• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

DRAM Pricing Jumps 50%, Only 70% of Orders Getting Filled

erek said:
View attachment 796557
Click to expand...
It's still a crazy price but MC also has similar TeamGroup ram (128GB of ddr5-6400) for $1499.

Also, from their own website:
1775941506907.png
 
BravO))) said:
How? By recounting the highs and lows of the computer market over the years? By implying 32gb of ram for 300 dollars is better than out of stock video cards, or them being 2x the price from scalpers? Then there is Thailand floods, Taiwan earthquakes, and other events that have impacted prices in the past. Duke is right, in that it isn't sustainable, and eventually this will be a phase, just like the problems of the past. If you are going to blame anyone, I would point the finger at dram manufacturing for not expanding. They applied lessons learned from the mining crash, but applied them to an expanding market. You also have to understand, big industrial players got screwed as well. Should we blame AI, when dram manufactures decided not to expand, when they could have? No one is benefiting from the shift in prices, except those making the chips. Do you think server farms related to AI want to pay 6x the price of what they were?
Click to expand...
I feel like I'm talking to a wall.
 

Samsung’s Memory Business Alone Is Now More Profitable Than Amazon, Meta, And Microsoft, Quietly Becoming One of the Biggest Winners of the AI Race​

Muhammad Zuhair
Apr 11, 2026 at 09:24am EDT

Samsung's DRAM Business Alone Has Managed To Achieve $37 Billion In Q1, And The Growth Won't Stop Here​

DRAM has become a commodity in the AI industry, in high demand, and suppliers like Samsung, SK hynix, and Micron are unable to meet customer requirements, driving a wider shortage across the supply chain. With that, a report by Counterpoint Research has talked about Samsung's Q1 financial results, but the interesting parallel drawn by them was the fact that the Korean giant's operating revenue was actually higher than the likes of Amazon, Microsoft, TSMC, and Meta, showing a clear indication that the memory supercycle has come in to boost the company's business.

There has been concern about the sustainability of this demand, but Samsung is also addressing it by signing up to 5-year LTAs with hyperscalers.

https://wccftech.com/samsung-memory...re-profitable-than-amazon-meta-and-microsoft/
 
erek said:
View attachment 796557
Click to expand...
I always wonder who the customer base is when you see crap like this. Are there actual people going into stores and buying $4200 kits? Businesses that need it to make money probably aren't walking into Microcenter (or whatever this place is) to buy computer stuff like this. Gotta say, the security level of that place has to be through the roof when you have display cases like this with maybe a quarter to half a million dollars "worth" of product
 
you2 said:
It is interesting you all talk about DRAM prices but stop and consider what AI is doing to Energy as well as water prices.
Click to expand...
As mentioned... yes we are thinking. And now we're getting slapped in the face (hard) by the age old capitalistic process of "If I'm willing to pay more than you, you don't get any"
 
Shoganai said:
I feel like I'm talking to a wall.
Click to expand...
Then I will say it plainly. You are being sensational and entitled. There are plenty of affordable options out there, and all run cloud based LLMs just fine. There are also plenty of options for under 1k that can do local LLMs and gaming just fine. You just want pro performance for a budget in a field with low inventory and high demand...
 
BravO))) said:
Then I will say it plainly. You are being sensational and entitled. There are plenty of affordable options out there, and all run cloud based LLMs just fine. There are also plenty of options for under 1k that can do local LLMs and gaming just fine. You just want pro performance for a budget in a field with low inventory and high demand...
Click to expand...
No, I'm not. You're being delusional and ignoring reality. You're also just flagrantly lying. You're not going to gaslight anywhere in here, no matter how much you want to. This entire thread is literally the antithesis of the gaslighting you're failing to do. But you do you.
 
I’m worried that even if the rumored 40% purchase of ram by OpenAI falls through, these ram manufacturers are going to just not manufacture or release that supply, given they can charge so much more now.
They have no incentive, since they are making multiple times more per chip now. That would more than make up for selling 40% less chips.
 
Shoganai said:
No, I'm not. You're being delusional and ignoring reality. You're also just flagrantly lying. You're not going to gaslight anywhere in here, no matter how much you want to. This entire thread is literally the antithesis of the gaslighting you're failing to do. But you do you.
Click to expand...
What have I said that is untrue?
 
zerogg said:
I’m worried that even if the rumored 40% purchase of ram by OpenAI falls through, these ram manufacturers are going to just not manufacture or release that supply, given they can charge so much more now.
They have no incentive, since they are making multiple times more per chip now. That would more than make up for selling 40% less chips.
Click to expand...
they are going for long term contracts — baking in the prices at the peak of the cycle now
 
zerogg said:
I’m worried that even if the rumored 40% purchase of ram by OpenAI falls through, these ram manufacturers are going to just not manufacture or release that supply, given they can charge so much more now.
They have no incentive, since they are making multiple times more per chip now. That would more than make up for selling 40% less chips.
Click to expand...
Kinda the problem with the entire market right now. When the AI bubble does pop, the entire market for computer hardware just crashes. DRAM manufacturers aren't stupid as they know not to over produce to prevent such a situation. I still think the market is shifting away from them all together, so who cares. The biggest problem is when companies like Intel only sell CPU's with ram built in. They do this already with Lunar Lake and Panther Lake. We not be able to upgrade out PCs in the future. At least not the affordable computers.
 
Last edited:

“Valve Engineer Improves Linux Memory Management for GPUs with 8 GB VRAM or Less​

by AleksandarK Today, 02:16 Discuss (8 Comments)
Valve's Linux developer, Natalie Vock, has introduced a new method for handling memory management on GPUs with 8 GB or less VRAM. Since a large part of the gaming community uses systems with limited resources, improving performance and VRAM allocation is an excellent way to maximize frames per second. According to Vock's solution, GPUs with 8 GB of VRAM or less on Linux will now benefit from better background VRAM management, allowing more free space for games. Several Linux kernel patches have been submitted to guide how Linux manages resources when a game is loaded. Previously, when a game used too much VRAM on a low-VRAM GPU, the Linux Graphics Translation Table (GTT) would move data to system RAM to prevent the game from crashing. Since Linux doesn't prioritize programs, it could sometimes move the actual game from VRAM to system RAM to make space for something like a browser or another third-party app.

In Vock's testing before these new patches, running Cyberpunk 2077 resulted in the system using only 6 GB of the 8 GB of available GPU VRAM, with about 1.37 GB spilling over to GTT and being allocated in system RAM. This could lead to game stuttering and frame-pacing issues. Now, Vock has developed a real-time dmemcg-booster, which stands for Device Memory Control Groups, to inform Linux when a program running on the GPU VRAM needs to stay there without being moved to GTT and then to system RAM. With the new patch, running Cyberpunk 2077 now uses almost the entire available system VRAM, reaching 7.4 GB, while GTT allocation has been reduced to 650 MB, cutting usage by more than half. In the picture below, you can check the "GameThread" behavior before and after the patch is applied.”
 
zerogg said:
I’m worried that even if the rumored 40% purchase of ram by OpenAI falls through, these ram manufacturers are going to just not manufacture or release that supply, given they can charge so much more now.
They have no incentive, since they are making multiple times more per chip now. That would more than make up for selling 40% less chips.
Click to expand...
Most of that "40%" was for wafer capacity yet to exist (it was a way to get it built/financed), it is not by OpenAI but an LLC that include MIcrosoft/Softbank/Oracle, they will use it (and/or sell some part to amd-nvidia-meta-amazon-etc...) how much will go to OpenAI is an open question.

And it will not be 40% anymore by the time they actually make chips.

DukenukemX said:
I still think the market is shifting away from them all together,
Click to expand...
That what you think when you see Samsung latest quarter result ? That sound like some no one go to the beach anymore, it is too crowded.

DukenukemX said:
biggest problem is when companies like Intel only sell CPU's with ram built in. They do this already with Lunar Lake and Panther Lake.
Click to expand...
Intel did it only for Lunar Lake, returned to cheaper off-package solution with their yield issues they could not continue to do it, that why they do stuff like this: https://www.tomshardware.com/pc-com...pus-as-generic-intel-graphics-in-task-manager.

Their packaging seem elite in term of result when it work for some things, but not apple/tsmc level for the on package memory.
 
Last edited:

“Microsoft Surface Laptops Get $200-300 Price Hike Due To RAMpocalypse​

by Cpt.Jank Today, 02:15 Discuss (5 Comments)
We've already seen large swathes of the PC hardware market get drastically affected by the ongoing DRAM and general hardware crisis, from Valve's upcoming Steam Machine to pre-built desktop PCs. Now, it seems as though it's Microsoft's turn with its Surface device line-up, which has seen a price increase of anywhere between $200 and $500, depending on the exact model. The price increases mean that the 13-inch Surface Pro with the Qualcomm Snapdragon X SoC, 16 GB of RAM, and 512 GB of storage increases by $300 to $1,499; while the 12-inch Surface Pro with the Snapdragon X, 16 GB RAM, and 256 GB of storage increases from $799 to $1,049; the 13-inch Surface Laptop with the Snapdragon X Plus, 16 GB RAM, and a 256 GB SSD increases from $899 to $1,199; the 13.8-inch Surface Laptop with the same SoC and RAM but 512 GB of storage increases from $1,199 to $1,499; and the Surface Laptop 15-inch with the Snapdragon X Elite SoC, 16 GB of RAM, and 256 GB of SSD storage sees a price hike from $1,499 to $1,599.

Microsoft confirmed in a statement to Windows Central that the price increases were a direct result of the current hardware constraints, with a Microsoft representative stating "Due to recent increases in memory and component costs, Surface is updating pricing on Microsoft.com for its current‑generation hardware portfolio. We remain committed to delivering value to customers and partners while upholding our standards for quality and innovation." This news comes just after we reported that DDR4 spot prices have seen a marginal decrease. Microsoft's Surface line-up has never been the best value when it comes to Windows laptops, and they'll likely be an even tougher sell after the price hikes, especially with Apple's recent introduction of the hyper-popular $600 MacBook Neo and with the somewhat limited app and feature support that comes with Qualcomm's Snapdragon SoCs and Windows on Arm.“
 
LukeTbk said:
That what you think when you see Samsung latest quarter result ? That sound like some no one go to the beach anymore, it is too crowded.
Click to expand...
Obviously not right now. AI boom is still booming.
Intel did it only for Lunar Lake, returned to cheaper off-package solution with their yield issues they could not continue to do it, that why they do stuff like this: https://www.tomshardware.com/pc-com...pus-as-generic-intel-graphics-in-task-manager.

Their packaging seem elite in term of result when it work for some things, but not apple/tsmc level for the on package memory.
Click to expand...
Might explain why their power efficiency is that great. I thought they did the same like with Lunar Lake.
 
DukenukemX said:
Might explain why their power efficiency is that great. I thought they did the same like with Lunar Lake.
Click to expand...
? on package tend to be the most power efficient way to do things, panther lake achieving it despite returning to the cheaper/less efficiant way is impressive.

Intel 18A, ribbonFET and powervia is just really good, lot more cache they added help reduce the energy cost of the off package memory has well.
 
Last edited:
DukenukemX said:
Obviously not right now. AI boom is still booming.

Might explain why their power efficiency is that great. I thought they did the same like with Lunar Lake.
Click to expand...
LukeTbk said:
? on package tend to be the most power efficient way to do things, panther lake achieving it despite returning to the cheaper/less efficiant way is impressive.

Intel 18A, ribbonFET and powervia is just really good, lot more cache help reduce the cost of the off package memory has well.
Click to expand...

"NVIDIA GeForce RTX 5060 and 5060 Ti Planned With 9 GB VRAM​

by Cpt.Jank Today, 19:33 Discuss (7 Comments)
It was recently revealed that NVIDIA plans to launch a version of the RTX 5050 GPU with 9 GB of VRAM—three 3 GB modules of GDDR7 over a 96-bit bus and 336 GB/s total memory bandwidth. Now, according to a new leak by Hakuitado on the Board Channels forum (via Gazlog), NVIDIA is planning to make the same move with the RTX 5060 and 5060 Ti in order to cater to budget consumers while avoiding high DRAM and BOM costs.

DRAM quantity aside, the move would also see NVIDIA decrease the bus width of the respective cards, going from 128-bit to 96-bit by using new 24 Gbit memory chips, resulting in an overall decrease in bandwidth and reduced memory performance. According to @harakuze5719 on X, the new 9 GB NVIDIA GeForce RTX 5060 and 5060 Ti GPUs are slated to launch sometime between May and June, with a Computex launch seemingly likely."
 
erek said:

"NVIDIA GeForce RTX 5060 and 5060 Ti Planned With 9 GB VRAM​

by Cpt.Jank Today, 19:33 Discuss (7 Comments)
It was recently revealed that NVIDIA plans to launch a version of the RTX 5050 GPU with 9 GB of VRAM—three 3 GB modules of GDDR7 over a 96-bit bus and 336 GB/s total memory bandwidth. Now, according to a new leak by Hakuitado on the Board Channels forum (via Gazlog), NVIDIA is planning to make the same move with the RTX 5060 and 5060 Ti in order to cater to budget consumers while avoiding high DRAM and BOM costs.

DRAM quantity aside, the move would also see NVIDIA decrease the bus width of the respective cards, going from 128-bit to 96-bit by using new 24 Gbit memory chips, resulting in an overall decrease in bandwidth and reduced memory performance. According to @harakuze5719 on X, the new 9 GB NVIDIA GeForce RTX 5060 and 5060 Ti GPUs are slated to launch sometime between May and June, with a Computex launch seemingly likely."
Click to expand...

"Microsoft Reveals Major Price Increase For All Surface PCs (windowscentral.com)16​

Posted by BeauHD on Tuesday April 14, 2026 @06:00PM from the sticker-shock dept.
Microsoft has sharply raised prices across its Surface lineup as RAM and component costs keep climbing. "Both its midrange and flagship Surface lines are now significantly more expensive than they were just a few weeks ago, with the flagship Surface Laptop 7 and Surface Pro 11 now starting at $500 more than they launched at in 2024," reports Windows Central. From the report:The Surface Pro 12-inch, which was previously Microsoft's cheapest modern Surface PC at $799, now starts at $1,049. The flagship Surface Pro 13-inch, which originally launched for $999, now starts at an eyewatering $1,499. It's the same story for the Surface Laptop lines, with the entry-level 13-inch model originally priced at $899, now starting at $1,149. The 13.8-inch flagship Surface Laptop launched at $999, but now costs $1,499, with the 15-inch model now starting at $1,599. This means that Microsoft's midrange devices now cost more than the flagships did when they launched in 2024.

[...] Microsoft has raised prices for all SKUs on offer, meaning the high end models are now more expensive too. A top end Surface Laptop 15-inch with Snapdragon X Elite, 64GB RAM and 1TB SSD storage now costs a staggering $3,649. To compare, the 16-inch MacBook Pro with an M5 Pro, 64GB RAM, and 1TB SSD is $3,299, and that comes with a significantly better display and much more power under the hood."
 

“Gamers Will Cut RAM Before Settling for SSDs Smaller Than 512 GB, Lexar Says​

by AleksandarK Today, 03:24 Discuss (16 Comments)
During a media tour at Lexar HQ in China, Digital Foundry spoke with Lexar's Europe General Manager, Grace Su, who explained what gamers are buying and where they are willing to compromise. Interestingly, Lexar found that, despite the pricing of DRAM and NAND Flash, users could purchase lower-capacity RAM kits without issues, while SSDs with capacities lower than 512 GB weren't selling well. Although this is not a direct comparison, when building a new PC system, you need both a RAM kit and an SSD. Based on their budget, enthusiasts often have to decide between a larger SSD or a bigger RAM kit. However, Lexar reports that users are so committed to having larger SSD space that they are even willing to revert to HDDs rather than buy an SSD smaller than 512 GB. This is understandable, considering that modern games can take up hundreds of gigabytes of space.

Currently, high RAM prices are driving PC gamers to invest considerably more in their gaming rigs to play smoothly. As 16 GB is somewhat considered the minimum for a Windows 11 system, some games and tasks are so demanding that PC DIY enthusiasts need to invest in 32 GB kits at prices that were unimaginable just a year ago. For example, we reported on the DDR4 spot market pricing, which saw the cost of a single 16 Gb module increase by about 2,200% over the past year, with only a minor 5% correction in March. We can only hope for more corrections soon so that PC gamers can finally see signs of relief, but with AI demand projected to continue rising, that seems far from reality.“
 
erek said:

“Gamers Will Cut RAM Before Settling for SSDs Smaller Than 512 GB, Lexar Says​

by AleksandarK Today, 03:24 Discuss (16 Comments)
During a media tour at Lexar HQ in China, Digital Foundry spoke with Lexar's Europe General Manager, Grace Su, who explained what gamers are buying and where they are willing to compromise. Interestingly, Lexar found that, despite the pricing of DRAM and NAND Flash, users could purchase lower-capacity RAM kits without issues, while SSDs with capacities lower than 512 GB weren't selling well. Although this is not a direct comparison, when building a new PC system, you need both a RAM kit and an SSD. Based on their budget, enthusiasts often have to decide between a larger SSD or a bigger RAM kit. However, Lexar reports that users are so committed to having larger SSD space that they are even willing to revert to HDDs rather than buy an SSD smaller than 512 GB. This is understandable, considering that modern games can take up hundreds of gigabytes of space.

Currently, high RAM prices are driving PC gamers to invest considerably more in their gaming rigs to play smoothly. As 16 GB is somewhat considered the minimum for a Windows 11 system, some games and tasks are so demanding that PC DIY enthusiasts need to invest in 32 GB kits at prices that were unimaginable just a year ago. For example, we reported on the DDR4 spot market pricing, which saw the cost of a single 16 Gb module increase by about 2,200% over the past year, with only a minor 5% correction in March. We can only hope for more corrections soon so that PC gamers can finally see signs of relief, but with AI demand projected to continue rising, that seems far from reality.“
Click to expand...
Can't install modern AAA games on a 256GB SSD. Maybe one game, but that's about it.
 
erek said:

“Valve Engineer Improves Linux Memory Management for GPUs with 8 GB VRAM or Less​

by AleksandarK Today, 02:16 Discuss (8 Comments)
Valve's Linux developer, Natalie Vock, has introduced a new method for handling memory management on GPUs with 8 GB or less VRAM. Since a large part of the gaming community uses systems with limited resources, improving performance and VRAM allocation is an excellent way to maximize frames per second. According to Vock's solution, GPUs with 8 GB of VRAM or less on Linux will now benefit from better background VRAM management, allowing more free space for games. Several Linux kernel patches have been submitted to guide how Linux manages resources when a game is loaded. Previously, when a game used too much VRAM on a low-VRAM GPU, the Linux Graphics Translation Table (GTT) would move data to system RAM to prevent the game from crashing. Since Linux doesn't prioritize programs, it could sometimes move the actual game from VRAM to system RAM to make space for something like a browser or another third-party app.

In Vock's testing before these new patches, running Cyberpunk 2077 resulted in the system using only 6 GB of the 8 GB of available GPU VRAM, with about 1.37 GB spilling over to GTT and being allocated in system RAM. This could lead to game stuttering and frame-pacing issues. Now, Vock has developed a real-time dmemcg-booster, which stands for Device Memory Control Groups, to inform Linux when a program running on the GPU VRAM needs to stay there without being moved to GTT and then to system RAM. With the new patch, running Cyberpunk 2077 now uses almost the entire available system VRAM, reaching 7.4 GB, while GTT allocation has been reduced to 650 MB, cutting usage by more than half. In the picture below, you can check the "GameThread" behavior before and after the patch is applied.”
Click to expand...
There is a ton more of this that can be done, and maybe it will be now that it starts making financial sense. Optimization is not a sexy word that any corporate manager wants to hear about.

The irony is that with AI code we can expect to see tons more spectacularly non-optimized code.
 
myzz said:
There is a ton more of this that can be done, and maybe it will be now that it starts making financial sense. Optimization is not a sexy word that any corporate manager wants to hear about.

The irony is that with AI code we can expect to see tons more spectacularly non-optimized code.
Click to expand...

The only hope is AI optimizing tools that do it all automagically and I don't that would even fix it.

Almost no one cares about optimization enough to make it a priority. If it isn't a priority the game won't be optimized. With games it's extremely easy to kill performance with one line of code or one asset. It's also easy for performance to slowly degrade as more code and assets get added to the game.

There are actually a lot of good performance tools, even free tools built into engines but people just ignore them because performance is not a priority. Even with all the tools and all the free knowledge out there the majority of game still aren't optimized for good performance. So I don't think AI is going to magically fix this.
 
Yes agree, at this stage the optimization that AI does still takes skill. You can tell it to optimize some basic well known algorithm but a game is a concert of multiple systems all working together. Requires exceptional (aka expensive) depth and breadth of knowledge and experience (in the human) to know where to look and what to do.
 
Except its not laughter, its a grimace I have. Because I managed to get my stuff before the world flipped I don't have these problems such as storage, ram, processing, etc. So because of the AIpocalypse I'll be forced into a similar boat for those newbs who want a computer today and can't afford a 4TB ssd
 
  • Like
Reactions: erek
like this
erek said:
even with compression?
Click to expand...
I don't think compression is going to help here. For a while we've seen games with install sizes bigger than 200GB. I hate to imagine what World of Warcraft install would be with classic and retail. Good thing I never touch retail.
sharknice said:
The only hope is AI optimizing tools that do it all automagically and I don't that would even fix it.

Almost no one cares about optimization enough to make it a priority. If it isn't a priority the game won't be optimized. With games it's extremely easy to kill performance with one line of code or one asset. It's also easy for performance to slowly degrade as more code and assets get added to the game.

There are actually a lot of good performance tools, even free tools built into engines but people just ignore them because performance is not a priority. Even with all the tools and all the free knowledge out there the majority of game still aren't optimized for good performance. So I don't think AI is going to magically fix this.
Click to expand...
Texture compression tools like Intel Texture Set Neural Compression would be a great idea, given that it wasn't favoring Intel hardware. It can work on AMD and Nvidia hardware, but obviously optimized for Intel. We need an industry standard for this so that we can finally have games that don't take up hundreds of gigs. Also, audio is becoming an issue as it's taking up a lot of space. Titalfall from 2014 has 35GB just for audio. We could also go back to have music that's generated on the fly which not only reduces file sizes but also brings back seamless audio transitions. A technology that was around since 1990 and was lost to time. That's not true, Nintendo still does it.

View: https://youtu.be/7N41TEcjcvM?si=01oqR71znRMFucr5
vegeta535 said:
You laugh but it is going to happen within 10 years.
Click to expand...
I heard this 10 years ago.
 
erek said:
“Reminder to enable ZRAM on Linux to optimize RAM usage.https://cnx-software.com/2026/04/15/reminder-enable-zram-on-your-linux-system-to-optimize-ram-usage/With the price of RAM getting out of control, it might be a good idea to enable ZRAM with zstd compression on your Linux machine before a potentially expensive upgrade. It's also a consideration when purchasing a Raspberry Pi or other SBC, although ZRAM is often enabled by default, for instance, on Raspberry Pi OS.”
Click to expand...
I've got zram enabled on both of my systems. On the 16GB system it may make a little difference, on the 32GB system I can't really notice much of a change TBH.

Running quad channel DDR4 on the 32GB system, I've got enough bandwidth to tie me over for a while yet, and a few tweaks left up my sleeve for more bandwidth should I feel the need.

Intel MLC results memory bandwidth.png
 
Last edited:
  • Like
Reactions: erek
like this

“NVIDIA GeForce RTX 3060 12 GB Returns in June, RTX 5050 9 GB Edition on Pause​

by AleksandarK Today, 05:48 Discuss (12 Comments)
NVIDIA's GeForce RTX 3060 12 GB "Ampere" GPU, which is two generations old, is expected to make a comeback in June this year. Meanwhile, the rumored RTX 5050 9 GB edition is reportedly on hold. According to well-known hardware leaker MEGAsizeGPU on X, NVIDIA is pausing the transition from its 8 GB RTX 5050 "Blackwell" version to a 9 GB model due to the reintroduction of the GeForce RTX 3060 12 GB "Ampere" GPU. Since both of these GPUs compete in the budget segment, the company will reportedly only release the older GeForce RTX 3060 12 GB SKU with a 192-bit wide memory bus. For this, NVIDIA will once again use Samsung's 8 nm DUVnode, as it has in the past. The entire NVIDIA "Ampere" architecture lineup was produced on the 8 nm DUV node, and its return after several years was unexpected. However, since NVIDIA has transitioned to TSMC for manufacturing its "Ada Lovelace" and latest "Blackwell" GPUs, becoming TSMC's largest customer using the 5 nm node, this move is intriguing.

The NVIDIA GeForce RTX 3060 features 3,584 CUDA cores and 12 GB of GDDR6 memory. In contrast, the GeForce RTX 5050 has newer CUDA cores on a smaller node with 2,560 cores on the GB207 die, but it has less memory, with only 9 GB of the rumored GDDR7. Since the resurrected GeForce RTX 3060 will use the older GDDR6 memory, sourcing GDDR6 may have been easier for NVIDIA, leaving the more expensive GDDR7 for its other GPUs.“
 
You must log in or register to reply here.
Back
Top