Table of Contents
When nVidia introduced the GeForce 10-Series, it began the process of limiting SLI support to certain GPUs. Since then, Team Green has been doing this at an ever-increasing pace. With the 2000-Series, nVidia made it so that only the GeForce RTX 2070 ‘Super’ and higher GPUs supported the feature. The GeForce RTX 2070 was never really considered an ‘enthusiast-class’ GPU to begin with, so setting the bar there was reasonable. Also, the overwhelming majority of modern games didn’t include SLI support, anyway.
What graphics cards are SLI compatible?
Well, none, technically. Not any new graphics cards, at least. Now, though, with the GeForce RTX 3000-series, if you want multi GPU support you are going to have to spend at least $3000 dollars. That’s right. The only nVidia GeForce 3-Series graphics card that supports NVLink (SLI’s successor) is the $1,499 dollar GeForce RTX 3090.
To add to the cost, most power supplies would struggle to support a single RTX 3090, let alone two of them.
Multi-GPU systems are still a thing, but only barely. Now only tech enthusiasts with fat wallets can afford the minimum GPU configuration needed to have a multi-GPU setup. This change, of course, will discourage game developers from including support for SLI/NVLink in their games due to the multi-GPU market shrinking to almost non-existent levels.
Nvidia is making this move after recently releasing its RTX 3000 series video cards which, in their defense, offers an amazing price to performance ratio over previous-generation hardware. The removal of SLI, such an adorned feature that has been around since 2004, is going to upset quite a few people.
What is SLI, anyway?
SLI, which stands for Scalable Link Interface, allows a user to combine the performance of two or more graphics cards to increase gaming performance by splitting the workload across multiple cards. This would, in turn, increase frame rates. Way ahead of its time, and using a restrictive PCI express interface, SLI never really hit the mainstream.
What happened to SLI? Why did SLI go away?
Even though it’s a great idea that works well both on paper and in practice, it always represented a small part of the gamer market due to the requirement to purchase two or more graphics cards. It was hard to find an SLI compatible game, as it was usually only reserved for AAA titles in the real of Battlefield and Call of duty.
What replaced SLI? Does it work on more cards than SLI?
SLI has been replaced by NVLink. NVLink provides several advantages, but unfortunately, it’s only available on the GeForce RTX 3090. NVLink provides higher GPU-to-GPU bandwidth and also allows each graphics card’s memory to effectively be summed together.
Is the RTX 3000 series worth it?
Absolutely. The nVidia RTX provides much better performance per dollar than its predecessor, the RTX 2000 series. In fact, the RTX 3070 performs better than a GeForce RTX 2080 Ti.
|RTX 2070||RTX 2080||RTX-2080 Ti||RTX 3070||RTX 3080||RTX 3090|
|Memory Bus Width (bits)||256||256||352||256||320||384|
|Memory Capacity (GB)||8||8||11||8||10||24|
|Core Frequency (MHz)||1410||1515||1350||1500||1440||1395|
|Core Boost Frequency (MHz)||1620||1710||1545||1725||1710||1695|
|Memory Frequency (MHz)||1720||1750||1750||1750||1188||1219|
|Memory Bandwidth GB/sec||410||417||574||417||708||872|
|Computational Power (TOPS)||7.46||10.07||13.45||20.31||29.77||35.58|
|Power Consumption (Watts)||215||276||320||215||350||350|
|Recommended PSU (Watts)||550||650||650||650||750||750|
|Cost per 1 Trillion Operations Per Second||$56.13||$78.36||$96.60||$24.56||$23.48||$42.13|
|Cost per GB per second of Memory Bandwidth||$1.02||$1.89||$2.26||$1.20||$0.99||$1.72|
|Cost per GB of Graphics Memory||$52.38||$98.63||$118.09||$62.38||$69.90||$62.46|
|Cost per Graphics Core||$0.18||$0.27||$0.30||$0.08||$0.08||$0.14|
What is GDDR6X?
GRRD6X is the lastest graphics card memory available. Micron co-developed GDDR6X in a close partnership with Nvidia. GDDR6X SGRAM isn’t an official memory standard, as it has not been recognized by JEDEC yet. GDDR6X offers double the bandwidth per clock-cycle as compared to conventional GDDR6.
What graphics cards have GDDR6X memory?
Nvidia is the only company with access to Micron’s GDDR6X memory modules so far. The first graphics cards to use GDDR6X are the Nvidia GeForce RTX 3080 and GeForce RTX 3090. GDDR6X consumes 15% less power per transferred bit than GDDR6, but because GDDR6X is implemented in faster configurations than GDDR6, power consumption is higher overall. GDDR6X is still lower-cost than HMB (High Bandwidth Memory)
If you like this article, you may want to check out this one about the RTX 3000.