GPUs with Expandable VRAM!

[@cube5380]

HEY WAIT THAT'S ME LMFAO

[@Axel.477]

no it aint

[@zappierpine890]

Ye it is

[@BlazingTonk]

balls

[@mastertang4200]

​@@Axel.477his @ is the same and I don't think it is changeable when you change your name

[@cube5380]

@@Axel.477 the second comment, the one that said "balls" also hi cane

[@Vathrex]

This existed before. I read it was common in the early 90s. There would be slots for extra vram chips. The S3 Trio had it

[@Kaillin]

I have one of these gpu's

[@justinpatterson5291]

Yeah. I was just thinking that really old agp cards had that at one point.

[@Vathrex]

​@@KaillinWhat a chad 😎

[@delleron1106]

Well th

[@Nick-123]

They removed that because it was too slow so it would not work on todays GPUs

[@dan8t669]

Expendable VRAM was a thing on workstation cards a long time ago. Welcome to Tech, Zach.

[@LabraDork-uj7ib]

Not at the speeds we see memory running today

[@OriginalName707]

"How come the consumer grade electronics guy doesn't know about enterprise level hardware"

[@Dr.W.Krueger]

A long time ago indeed.

[@diamonshade7484]

​@@OriginalName707probably forgot

[@nikoheino3927]

​@@OriginalName707well, but he doesnt even understand the technicalities of consumer products. its simply impossible to have upgradeable vram unless its in the form of soldering new chips yourself. modern vram operates at very high speeds and requires perfect signal integrity, which requires the chips to be soldered right next to the gpu die itself.

[@Tempestan]

As others have posted, yes, they had this feature back in the late 90s and early 2000s. There is a reason why you do not see them on cards today.

[@AgentK-im8ke]

Money ?

[@Tempestan]

@@AgentK-im8ke terrible performance, ok, not exactly terrible, but it preforms subpar to direct soldered chips on the card, and it is harder/costlier to manufacture. So, yes, money in part.

[@user-hr4hu8xb5f]

@@diamonshade7484 Ram need to be physically close to the chip to be fast, thats why Dram is ultra slow comparing to cache and Vram.

[@petrosarvanitis2800]

A guy added an extra 8 gb of vram on a 3070. I'm sure if companies wanted to, they could develop gpus with "empty memory slots"

[@user-hr4hu8xb5f]

@petrosarvanitis2800 Let average stupid customers get a chance to disassemble the cooler, break the card, and then blame the company? I won't do that if I run the company lol.

[@exclar]

The main reason they haven't done this is that it would make the RAM slower. As you can see, the RAM chips on the GPU are very close to the GPU chip itself to reduce latency. This is why this is not recommended.

[@exclar]

If expandable memory were introduced, it could potentially lead to increased latency and hinder the GPU's performance.

[@cardboard_boi]

Yes

[@cardboard_boi]

This is why desktop ram can’t reach the same speed as vram

[@Kaillin]

​@@cardboard_boiit is more because cpu's do not need high frequency ram, but yes soldered ram is faster

[@AMPS1]

Yes, exactly. Even if they would have added "extra" ram slots besides the solderer vram, not only would it be a nightmare for the engineers and bios devs, the extra vram would most likely not really useful as it's slower and lowers the gpu performance.

[@GabeDaBabe930]

Balls

[@ideeAMxCC]

no

[@JayJadhav]

Banana

[@safabi7577]

Balls

[@SNZ-Clan-GERMAN]

Balls

[@LockerGames]

Ty

[@GhostyKingdom]

Engineer here. Unfortunatly, this cant happen (yet) because ddr6 VRAM is highly specialized and not commercially available and tends to be built differently between GPU's.

[@ozgurpeynirci]

Just like there is a JEDEC standard for RAM, we can have it for GPUs especially Intel coming out for competition, there could be an aggreement.

[@triynizzles]

They would have to completely redo, the memory controller architecture. Currently if you have a 64 bit bus(small number to make the example easier), then there’s two chips on the pcb. The only thing you could do is potentially clamshell to double size, but maintain speed or go from 18gbps to 19gbps (faster frequency) but the main thing that will give you more bandwidth and vram size is a larger bit bus, so there are more chips providing data simultaneously.

[@ozgurpeynirci]

This doesn't make sense, if it requires 8 chips then there will be VRAMs sold with 8 chips in 1 unit.

[@arkgaharandan5881]

thank you, someone who is not dumb and has an idea wtf he is talking about, unlike this guy in the video.

[@user-nm1sw2fc9f]

They have done this, in fact it used to be quite common. The reason it stopped was the same reason why it’s difficult to find the highest speeds of DDR5 in laptops with slotted memory: it was becoming too difficult to increase the speed of the memory with the much longer traces that you need for slots. When the memory is soldered directly to the board, the traces can be much shorter because they’re able to put the chips right next to the GPU. That’s why you see a circle of memory chips around the GPU if you take off the heatsink of a modern graphics card, it allows them to all be as close as possible and the traces as short as possible.

[@Farouk.khettab]

I would love to see a combo of fast + slightly slower memory, just like how 3DV cache works with RAM.

[@doggo_woo]

​@@Farouk.khettabThat would be sick. Fast vram for important or constantly used assets and slower expandable vram for the rest. Kind of like how we have swap for when we run out of actual ram.

[@stevy2]

@@Farouk.khettab Basically the GTX 970. People exaggerated but that last 512mb was still faster than swapping to RAM.

[@anthonypolanc0]

Dude watches the tech tips

[@narwahlssb]

I was about to lay this down but you beat me to it. I'm fairly sure they had some older cars in the 90s that did have expansion memory.

[@hippityhop9522]

While normal RAM isn't really affected by distance from CPU, VRAM is very dependant on the distance between it and the core of GPU.

[@arne_666]

Normal RAM is actually affected, for example laptops using DDR5 had some problems with the speed and so are now soldering it to the board directly.

[@sarowie]

@@arne_666 let us phrase it differently: A modern CPU and operating system has such a complex caching schema, that the average user does not notice the speed difference of a socketed CPU and socketed RAM while using chrome. (we are using real world applications, not benchmarks) Meanwhile in gaming... yeah: Gaming is gaming and benchmarks are aqurate. There might be a difference in AI, Mining and Gaming performanc depending on...

[@wolvreigns]

​@@arne_666also why most high end over the top overclocking motherboards only have 2 ram modules. They are closer to the CPU and of course 2 ram sticks is more stable then 4

[@ARCAD3BLOOD]

both rams are affected. Vram and ram are pretty similiar things. If your prediction would be true, then why they don't mount ram slots anywhere but close to the cpu?

[@josephdias3968]

Just make detachable vram chiplets that clip on and off right next to the gpu core

[@BF26595]

Expandable memory GPUs actually existed a while ago... the S3 Virge had 2MB on board and you could add a daughter board with another 2MB of memory :) I still have it in my DOS Stuff box 😇

[@daredog509]

mad respect to the guy who commented "balls" and then got featured in the video

[@LexyDaShmexxy]

they used to do that but there was a limitation in the speed and bandwidth, it's just too slow to implement, that's why vram chips are so close to the gpu die

[@walkinmn]

and that's why Apple put the ram chips so close to the SoC too and some other manufacturers are starting to do similar things and I'm worried this trend could catch on, I mean, yes, the speed and performance gains are cool but is it really that hard to make upgradeable ram and vram modules that can bring these upgrades without losing the advantages of letting us upgrade the ram? The issue is the industry probably has more interest in not exploring that idea and using the performance excuse to make all the chips non upgradeable than to give us something better and modular.

[@zacker150]

​@@walkinmnyes. It's not just hard. it's basically impossible. You are literally butting up against physical constraints like the speed of light

[@walkinmn]

@@zacker150 do you have any sources to back that up? Because I actually know physics and I'm pretty sure that's not true

[@zacker150]

It's common knowledge. Data has to physically move from the memory to the processor and can only move a a fraction of the speed of light divided by the material's dielectic constant. As you increase the frequency, information has less time to travel from the GPU and the VRAM. At a frequency of 2335Mhz (the frequency of the VRAM on the 4090), light can only travel 12.8 cm, so the VRAM must be at least 6 cm of the die.

[@MiguelAngel-yr5zb]

You can search for transmission line theory and length matching for high speed pcb, I´m currently studying about that. Sorry in advance for misspeaking, I´m not native speaker@@walkinmn

[@ossiaaltola4578]

The reason that they don't have expandable VRAM is that it would make the traces from the VRAM to the gpu die too long. The closer the VRAM is to the GPU the faster you can make the ram. And also they want your money.

[@kilgarragh]

you put them on the back so the slots are right next to the die

[@_ace____w_w__________]

​@@kilgarraghexactly 💯 I was thinking the same

[@ossiaaltola4578]

@@kilgarragh not quite how that works

[@sarowie]

@@kilgarragh very funny. Even just the fan out to and from the slot is a significant signal length. Even just the pins within in the slot. Yes, you can make the pitch (pin to pin distance) smaller... but did we not start out with a customer upgradeble slot? Now we make it so fragile, that I do not want to see it on the assembly floor?

[@ARCAD3BLOOD]

then make it double. You can have faster and slower vram in the same system.

[@daydream605]

two words to describe why this hasn't happened Cost Latency

[@airbusaviation6548]

I really want to see zach making a pc with a GTX 1080 Ti. Still very good to this day and its cheap.

[@kristofmadarasz3166]

Fun fact :Back in day's they did expandable vram on gpu's just they stoped using it because it was slow and expensive .

[@sarowie]

it was not slow. I mean: It was fast enough at the time, but with faster GPU and RAM speeds, the margin for unnecessary connections and traces is getting smaller and smaller.

[@notisac3149]

That would be amazing, especially as newer games require sooo much more vram. There would also be literal tons less e-waste as older cards would remain viable for possibly years longer! Edit: It seems this used to be a thing on older cards. The prevailing thought is that expanded vram is quite a bit slower. Not sure how much validity there is in that but I’m sure there’s at least some truth to it. Still, it’s fun to see people replace parts in older cards to install more vram themselves, just not as quick and easy as a little slot like an sd card or what have you.

[@im4d3m0n_]

The issue isnt really that gpu companies want to make money, its that the further away the ram sits from the gpu, the higher the latency and the slower the data transfer, which is an impossible issue to result due to physics, we physically do not yet have a way to make the type of ram that gpus need expandable if not by soldering new ram chips on it or by significantly reduce the gpu performance due to incredibly slow ram

[@im4d3m0n_]

​@@filipionescu4543my friend, u point fingers but dont do research

[@Dell-ol6hb]

@@filipionescu4543 that’s not really why, it’s more of a performance reason, graphics card still last years anyway, expandable vram isn’t a thing anymore because to achieve the speed and therefore performance we see in modern graphics cards you need to have the memory chips soldered to the board as close as possible to the gpu itself. This is why vram is way faster than normal ram you put in your motherboard, it’s because it’s literally right next to the processor

[@haywagonbmwe46touring54]

@@filipionescu4543 Oh boy... not how any of this works

[@ARCAD3BLOOD]

@@filipionescu4543 Performance of a card is not solelly dependant on vram. Amd is just better at matching their cards performance with corresponding vram values. They could also create their own branded sticks of vram, and nvidia could, as always, close off the market to only their vram sticks.

[@No_one448]

"I like money" all GPU companies

[@miaugato93]

i feel old now. this used to be normal in the 90s.

[@verhulstak]

They actually used to do that but a socket cant do the speeds for it

[@LyronAguiar]

Linus did a video explaining why this would not work

[@uh-nuh]

it would work, just not gonna perform good

[@uh-nuh]

it would be like those render gpus, they are not fast ans thwy cant run games well but they has lots of vram

[@arakwar]

He probably screwed up that data too.

[@RuyGedares_GuyRedares]

​@@arakwarXD😂

[@norliasmith]

Graphics cards in the past: *Whistling non-chalantly*

[@InventionPR]

Who remember Cache Memory Sticks? Good old days

[@edwinma2910]

GPU companies just want your money, and don't really care about things like this

[@cardboard_boi]

There is actually a problem with this though, it’s signal degradation. That’s why desktop ram just can’t be as fast as gpu vram. At least that’s the biggest thing I’ve read. Though I’m sure the money is a huge factor as always.

[@berryblade1305]

No its not only for that, its not practical and will be slow

[@cardboard_boi]

@@berryblade1305 yup

[@Alpine_flo92002]

Not only is it practically impossible, if there was a solution to it there would be so much user error we would have the same uproar of User caused failures like with the high power 12V connectors

[@ARCAD3BLOOD]

@@cardboard_boi Usually it's a deal for deal. Vram up to ddr5 trade speed for latency. Ram is not as effective with high latency.

[@krishahirwar7326]

They haven't done it coz it will increase the response time .even just by some nano second but it will be significant 😊😊😊

[@BlueSheep777]

They didn't do it because: AMD: It's good enough that all the power is used. NVidia: They didn't do it so you would buy a better model. Also the 16gb 4060ti's vram is more expensive then the actual vram cost.

[@soju506]

Lots of graphics cards a couple of decades ago had a design like that where there were slots on the graphics card similar to ram on your motherboard. The reason they don’t do this anymore is because of signal integrity. One of the main factors that makes vram so much faster than normal ram you slot into your motherboard is that it is closer to the die. The typical rule of thumb is that the faster you want something to be, you will have to move it closer to the die to maintain signal integrity.

[@triandot]

"yeah that would be a pretty cool life-changing invention, but it has to be prettyyyyy"

[@thatonecubeguy4499]

That is a bad idea , the chips have to be as close as possible to the die and adding vram through slots would make the vram much slower than embedded vram.

[@ARCAD3BLOOD]

I believe with current technology, not alot would be lost. You can always make slots also close to the chip. Maybe not a concept for next rtx 5090, but for something like next rx 8600 xt. also who said you couldn't have two sets of vram? Esram worked pretty well.

[@thatonecubeguy4499]

The issue is that the physical length of the "wires" connecting to the gpu die would be too long and delay the vram's response time @@ARCAD3BLOOD

[@ARCAD3BLOOD]

Similar technology was used in creation of i7 5775c. It had it's own ram, but also allowed normal ram to function. Xbox one used similar technology. You can always put them on the back of the card, reducing the length. And even if it would "loose" some speed, you can compensate with built-in ram, that would be smaller and closer, which would allow it to be even faster. @@thatonecubeguy4499

[@ryankline8188]

Dude. Seeing a fully loaded gpu with all its slots filled, would look badass. Like that old sega genesis!😂🤣

[@hinterhaeltiger]

They actually did that. But as latency got more, they realized that expansion slots really eat away at latency. And since the graphics card is one of the most latency-critical components in your computer, it just isn't feasible anymore.

[@ImTikhon]

We are going to have to build GPUs in the future 💀

[@syn010110]

this was hella common in the 90s

[@AzureTempest]

*looks nervously at folks soldering more ram onto the 4060’s*

[@Zenefor]

maybe the extra slottable memory could act as an active cached shader on the gpu rather than on the m.2.

[@Johnny_C137]

Expandable VRAM would likely be slower clock speeds than integrated VRAM, but if it allows you to go balls out overkill on VRAM, it might still be worth it.

[@crazytelevisiontm2969]

whos the random guy writing balls lmaoooo

[@davidgrunga]

Would be the most consumer friendly tech drop in years. Would love to see ut

[@Zorisura]

"But I digress" Bro thinks he's JayzTwoCents

[@asherthefloofer]

Water cooling will be in another level

[@applemirer3937]

I seem to remember a card that you could add hundreds of more megabytes to, and it was cool for the time. They said that they stopped because of the number of traces needed for modern vram. HBM needs even more traces and is close to the gpu. So they're going in the opposite direction.

[@ruebo2817]

Won’t happen because of performance. You can even see it in laptop with DDR5. The SODIMM versions aren’t as fast as the soldered ones. You’d have to somehow invent a slot that has next to no latency and be really close to the processor. You can’t really do it on the front of the GPU because of cooler designs and therefore it would have to be on the back. Thus adding more distance and more latency, leaving out the latency the slot would inherently introduce. I would be really surprised if we were to se such a GPU.

[@hanro50]

This use to be a thing. Issue is that modern vram has to sit so close to the main GPU chip that this is no longer possible without sacrificing performance significantly

[@Paradigmfusion]

They did have this once upon a time, the first PC I built had a 256kb Trident VGA card in it that had modules for VRAM upgrading.

[@amaurytt]

Bought an ATI in 1996 that had an expansion VRAM module from 4 to 8MB

[@BubbyGamingOfficial]

me: uses the M.2 Nvme as more Vram

[@geekshed1674]

They used to have expandable memory back in the nineties, i expanded the VRAM in my S3 trio

[@toyotaae86trueno]

Simple, the chips on the cards anre directly connected to where they would have been on the main board of the gpu. Just make a dimm/sodimm-style connector that connects both together Kinda liie how game cartridges worked on early game consoles like the NES or the Gameboy.

[@stuartwilson4960]

Great idea, along with room temperature superconductors...... 😂

[@idontknowagoodnam3]

They should add ram slots on ram

[@icerink239]

Being detachable increases physical dimensions of the connection which also increases the latency of the memory

[@rtechie]

Back during the VESA Local Bus days (1992) graphics cards had memory slots and you could add VRAM. The reason why this died is because the slots themselves actually slowed memory access and nowadays with PCI Express that effect would be even worse.

[@hafiszaky1917]

bro casually comment "Balls" instead of tech stuff

[@Renagale]

The ram in gpus used to not only be expandable but it was also not brand specific and if you're willing to modify your gpu and void warranty you can still expand your gpu. The problem is if they allow your to easily upgrade or add your own ram they can't push newer models on you. The 1080ti was a gpu that had 12gb of vram that when overclocked can still today compete if not pass cards in the 30series that only have 8gb of v ram. So manufacturers arnt likely to allow users to upgrade their gpus for $100 when they could sell you a new "better" gpu for another thousand dollars.

[@bush2137]

They have done that, back in the days some gpu's had sodimm-like slots. That was discontinued due to the distance from the memory chips to the dye. Every additional milimiter can cause performance drops

[@kerfumblelumble]

I've learned so much about PCs. I didn't even know GPUs had their own RAM. Wow.

[@IUSAI]

Ram slots on top of the GPU could look sick though.

[@Plankton69]

Let's see how many "balls" we can get in here

[@taxuanbach0908]

The upgradable VRAM was a thing back then in 1990s 2000s lol

[@ELACEITUNERO]

The bus is a problem. It would set the roof of how much ram could have a GPU.

[@mlodikk]

Around Year 2000 I remember that my older Brother upgraded our S3 Virge card from 4 do 8MB 😂 it had 4 slots for additional chips :)

[@ninjasiren]

Late 90s and Early 2000s GPUs and 3D Accelerator cards some of those has upgradable video memory slots. Matrox G200, some of the S3 GPUs Though obviously because of the distance and latency of using a daughter board or slotted VRAM back then was slower, and far more cost effective. But would love to see its return back

[@achinti9]

I feel as if all of these can be solved with just more options. This would apply to many components, and even entire pcs.

[@vladimirlesevic6833]

BAWLLS!!!LIKE A BAWLS!!!

[@troudbalos333]

Did he just call a base level gpu 8gb of ram!!?

[@commandershadow8117]

The m.2 explanation was because of the extra PCIE lanes available from the slot not because of the GPU itself. Expandable storage on gaming cards leads to a lot of issues because of timing. Swappable memory is generally slower then soldered memory and the extra steps can cause worse fps because it can't send out data fastest enough or inconsistent frame times

[@cobaltretrotech]

I love the idea of user replaceable ram, but manufacturers stopped doing this after about the 90s or so because it is a lot harder to push high speeds that today's graphics cards need on removable chips. There is a genuine speed advantage to soldered memory, or so the story goes

[@user-ei4lc2wg2z]

Like CPU you can change 12th gen to 14th gen by only one lever like thing that connects thats make production cheap and reliable.

[@brandonmartino1363]

GPU codes are hard written for the RAM type. An example is the 3080 that has few rams upgraded to the ones that were in the 4060. The graphics card was able to see all the ram but was unable to use all the ram available and hardcode linked out when it hit or topped over the ram that was originally set in the hardcode E prom. Well, gravis car could easily do this by fixing and opening up their bios. They are considered encrypted trade secrets by the manufacturer

[@xxjjgdrgn]

The latency would be wild

[@SignalJones]

I imagine something like camm memory modules would be awesome for this. I would definitely mount it directly on the exact opposite side of the card as the die though. You want really short data runs to the GPU.

[@aftabansari606]

I've heard of this kind of GPU before. They used to have RAM like slots or something but removed it to make VRAM closer to the GPU for much better performance and lower latency and faster speeds for VRAM.

[@AnimilesYT]

I think visible expandable vram would add to the aesthetic

[@hrbt78]

In the past there were graphics cards with expandable VRAM. already twenty years ago. -man, I'm old

[@LabraDork-uj7ib]

Y’all need to read a white paper. The transfer times dramatically increase when you start putting memory on dimms or any transfer medium that isn’t directly soldered.

[@ahmedsamih]

It used to be expandable in the 90’s my first VGA card was sold with 1MB VRAm , and i later upgraded it to 2MB

[@lelkasa361]

BACK TO THE 80S WITH THIS EXPANDABLE VRAM THING!

[@scarfaceReaper]

sounds like the new phones

[@DigBipper188]

There are 3 core reasons why socketed / expandable VRAM and GPU dies aren't a thing. 1: Cost It's more expensive to add a socket that the chips seat into than creating a BGA footprint and soldering the parts directly onto the board. It also takes more manufacturing time and adds complexity. Both drive up the end cost. 2: Latency The further away an electrical signal must travel, the longer it takes for it to make it from point A to B. In RAM this causes latency. This increased latency reduces the end performance of the memory and therefore can slow the GPU if it is performing latency sensitive tasks (which it's always doing) 3: Reliability Increased trace length, plus the existence of the socketed interface both introduce the potential for the introduction of noise on a signal. This noise has the potential to cause memory errors and therefore would increase the chances of corruption or general hardware instability. GDDR memory, especially newer standards such as GDDR5 and GDDR6 are both extremely high speed, high bandwidth components. This high speed and bandwidth makes them extremely susceptible to noise. It's better to keep the parts right next to the GPU package soldered directly to the board for these reasons. Soldering creates a stronger electrical connection than socketing and since you don't have added distance it reduces the chances of an interference induced error.

[@0AThijs]

Tbh, ram dim on top of a gpu would look crazy!

[@direkta010]

The reason is because when the vram is closer is reduces the trails which interns gives you more FPS. If it was upgradable it would make the traces longer, increases costs and complexity of the trails. For the companies it would increase r&d costs and lower costumer satisfaction because it would need to be bulkier, uglier and it would be in sockets which could fail. Like if they made it a 8 vram GPU would go for 500usd. (I am not an engineer of sorts)

[@yes93924]

I'm going make my own vertical gpu out of Lego... wish me luck

[@DxrkMxntality]

What’s next? expansion slots to CPUS CRAM 💀

[@qinntt]

This used to be pretty common and they stopped because of the latency

[@w00_die49]

Honestly sounds pretty easy just create your own pcb that connects to the vram traces then figure out how you want to connect the new modular vram

[@techorgames]

the main reason they haven't done it is because the vram would be a lot slower. if it is soldered on like it is in the present then that vram can be accessed by the gpu much more quickly and so can be used quicker, resulting in higher speeds. If they were to add slots than it would slow it down by a huge margin, although they could probably have some weird hybrid thing work somehow...

[@madsfrederiktoft5808]

That existed one GPU maker in the 1990s that allowed it. Forgot the name .

[@AdRoPol]

i want to see slots for extra cache on CPU’s 😂😂

[@hmuy0608]

The same reason why soldered LPDDR RAM is usually much faster than slotted RAM on laptops: speed and latency.

[@FLoAtYno]

Bro really said a 3070 is a basic entry level gpu 😂

[@Kush_Amirash_Parekh]

What do you think about dual chipset in Graphic cards

[@boodrow24m]

I had that thought several months back. It would be awesome.

[@jasonsandhu8141]

It would be a lot slower and more expensive

[@thirtyfivefps]

it used to be like that but doing that would be very expensive and those rams won't be as fast as the ones soldered on the pcb

[@noxproductions6851]

It would be a good idea for when the cards integrated vram is used up if the expandable secondary is that close to the gpu. It wouldn't work as a good as integrated vram but would definitely work better than when the gpu has to use the dram in the pc as vram

[@gamerstewart1660]

Bro stole my idea... Exactly what I was thinking

[@aaronpalmer7425]

They did used to do that but it was more expensive and harder to optimize

[@michaelsmithers3226]

This is a really cool idea! It actually used to exist but they decided to do away with it because the memory speed was not fast enough or at least as fast as having all of the memory soldered

[@dirkjewitt5037]

They tried this. I can't remember what company it was, but LTT did a video on it. They also tried Wi-Fi on a GPU.

[@alexandre_1a442]

"I wAnT a CpU wItH eXpEnDaBlE cOrEs !"