is moot and poorly worded, but I was speaking of an OC result outside of any driver influence. I was just trying to get at the 'free' performance upgrade someone can gain from a card whether it's a day one OC or incremental FPS improvements over several driver revisions

Fluffmeister · Mar 18, 2019

It's always been a myth, Kepler just happened to be the start of when their XX104 tier chips competed with AMD's high-end, and they have be livid ever since.

AmioriK · Mar 18, 2019

I just wanted to say that there are some anomalies with Kepler performance in modern titles. I have seen HD7970 trading blows with/ beating 780ti in a game but I can't remember what it is and I'm in bed half asleep so CBA to trawl the internet.

It's not deliberate gimping for sure, but I do believe that there are some uArch weaknesses with Kepler that result in underperforming in certain games.

The Kepler SMX is a big and complex beast that's apparently a lot harder to keep fed than the Maxwell SM, which focuses on higher granularity and fixed quads inside the SM that are equal in size to a 32 thread warp resulting in simpler scheduling and higher utilisation. Mh...

king of swag187 · Mar 18, 2019

Its not really gimping, its more of a lack of focus on newer drivers being optimized for Kepler's frankly dated arch. Its not a loss of performance per say, but more a loss of performance compared to what you would get with more optimized drivers

mcraygsx · Mar 20, 2019

king of swag187 said:
Its not really gimping, its more of a lack of focus on newer drivers being optimized for Kepler's frankly dated arch. Its not a loss of performance per say, but more a loss of performance compared to what you would get with more optimized drivers

+1

It's not gimping Kepler, it's merely Neglect. Game Ready drivers only apply to NV's latest GPUs basically. However, the end result is the same, in that older Kepler gets left behind via neglect. While AMD Drivers's are opposite of this.

Caring1 · Mar 20, 2019

If FPS drop in a game that previously had good FPS after switching to the latest drivers, then it is gimping.
Drivers are meant to be improved upon, each iteration building on the previous.
People talk about proof, but I haven't seen any valid proof yet, just talk claiming there is no gimping.

rtwjunkie · Mar 21, 2019

Caring1 said:
If FPS drop in a game that previously had good FPS after switching to the latest drivers, then it is gimping.
Drivers are meant to be improved upon, each iteration building on the previous.
People talk about proof, but I haven't seen any valid proof yet, just talk claiming there is no gimping.

You can’t get blood from a stone. Each new driver squeezes a little more performance from the hardware until there isn’t any more to get.

EarthDog · Mar 21, 2019

rtwjunkie said:
until there isn’t any more to get.

and/or they simply stop and move on.

I don't see where any regressed in the testing (FPS drop he's talking about), so I am not sure what Caring1 is missing here about not seeing valid proof. In this test, not one title lost performance.

John Naylor · Mar 21, 2019

Im not sure that anything is gained by rehashing this topic every 3-4 months ... the last one i thought perty much put it to bed once and for all.

https://www.techpowerup.com/forums/...-353-62-vs-376-33-vs-398-36-vs-411-70.248256/

hat · Mar 21, 2019

EarthDog said:
Right.. is there a Cliff's notes? Not clicking through a half dozen+ Polish links for the high level answer.

I thank you cucker, for beating the dead horse , but maybe a summary and some links will get people discussing this subject (again).

Edit: geez... the last page even summarizes it... for the gtx 780 in the games tested, there was a 7% increase (average sans fc primal) over time for release to their last driver tested.
https://ithardware.pl/testyirecenzj...ych_kart_graficznych_odc_3_kepler-8591-8.html

nVidia FineWIne confirmed

System Name	Purple rain
Processor	10.5 thousand 4.2G 1.1v
Motherboard	Zee 490 Aorus Elite
Cooling	Noctua D15S
Memory	16GB 4133 CL16-16-16-31 Viper Steel
Video Card(s)	RTX 2070 Super Gaming X Trio
Storage	SU900 128,8200Pro 1TB,850 Pro 512+256+256,860 Evo 500,XPG950 480, Skyhawk 2TB
Display(s)	Acer XB241YU+Dell S2716DG
Case	P600S Silent w. Alpenfohn wing boost 3 ARGBT+ fans
Audio Device(s)	K612 Pro w. FiiO E10k DAC,W830BT wireless
Power Supply	Superflower Leadex Gold 850W
Mouse	G903 lightspeed+powerplay,G403 wireless + Steelseries DeX + Roccat rest
Keyboard	HyperX Alloy SilverSpeed (w.HyperX wrist rest),Razer Deathstalker
Software	Windows 10
Benchmark Scores	A LOT

System Name	Not an Intel Piece of Shite
Processor	Superior AMD Glorious Master Race 2700SEX
Motherboard	Glorious low cost Awesome Motherboard 4
Cooling	A piece of metal that cools the amazing Ryzen CPU
Memory	SAMMY BEEE DAI BABEH
Video Card(s)	Turding
Storage	irelevant
Display(s)	monitor
Case	It's red because AMD = red and AMD = awesome
Power Supply	1000W,. but not needed as Glorious RYZEN CPU is extremely afficient unlike that recylced 14nm++ Junk
Mouse	gets cat
Keyboard	RUHGUBUH!
Software	Not Linux
Benchmark Scores	Higher than Intel shite

System Name	Indis the Fair (cursed edition)
Processor	11900k 5.1/4.9 undervolted.
Motherboard	MSI Z590 Unify-X
Cooling	Heatkiller VI Pro, VPP755 V.3, XSPC TX360 slim radiator, 3xA12x25, 4x Arctic P14 case fans
Memory	G.Skill Ripjaws V 2x16GB 4000 16-19-19 (b-die@3600 14-14-14 1.45v)
Video Card(s)	EVGA 2080 Super Hybrid (T30-120 fan)
Storage	970EVO 1TB, 660p 1TB, WD Blue 3D 1TB, Sandisk Ultra 3D 2TB
Display(s)	BenQ XL2546K, Dell P2417H
Case	FD Define 7
Audio Device(s)	DT770 Pro, Topping A50, Focusrite Scarlett 2i2, Røde VXLR+, Modmic 5
Power Supply	Seasonic 860w Platinum
Mouse	Razer Viper Mini, Odin Infinity mousepad
Keyboard	GMMK Fullsize v2 (Boba U4Ts)
Software	Win10 x64/Win7 x64/Ubuntu

System Name	Desktop \| Laptop
Processor	AMD Ryzen 7 5800X3D \| Intel Core i7 7700HQ
Motherboard	MAG X570S Torpedo Max\| Neptune KLS HM175
Cooling	Corsair H100x \| Twin fan, fin stack & heat pipes
Memory	32GB G.Skill F4-3600C16-8GVK @ 3600MHz / 16-16-16-36-1T \| 16GB DDR4 @ 2400MHz / 17-17-17-39-2T
Video Card(s)	EVGA RTX 3080 Ti FTW3 Ultra \| GTX 1050 Ti 4GB
Storage	Kingston KC3000 1TB + Kingston KC3000 2TB + Samsung 860 EVO 1TB \| 970 Evo 500GB
Display(s)	32" Dell G3223Q (2160p @ 144Hz) \| 17" IPS 1920x1080P
Case	Fractal Meshify 2 Compact \| Aspire V Nitro BE
Audio Device(s)	ifi Audio ZEN DAC V2 + Focal Radiance / HyperX Solocast
Power Supply	Super Flower Leadex V Platinum Pro 1000W \| 150W
Mouse	Razer Viper Ultimate \| Logitech MX Anywhere 2
Keyboard	Razer Huntsman V2 Optical (Linear Red)
Software	Windows 11 Pro x64

System Name	Sham Pc
Processor	i5-2500k @ 4.33
Motherboard	INTEL DZ77SL 50K
Cooling	2 bay res. "2L of fluid in loop" 1x480 2x360
Memory	16gb 4x4 kingstone 1600 hyper x fury black
Video Card(s)	hfa2 gtx 780 @ 1306/1768 (xspc bloc)
Storage	1tb wd red 120gb kingston on the way os, 1.5Tb wd black, 3tb random WD rebrand
Display(s)	cibox something or other 23" 1080p " 23 inch downstairs. 52 inch plasma downstairs 15" tft kitchen
Case	900D
Audio Device(s)	on board
Power Supply	xion gaming seriese 1000W (non modular) 80+ bronze
Software	windows 10 pro x64

Kepler performance gimping myth.

cucker tarlson

AmioriK

GorbazTheDragon

Naito

Shambles1980

Naito

EarthDog

Naito

EarthDog

cucker tarlson

Naito

64K

EarthDog

rtwjunkie

PC Gaming Enthusiast

Naito

Fluffmeister

AmioriK

king of swag187

mcraygsx

Caring1

rtwjunkie

PC Gaming Enthusiast

EarthDog

John Naylor

hat

Enthusiast

Processor	i7 7700k
Motherboard	MSI Z270 SLI Plus
Cooling	CM Hyper 212 EVO
Memory	2 x 8 GB Corsair Vengeance
Video Card(s)	MSI RTX 2070 Super
Storage	Samsung 850 EVO 250 GB and WD Black 4TB
Display(s)	Dell 27 inch 1440p 144 Hz
Case	Corsair Obsidian 750D Airflow Edition
Audio Device(s)	Onboard
Power Supply	EVGA SuperNova 850 W Gold
Mouse	Logitech G502
Keyboard	Logitech G105
Software	Windows 10

System Name	Bayou Phantom
Processor	Core i7-8700k 4.4Ghz @ 1.18v
Motherboard	ASRock Z390 Phantom Gaming 6
Cooling	All air: 2x140mm Fractal exhaust; 3x 140mm Cougar Intake; Enermax T40F Black CPU cooler
Memory	2x 16GB Mushkin Redline DDR-4 3200
Video Card(s)	EVGA RTX 2080 Ti Xc
Storage	1x 500 MX500 SSD; 2x 6TB WD Black; 1x 4TB WD Black; 1x400GB VelRptr; 1x 4TB WD Blue storage (eSATA)
Display(s)	HP 27q 27" IPS @ 2560 x 1440
Case	Fractal Design Define R4 Black w/Titanium front -windowed
Audio Device(s)	Soundblaster Z
Power Supply	Seasonic X-850
Mouse	Coolermaster Sentinel III (large palm grip!)
Keyboard	Logitech G610 Orion mechanical (Cherry Brown switches)
Software	Windows 10 Pro 64-bit (Start10 & Fences 3.0 installed)

Processor	AMD Ryzen 7 3700X
Motherboard	MSI MAG B550 TOMAHAWK
Cooling	AMD Wraith Prism
Memory	Team Group Dark Pro 8Pack Edition 3600Mhz CL16
Video Card(s)	NVIDIA GeForce RTX 3080 FE
Storage	Kingston A2000 1TB + Seagate HDD workhorse
Display(s)	Samsung 50" QN94A Neo QLED
Case	Antec 1200
Power Supply	Seasonic Focus GX-850
Mouse	Razer Deathadder Chroma
Keyboard	Logitech UltraX
Software	Windows 11

System Name	Bruh
Processor	10700K 5.3Ghz 1.35v\| i7 7920HQ 3.6Ghz -180Mv \|
Motherboard	Z490 TUF Wifi \| Apple QMS180 \|
Cooling	EVGA 360MM \| Laptop HS \|
Memory	DDR4 32GB 3600Mhz CL16 \| LPDDR3 16GB 2133Mhz CL20 \|
Video Card(s)	Asus ROG Strix 3080 (2100Mhz/18Ghz)\|Radeon Pro 560 (1150Mhz/1655Mhz)\|
Storage	Many SSDs, ~24TB HDD/8TB SSD
Display(s)	S2719DGF, HP Z27i, Z24n\| 1800P 15.4" + ZR30W + iPad Pro 10.5 2017
Case	NR600 \| MBP 2017 15" Silver \| MSI GE62VR \| Elite 120 Advanced
Audio Device(s)	Lol imagine caring about audio
Power Supply	850GQ \| Apple 87W USB-C \|
Mouse	Whatever I have on hand + trackpads (Lanchead TE)
Keyboard	HyperX Origins Alloy idk
Software	W10 20H2\|W10 1903 LTSC/MacOS 11
Benchmark Scores	No.

Processor	Intel 7700K 5.1Ghz (Intel advised me not to OC this CPU)
Motherboard	Asus Maximus IX Code
Cooling	Corsair Hydro H115i Platinum
Memory	48GB G.Skill TridentZ DDR4 3200 Dual Channel (2x16 & 2x8)
Video Card(s)	nVIDIA Titan XP (Overclocks like a champ but stock performance is enough)
Storage	Intel 760p 2280 2TB
Display(s)	MSI Optix MPG27CQ Black 27" 1ms 144hz
Case	Thermaltake View 71
Power Supply	EVGA SuperNova 1000 Platinum2
Mouse	Corsair M65 Pro (not recommded, I am on my second mouse with same defect)
Software	Windows 10 Enterprise 1803
Benchmark Scores	Yes I am Intel fanboy that is my benchmark score.

System Name	Black Box
Processor	Intel Xeon E3-1260L v5
Motherboard	MSI E3 KRAIT Gaming v5
Cooling	Tt tower + 120mm Tt fan
Memory	G.Skill 16GB 3600 C18
Video Card(s)	Asus GTX 970 Mini
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

System Name	Starlifter :: Dragonfly
Processor	i7 2600k 4.4GHz :: i5 10400
Motherboard	ASUS P8P67 Pro :: ASUS Prime H570-Plus
Cooling	Cryorig M9 :: Stock
Memory	4x4GB DDR3 2133 :: 2x8GB DDR4 2400
Video Card(s)	PNY GTX1070 :: Integrated UHD 630
Storage	Crucial MX500 1TB, 2x1TB Seagate RAID 0 :: Mushkin Enhanced 60GB SSD, 3x4TB Seagate HDD RAID5
Display(s)	Onn 165hz 1080p :: Acer 1080p
Case	Antec SOHO 1030B :: Old White Full Tower
Audio Device(s)	Creative X-Fi Titanium Fatal1ty Pro - Bose Companion 2 Series III :: None
Power Supply	FSP Hydro GE 550w :: EVGA Supernova 550
Software	Windows 10 Pro - Plex Server on Dragonfly
Benchmark Scores	>9000