Jump to content

Why Crossfire and SLi is no longer available?

Hunter-
 Share

Why dual-gpu era is dead since 2013-2014?For example,i miss time when HD 5970, HD 7990 or gtx 690 was present.
For me Multi GPU is fantastically useful for many, many things,is raw power in a single motherboard slot.I mean what is the reason behind discontinuing dual GPU cards?I want to understand.Since 2013 both company Radeon and GeForce stopped their production of multi gpu solution for consumers.Why they don't want to revive again this solution in present?They are focusing only on single core GPU card

Crossfire-SLi.jpg

Link to comment
Share on other sites

Link to post
Share on other sites

Games just don't make use of multiple and for the price and how much more performance you'd get you were better off getting the higher performance gpu

Link to comment
Share on other sites

Link to post
Share on other sites

Multi-GPU in games is more effort to implement than it's worth. Single cards are so fast now that we can do away with 2+ GPU setups, and so AMD and Nvidia have.

Intel HEDT and Server platform enthusiasts: Intel HEDT Xeon/i7 Megathread 

 

Big PC (disassembled pending re-rebuild) - i9 7980XE @4.4Ghz/~1.18v - EKWB Quantum-something block - EVGA X299 Dark - 4x8GB HyperX Predator DDR4 @2400Mhz - Intel ARC A770 LE - 1TB 970 Evo  + 1TB MX500 + 500GB XPG Atom30 + LG WH14NS40 + Lite-On IHAS124-14 - EVGA 1600W T2 - Corsair 750D Airflow - CPU-only custom loop w/360 slim rad + 280mm thick (60mm) rad & Noctua Redux NF-P12/P14s fans - Windows 11

 

Little PC - i7 6950X @4GHz core/3.5GHz uncore 1.2v - Heatkiller IV block - EVGA X99 Micro 2 - 4x8GB 3200MHz DDR4 - 2060 Super FE w/Corsair block - 1TB MX500 + 1TB MP34 + 500GB MX300 - EVGA 1000W G3 - ASUS AP201 - Custom loop w/360mm GTS rad - 5x Noctua iPPC NF-F12 - Windows 11 (TPM req removed)

 

Rest of the HEDT collection - Xeon X5675/EVGA X58 Classified 4-Way SLI - i7 4930K/EVGA X79 Dark 

 

Mobile/Work Devices: Late 2014 iMac 5K/2015 15" MacBook Pro (work computers) - iPhone 13 Pro Max - Apple Watch S3

 

Other Misc Devices: 2012 MacBook Pro 13" running Mac OS X Mountain Lion - Mac Pro 1,1 (2x Quad Core Xeons, 32GB DDR2, HD4870 512MB) running Mac OS X Snow Leopard - iPod Video (Gen 5.5E, swapped to 128GB SD card and running Rockbox firmware)

Link to comment
Share on other sites

Link to post
Share on other sites

Too many issues with it, too little support for it in general, too much work to maintain for a tiny user base,...

F@H
Desktop: i7-5960X 4.4GHz, Noctua NH-D14, ASUS Rampage V, 32GB, RTX3080, 2TB SX8200Pro, 2x16TB Ironwolf RAID0, Corsair HX1200, Thermaltake Overseer RX1, Samsung 4K curved 49" TV, 23" secondary, Mountain Everest Max

Mobile SFF rig: i9-9900K, Noctua NH-L9i, Asrock Z390 Phantom ITX-AC, 32GB, GTX1070, 2x1TB SX8200Pro RAID0, 2x5TB 2.5" HDD RAID0, Athena 500W Flex (Noctua fan), Custom 4.7l 3D printed case

 

Asus Zenbook UM325UA, Ryzen 7 5700u, 16GB, 1TB, OLED

 

GPD Win 2

Link to comment
Share on other sites

Link to post
Share on other sites

It was never 100% functional

Was an additional cost to implement in games for basically almost no userbase

Big development cost for a low return

Single better card always was better than 2 mid cards

Was up to the dev to implement and maintain which meant no quality guarantee in the limited games that got it

 

For productivy its been tried over and over and even there there are only a small ammount of applications that use multiple gpus

 

So the cost for the extra devtime needed for these is simply not worth it in any market

Link to comment
Share on other sites

Link to post
Share on other sites

26 minutes ago, Hunter- said:

I mean what is the reason behind discontinuing dual GPU cards?

Current GPUs use way too much power and require tons of cooling. Can you imagine a single PCB with 2x 4090s? Over 1kW peak usage and tons of heat to dissipate.

Games, the main target for those GPUs at the time, also don't play that well with mGPU since it's hard to keep everything in sync, so people just stopped caring about it.

 

For other uses tho, mGPU is still really useful, like machine learning, just plug 2 or more GPUs and have the workload distributed among them without many headaches.

28 minutes ago, Hunter- said:

,is raw power in a single motherboard slot

Get a PCIe splitter and you can have 2x x8 devices in a single x16 slot 😛 

 

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to comment
Share on other sites

Link to post
Share on other sites

16 minutes ago, igormp said:

Current GPUs use way too much power and require tons of cooling. Can you imagine a single PCB with 2x 4090s? Over 1kW peak usage and tons of heat to dissipate.

Power draw is completely irrelevant to why SLI/Crossfire was dropped. Cooling is a non-issue either, waterblocks exist.

 

My GTX 780 Classifieds will each pull over 300W, and they support 4-way SLI, so if I bought 2 more I could put those on my X58 Classified SLI 4-Way and have 1200W+ of GPU. Now imagine folks pushing these on water or LN2 for benchmark scores. Even my 2 on air with a stock X5675 hexacore pull 750W from the wall through an 80+ Titanium unit, so that's pretty close to the actual wattage draw of the components themselves.

 

The reasons SLI was dropped are the ones mentioned already, it didn't work great and single cards are now fast enough that there's no point in trying to make it work. Purely that, not because of power draw or heat or cost or really any other reason, the base tech just wasn't worth the effort as cards got faster.

Intel HEDT and Server platform enthusiasts: Intel HEDT Xeon/i7 Megathread 

 

Big PC (disassembled pending re-rebuild) - i9 7980XE @4.4Ghz/~1.18v - EKWB Quantum-something block - EVGA X299 Dark - 4x8GB HyperX Predator DDR4 @2400Mhz - Intel ARC A770 LE - 1TB 970 Evo  + 1TB MX500 + 500GB XPG Atom30 + LG WH14NS40 + Lite-On IHAS124-14 - EVGA 1600W T2 - Corsair 750D Airflow - CPU-only custom loop w/360 slim rad + 280mm thick (60mm) rad & Noctua Redux NF-P12/P14s fans - Windows 11

 

Little PC - i7 6950X @4GHz core/3.5GHz uncore 1.2v - Heatkiller IV block - EVGA X99 Micro 2 - 4x8GB 3200MHz DDR4 - 2060 Super FE w/Corsair block - 1TB MX500 + 1TB MP34 + 500GB MX300 - EVGA 1000W G3 - ASUS AP201 - Custom loop w/360mm GTS rad - 5x Noctua iPPC NF-F12 - Windows 11 (TPM req removed)

 

Rest of the HEDT collection - Xeon X5675/EVGA X58 Classified 4-Way SLI - i7 4930K/EVGA X79 Dark 

 

Mobile/Work Devices: Late 2014 iMac 5K/2015 15" MacBook Pro (work computers) - iPhone 13 Pro Max - Apple Watch S3

 

Other Misc Devices: 2012 MacBook Pro 13" running Mac OS X Mountain Lion - Mac Pro 1,1 (2x Quad Core Xeons, 32GB DDR2, HD4870 512MB) running Mac OS X Snow Leopard - iPod Video (Gen 5.5E, swapped to 128GB SD card and running Rockbox firmware)

Link to comment
Share on other sites

Link to post
Share on other sites

It's not the manufacturer's fault.

 

Game devs are forced to release games fast, thus some things are skipped in development: optimisation, mGPU compatibility, file compression. Not to mention most AAA games are developed to work in consoles.

Look at what happened to Cyberpunk, devs were pushed and couldn't optimise the game, the result: it runs like shit even if you use the highest end cards available, applying some fake resolution copium with the drivers isn't really making the game run as intended.

 

nvidia and AMD had no reason to keep driver support for that tech if no compatible games were made anymore.

 

Tandem cards and dual GPU in a same PCB are completely different things, for the second to work there's certain limitations like heatsink size and power.

Tender is the night

Link to comment
Share on other sites

Link to post
Share on other sites

46 minutes ago, Zando_ said:

Power draw is completely irrelevant to why SLI/Crossfire was dropped.

I was talking about mGPU in a single board, not mGPU as a whole, since those are still plenty used in professional scenarios.

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, igormp said:

I was talking about mGPU in a single board, not mGPU as a whole, since those are still plenty used in professional scenarios.

Ahh. Yeah, could be a struggle with 4090s, 4080s or below should be doable (they look around 300W under normal load). AMD still makes dual-GPU single-PCB workstation parts with 200W GPUs (so 400W card total) and that's running stock. Go back a bit and you have the 475W Radeon Pro Vega II Duo, back further and you have the 295x2 that's 500W stock. And overclockable, a friend had two of them and pushed 'em with dual 1200W PSUs.

Intel HEDT and Server platform enthusiasts: Intel HEDT Xeon/i7 Megathread 

 

Big PC (disassembled pending re-rebuild) - i9 7980XE @4.4Ghz/~1.18v - EKWB Quantum-something block - EVGA X299 Dark - 4x8GB HyperX Predator DDR4 @2400Mhz - Intel ARC A770 LE - 1TB 970 Evo  + 1TB MX500 + 500GB XPG Atom30 + LG WH14NS40 + Lite-On IHAS124-14 - EVGA 1600W T2 - Corsair 750D Airflow - CPU-only custom loop w/360 slim rad + 280mm thick (60mm) rad & Noctua Redux NF-P12/P14s fans - Windows 11

 

Little PC - i7 6950X @4GHz core/3.5GHz uncore 1.2v - Heatkiller IV block - EVGA X99 Micro 2 - 4x8GB 3200MHz DDR4 - 2060 Super FE w/Corsair block - 1TB MX500 + 1TB MP34 + 500GB MX300 - EVGA 1000W G3 - ASUS AP201 - Custom loop w/360mm GTS rad - 5x Noctua iPPC NF-F12 - Windows 11 (TPM req removed)

 

Rest of the HEDT collection - Xeon X5675/EVGA X58 Classified 4-Way SLI - i7 4930K/EVGA X79 Dark 

 

Mobile/Work Devices: Late 2014 iMac 5K/2015 15" MacBook Pro (work computers) - iPhone 13 Pro Max - Apple Watch S3

 

Other Misc Devices: 2012 MacBook Pro 13" running Mac OS X Mountain Lion - Mac Pro 1,1 (2x Quad Core Xeons, 32GB DDR2, HD4870 512MB) running Mac OS X Snow Leopard - iPod Video (Gen 5.5E, swapped to 128GB SD card and running Rockbox firmware)

Link to comment
Share on other sites

Link to post
Share on other sites

1 hour ago, Zando_ said:

AMD still makes dual-GPU single-PCB workstation parts with 200W GPUs (so 400W card total) and that's running stock.

Do they? I thought the pro duo was the last one, custom made for apple.

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to comment
Share on other sites

Link to post
Share on other sites

Before DX12 this was implemented by Nvidia and AMD.  DX12 puts it in the developers hands, most of which can barely even release a complete game that doesn't need months or even years of patching just to be functional.

AMD 7950x / Asus Strix B650E / 64GB @ 6000c30 / 2TB Samsung 980 Pro Heatsink 4.0x4 / 7.68TB Samsung PM9A3 / 3.84TB Samsung PM983 / 44TB Synology 1522+ / MSI Gaming Trio 4090 / EVGA G6 1000w /Thermaltake View71 / LG C1 48in OLED

Custom water loop EK Vector AM4, D5 pump, Coolstream 420 radiator

Link to comment
Share on other sites

Link to post
Share on other sites

1 hour ago, igormp said:

Do they? I thought the pro duo was the last one, custom made for apple.

Pro Vega II Duo and now a Pro W6800X Duo. 475W and 400W respectively, so 200-225W per GPU/VRAM combo. Likely limited by the cooler set up of Apple's MPX cards for noise sake, if noise weren't as much of a concern then either higher speed intake fans or a beefy axial cooler could dump more heat.

Intel HEDT and Server platform enthusiasts: Intel HEDT Xeon/i7 Megathread 

 

Big PC (disassembled pending re-rebuild) - i9 7980XE @4.4Ghz/~1.18v - EKWB Quantum-something block - EVGA X299 Dark - 4x8GB HyperX Predator DDR4 @2400Mhz - Intel ARC A770 LE - 1TB 970 Evo  + 1TB MX500 + 500GB XPG Atom30 + LG WH14NS40 + Lite-On IHAS124-14 - EVGA 1600W T2 - Corsair 750D Airflow - CPU-only custom loop w/360 slim rad + 280mm thick (60mm) rad & Noctua Redux NF-P12/P14s fans - Windows 11

 

Little PC - i7 6950X @4GHz core/3.5GHz uncore 1.2v - Heatkiller IV block - EVGA X99 Micro 2 - 4x8GB 3200MHz DDR4 - 2060 Super FE w/Corsair block - 1TB MX500 + 1TB MP34 + 500GB MX300 - EVGA 1000W G3 - ASUS AP201 - Custom loop w/360mm GTS rad - 5x Noctua iPPC NF-F12 - Windows 11 (TPM req removed)

 

Rest of the HEDT collection - Xeon X5675/EVGA X58 Classified 4-Way SLI - i7 4930K/EVGA X79 Dark 

 

Mobile/Work Devices: Late 2014 iMac 5K/2015 15" MacBook Pro (work computers) - iPhone 13 Pro Max - Apple Watch S3

 

Other Misc Devices: 2012 MacBook Pro 13" running Mac OS X Mountain Lion - Mac Pro 1,1 (2x Quad Core Xeons, 32GB DDR2, HD4870 512MB) running Mac OS X Snow Leopard - iPod Video (Gen 5.5E, swapped to 128GB SD card and running Rockbox firmware)

Link to comment
Share on other sites

Link to post
Share on other sites

5 hours ago, deadlou666 said:

Games just don't make use of multiple and for the price and how much more performance you'd get you were better off getting the higher performance gpu

I wonder if the price case would still be true if SLI were still supported by all cards, given we now live in an age of $1200 4080s.

Link to comment
Share on other sites

Link to post
Share on other sites

1 hour ago, Zando_ said:

Pro Vega II Duo and now a Pro W6800X Duo. 475W and 400W respectively, so 200-225W per GPU/VRAM combo. Likely limited by the cooler set up of Apple's MPX cards for noise sake, if noise weren't as much of a concern then either higher speed intake fans or a beefy axial cooler could dump more heat.

Didn't know about that second one, interesting. Are any of those available outside of purchase for the mac pro?

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to comment
Share on other sites

Link to post
Share on other sites

2 hours ago, igormp said:

Didn't know about that second one, interesting. Are any of those available outside of purchase for the mac pro?

AMD's page for them only shows the Mac Pro, so those are likely made just for Apple. AMD makes a lot of one-off models of hardware like that. Point was just more that there are still dual-GPU single-PCB cards in production, with pretty high wattage draw, and even higher wattage overclockable ones used to be made. So thermals/power draw wasn't the killer issue for them, and doesn't block us from making them today. Just no point for gaming, as AFR has issues and any other solution to rendering frames on 2 GPUs has even more issues, so it's not worth implementing when a single card is fast enough.

Intel HEDT and Server platform enthusiasts: Intel HEDT Xeon/i7 Megathread 

 

Big PC (disassembled pending re-rebuild) - i9 7980XE @4.4Ghz/~1.18v - EKWB Quantum-something block - EVGA X299 Dark - 4x8GB HyperX Predator DDR4 @2400Mhz - Intel ARC A770 LE - 1TB 970 Evo  + 1TB MX500 + 500GB XPG Atom30 + LG WH14NS40 + Lite-On IHAS124-14 - EVGA 1600W T2 - Corsair 750D Airflow - CPU-only custom loop w/360 slim rad + 280mm thick (60mm) rad & Noctua Redux NF-P12/P14s fans - Windows 11

 

Little PC - i7 6950X @4GHz core/3.5GHz uncore 1.2v - Heatkiller IV block - EVGA X99 Micro 2 - 4x8GB 3200MHz DDR4 - 2060 Super FE w/Corsair block - 1TB MX500 + 1TB MP34 + 500GB MX300 - EVGA 1000W G3 - ASUS AP201 - Custom loop w/360mm GTS rad - 5x Noctua iPPC NF-F12 - Windows 11 (TPM req removed)

 

Rest of the HEDT collection - Xeon X5675/EVGA X58 Classified 4-Way SLI - i7 4930K/EVGA X79 Dark 

 

Mobile/Work Devices: Late 2014 iMac 5K/2015 15" MacBook Pro (work computers) - iPhone 13 Pro Max - Apple Watch S3

 

Other Misc Devices: 2012 MacBook Pro 13" running Mac OS X Mountain Lion - Mac Pro 1,1 (2x Quad Core Xeons, 32GB DDR2, HD4870 512MB) running Mac OS X Snow Leopard - iPod Video (Gen 5.5E, swapped to 128GB SD card and running Rockbox firmware)

Link to comment
Share on other sites

Link to post
Share on other sites

Better for companies to appeal to the masses. Most people have low to mid range hardware. No reason to invest in keeping something alive if there isnt a big enough return. 
Especially see where cards are now. People have been complaining about card prices for years. They would continue to do so about stuff they coudlnt afford anyway. 

Main RIg Corsair Air 540, I7 9900k, ASUS ROG Maximus XI Hero, G.Skill Ripjaws 3600 32GB, 3090FE, EVGA 1000G5, Acer Nitro XZ3 2560 x 1440@240hz 

 

Spare RIg Lian Li O11 AIR MINI, I7 4790K, Asus Maximus VI Extreme, G.Skill Ares 2400 32Gb, EVGA 1080ti, 1080sc 1070sc & 1060 SSC, EVGA 850GA, Acer KG251Q 1920x1080@240hz

 

Link to comment
Share on other sites

Link to post
Share on other sites

1 hour ago, Zando_ said:

So thermals/power draw wasn't the killer issue for them, and doesn't block us from making them today. Just no point for gaming, as AFR has issues and any other solution to rendering frames on 2 GPUs has even more issues, so it's not worth implementing when a single card is fast enough.

But they're not a thing in other places, not only for games. Nvidia doesn't make any dual GPU in a single PCB anymore, even though there are plenty of systems of theirs with 8, 16 and even more GPUs, and I still believe that's still due to power draw and cooling limitations.

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to comment
Share on other sites

Link to post
Share on other sites

12 minutes ago, igormp said:

But they're not a thing in other places, not only for games.

That's fair, those Mac Pro ones are the last holdout I know of on the workstation side.

Intel HEDT and Server platform enthusiasts: Intel HEDT Xeon/i7 Megathread 

 

Big PC (disassembled pending re-rebuild) - i9 7980XE @4.4Ghz/~1.18v - EKWB Quantum-something block - EVGA X299 Dark - 4x8GB HyperX Predator DDR4 @2400Mhz - Intel ARC A770 LE - 1TB 970 Evo  + 1TB MX500 + 500GB XPG Atom30 + LG WH14NS40 + Lite-On IHAS124-14 - EVGA 1600W T2 - Corsair 750D Airflow - CPU-only custom loop w/360 slim rad + 280mm thick (60mm) rad & Noctua Redux NF-P12/P14s fans - Windows 11

 

Little PC - i7 6950X @4GHz core/3.5GHz uncore 1.2v - Heatkiller IV block - EVGA X99 Micro 2 - 4x8GB 3200MHz DDR4 - 2060 Super FE w/Corsair block - 1TB MX500 + 1TB MP34 + 500GB MX300 - EVGA 1000W G3 - ASUS AP201 - Custom loop w/360mm GTS rad - 5x Noctua iPPC NF-F12 - Windows 11 (TPM req removed)

 

Rest of the HEDT collection - Xeon X5675/EVGA X58 Classified 4-Way SLI - i7 4930K/EVGA X79 Dark 

 

Mobile/Work Devices: Late 2014 iMac 5K/2015 15" MacBook Pro (work computers) - iPhone 13 Pro Max - Apple Watch S3

 

Other Misc Devices: 2012 MacBook Pro 13" running Mac OS X Mountain Lion - Mac Pro 1,1 (2x Quad Core Xeons, 32GB DDR2, HD4870 512MB) running Mac OS X Snow Leopard - iPod Video (Gen 5.5E, swapped to 128GB SD card and running Rockbox firmware)

Link to comment
Share on other sites

Link to post
Share on other sites

well people use to sli cheap cards and that came with heat problems do to being too close. they did fix it and add more space but then gpus got fatter and fatter and tg was used instead. not only that cost of gpus went way up. why sell 2 cards at $1000 when you can sell a titan at that cost... and ya game were clear that they were coming out in unlpayable states that they would never get time to even think about sli. that and esports that dont need fast stuff was also focused on. then PHONE games... the time and moeny is now spent on micro transactions and phone games... also porting games from consoles probably has something to do with it too.

I have dyslexia plz be kind to me. dont like my post dont read it or respond thx

also i edit post alot because you no why...

Thrasher_565 hub links build logs

Corsair Lian Li Bykski Barrow thermaltake nzxt aquacomputer 5v argb pin out guide + argb info

5v device to 12v mb header

Odds and Sods Argb Rgb Links

 

Link to comment
Share on other sites

Link to post
Share on other sites

On 2/5/2023 at 1:40 AM, Zando_ said:

That's fair, those Mac Pro ones are the last holdout I know of on the workstation side.

Turns out nvidia just released such a gpu, with not only two, but four GPUs in a single pcb:

https://www.servethehome.com/nvidia-a16-with-4x-ampere-16gb-gpus-onboard-quick-look/

 

Those are likely meant for VDI, so not being used as a single gpu, but pretty interesting nevertheless. 

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share


×