• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Intel Alder Lake to launch on October 27th. On Intel 7. PCIe Gen 5. DDR 5. DMI 4. 20% IPC. Xe Graphics iGPU. Hybrid Arch. Power Efficient. LGA 1700.

Leonidas

Member

Nice to see advancements in the CPU space. Some of the leaked benchmarks for single core performance have been insane (way faster than any currently released CPU, if true).

Can't wait to see the CPU landscape this fall. I'm due for a CPU upgrade soon :goog_smile_face_eyes:
 

supernova8

Banned
"Intel 7"

"Guys we don't need to move to 7nm if we just start calling our 10nm process 7nm! Nobody will ever notice!"

veronica mars wink GIF


Also I hope they change the product naming scheme itself. I mean..

Intel® Core™ i5‑1145G7E6

Really rolls off the tongue!
 

martino

Member
Pretty sure TSMC “7 nm” is closer to Intel’s 10 nm. There’s not a standardized methodology of measurements across manufacturers.
yes also
edit: there was a better reference in the news
 
Last edited:

ethomaz

Banned
"Intel 7"

"Guys we don't need to move to 7nm if we just start calling our 10nm process 7nm! Nobody will ever notice!"

veronica mars wink GIF


Also I hope they change the product naming scheme itself. I mean..

Intel® Core™ i5‑1145G7E6

Really rolls off the tongue!
Intel 10nm was always a similar node to TSMC 7nm.
TSMC names are more marketing and not the size or density used.

It was an actual node size until around 2009... after that:

"The nanometers used to name process nodes has become more of a marketing term that has no relation with actual feature sizes nor transistor density (number of transistors per square millimeter). For example, Intel's 10 nm process actually has features (the tips of FinFET fins) with a width of 7 nm, Intel's 10 nm process is similar in transistor density to TSMC's 7 nm processes, while GlobalFoundries' 12 and 14 nm processes have similar feature sizes."
 
Last edited:

Armorian

Banned
amd is so dead

the next am5 ryzen dont even have pcie5.0 🤷‍♀️

PCIE5 won't be needed for much things even after Zen 5 launches.

Hell, PCIE4 and even PCIE3 NVMEs are pretty much useless for gaming right now, they have almost the same performance as SATA drives. Direct Storage will change that but still, Xbox speed of 2GB/s will be the lowest common denominator...

And GPUs are still below PCIE3 limits at this point.

That said, AL looks very interesting.
 
PCIE5 won't be needed for much things even after Zen 5 launches.

Hell, PCIE4 and even PCIE3 NVMEs are pretty much useless for gaming right now, they have almost the same performance as SATA drives. Direct Storage will change that but still, Xbox speed of 2GB/s will be the lowest common denominator...

And GPUs are still below PCIE3 limits at this point.

That said, AL looks very interesting.
lmao wut.
 

Kenpachii

Member

Nice to see advancements in the CPU space. Some of the leaked benchmarks for single core performance have been insane (way faster than any currently released CPU, if true).

Can't wait to see the CPU landscape this fall. I'm due for a CPU upgrade soon :goog_smile_face_eyes:

Like i already said, if this is true, pci-e 4.0 was more useless then ever and just pushed by amd to get some flack on intel.

Will be interesting to see how AMD will hold up with there next cpu's or if its going to be a 3000/9000 series situation again.
 

Kenpachii

Member
PCIE5 won't be needed for much things even after Zen 5 launches.

Hell, PCIE4 and even PCIE3 NVMEs are pretty much useless for gaming right now, they have almost the same performance as SATA drives. Direct Storage will change that but still, Xbox speed of 2GB/s will be the lowest common denominator...

And GPUs are still below PCIE3 limits at this point.

That said, AL looks very interesting.

It's not about gpu's its about SSD's at this point.

We know directstorage is going to be pci-e 3.0 focused, 4.0 would be nothing but a improvement over it but if 5.0 will release the only advantage 4.0 had instantly dissapears in tin air.

In my opinion, 4 should have never been launched. It makes the whole industry a mess. I rather had them focus on 3.0 longer and straight move to 5.0 exactly what intel said back in the day. but AMD needed a reason for people to stick with it after there lackbusting 3000 reveal performance wise. Sadly people bought into it because PS5 and 5,5gbps ssd so they wanted to 1 up the competition.

As of now the only good cpu lineup AMD created that outperformed is 5000 series, the 3000 was only interesting for a budget standpoint. but because they also killed the value of those cpu''s drastically by upping the price, AMD can officially suck a nut.

I got the feeling its going to be a 3000/9000 situation all over again. WIth intel being the better choice even if amd could compete performance wise. 5.0 is going to be the new focus for PC high end gaming. That is, if directstorage and all the console crap nvme > gpu shit is actually going to mean something for PC gaming.
 
Last edited:

Rikkori

Member
After seeing their talks about programming for new cores & what "assumptions should be discarded" I'm more convinced than ever that Alder Lake will disappoint, at least for gaming. I think we're going to see some particularly awful results with older games. Alas.

Stack cache or GTFO.

Money Hollywood GIF by Anna Moon
 
No doubt the big cores will beat current Amd, and a node bump is huge. I'm just worried about the little cores and the scheduling of such mucking things up. Cautiously optimistic here.
 

Unknown?

Member
Can we get these CPUs without Intel Management Engine like the government does or are we stuck with this all access OS that operates under Windows?
 

Armorian

Banned
I wonder if he changed his BIOS settings to take advantage of x4, x2 instead of SATA III. I'm fairly certain this is overlooked at times.



It's like that because without Direct Storage all games are CPU limited, and standard SATA drives are fast enough to stay within that limit. People are buying NVMEs and not really using their speeds because outside of some specific tasks (vopy-paste) they are not faster for use than SATA drives. I have PCIE3 NVME, I bought because it was cheaper than SATA drives at the time - it's 4x times faster than my S3 SSD and yet it performs the same in games.



THIS is how jump from SATA SSD to NVME should look like with DX Storage:

 


It's like that because without Direct Storage all games are CPU limited, and standard SATA drives are fast enough to stay within that limit. People are buying NVMEs and not really using their speeds because outside of some specific tasks (vopy-paste) they are not faster for use than SATA drives. I have PCIE3 NVME, I bought because it was cheaper than SATA drives at the time - it's 4x times faster than my S3 SSD and yet it performs the same in games.



THIS is how jump from SATA SSD to NVME should look like with DX Storage:


Ok, so if the Storage is CPU bound (which should change via Direct Storage) then whatever performance and bandwidth gains these new processors have due to PCIe 5.0 should be significant.
 

Kenpachii

Member
Its not CPU limited, its software limited. Software doesn't allocate all cores to the load proces otherwise people with 5950x would see a massive gain from people with a 4/4 thread cpu. which isn't the case.
It's the same why games don't allocate memory to the ram to speed things up, because devs give zero shits.

3rd party's only started to care when sony started to push things forwards, because nobody was complaining about them anyway on PC. i would argue connections are bigger issue's then load times in games.

Still nvme drivers have there uses over sata with microstutter. Less chance of data getting strung up specially if you do more things at the same time with your gaming.
 
Last edited:

SantaC

Member
Ok, so if the Storage is CPU bound (which should change via Direct Storage) then whatever performance and bandwidth gains these new processors have due to PCIe 5.0 should be significant.
Nobody knows the gains because direct storage isnt out yet and nobody knows when it is coming. We havent even seen it demonstrated in a game yet.
 
Lol calling it Intel 7 when it is on 10nm
Intel 10nm isn't bigger than AMD 7nm, you're just measuring wrong | PC Gamer

Intel is likely to be keen on adopting a new standard metric as it tends to come out on top in raw transistor density as it stands today. Intel reports a density of 100.76MTr/mm2 (mega-transistor per squared millimetre) for its 10nm process, while TSMC's 7nm process is said to land a little behind at 91.2MTr/mm2 (via Wikichip). Not that it'll do you much good on desktop—Intel's yet to produce 10nm desktop processors. AMD Ryzen on the other hand….
Intel processors are already toe to toe with AMD at the aging14nm+(repeating) process. At 10nm they're only 9.56MTr/mm2 more dense than AMD. I expect to see a huge performance increase from Intel.
 
Last edited:

UltimaKilo

Gold Member
"Intel 7"

"Guys we don't need to move to 7nm if we just start calling our 10nm process 7nm! Nobody will ever notice!"

veronica mars wink GIF


Also I hope they change the product naming scheme itself. I mean..

Intel® Core™ i5‑1145G7E6

Really rolls off the tongue!
To be fair, they should have done this a long time ago. Their density is above anyone else on the market who uses the nanometer moniker for marketing.
 

JohnnyFootball

GerAlt-Right. Ciriously.
All kidding aside, this does look like it could be Intel's next 2600K. The CPU that just crushes it in terms of value. I look forward to this coming out.

I'd like to think that Lisa Su is too smart to just sit idle and let Intel take back the performance crown. AMD most likely has some tricks up their sleeves.

For example, they have a TON of room to drop the prices of their 5000 series.
 
I hope that table is accurate and we can slot Raptor, if its a tangible improvement over Alder into z690 boards. If so I might just get 12 series and not wait for Zen 4, coz you're already getting the holy triplet DDR5, PCI5, Gen5 NVME. Who knows how far away Zen 4 is.

ihopeitstruegfdgdrggr.png
 

SantaC

Member
Intel 10nm isn't bigger than AMD 7nm, you're just measuring wrong | PC Gamer


Intel processors are already toe to toe with AMD at the aging14nm+(repeating) process. At 10nm they're only 9.56MTr/mm2 more dense than AMD. I expect to see a huge performance increase from Intel.
TMSC has improved lately, it was back a few years ago when AMD used glofo it was true that their node process was behind.

Now TMSC is so godly even intel wants to use them because their own fabs are shit.
 
Last edited:

Md Ray

Member
Intel Alder Lake will support DDR4 as well as DDR5 memory technology.
Interesting. How about GDDR6? Hear me out...

How about a new class of desktop/laptop PC for gaming purposes with GDDR6 as system memory, as a new option? The industry will have to come together to create this new standard:

1. Make GDDR6 memory sticks available in the same form factor as DDR4/5 RAM sticks with different configs 8/16GB, etc.
2. Create motherboards to be able to accept GDDR6 type memory, implement beefier power delivery, VRMs to support powerful APUs (more on this below)
3. Implement GDDR6 memory controller in the SoC
4. Launch powerful APUs with big GPUs integrated with the CPU (I know, atm AMD is in a better position to pull something like this off than Intel)
with higher TDP in mind.

Right now, AMD APUs in the desktop space cap out at 8 Vega CUs. We'll need higher CU count than that for gaming, imagine 8C/16T and 6C/12T CPUs being integrated with 36 CU, 40 CU (6700 XT), 60 CU (6800), 72 CU (6800 XT) GPUs.

The architecture will be very similar to consoles, with more horsepower. It should also be cheaper for the end-user to build than a similarly specced dGPU system. For one, It eliminates the need for a separate DDR4, no need to pay for it. You get a single pool of memory (devs will love it). Most of the cost that goes into a dGPU will now be included with the CPU, the rest will shift towards the motherboard and the CPU cooler you get. I'm not saying we should ditch DDR4/5 and dGPUs. It's just something that occurred to me after looking at systems like Subor Z Plus and AMD 4700S kit running Windows, would be cool to see something like this become a proper standard that's geared specifically towards the PC gaming crowd who only just want to game on it.
 
Last edited:

Md Ray

Member
Its not CPU limited, its software limited. Software doesn't allocate all cores to the load proces otherwise people with 5950x would see a massive gain from people with a 4/4 thread cpu. which isn't the case.
It's the same why games don't allocate memory to the ram to speed things up, because devs give zero shits.

3rd party's only started to care when sony started to push things forwards, because nobody was complaining about them anyway on PC. i would argue connections are bigger issue's then load times in games.

Still nvme drivers have there uses over sata with microstutter. Less chance of data getting strung up specially if you do more things at the same time with your gaming.
It is also CPU-limited according to Microsoft and NVIDIA. Fast loading also involves asset decompression, this asset decompression at NVMe's multi GB/s speeds require a lot of cores. This is why they're introducing RTX I/O to offload the decompression work to GPU.
 
Last edited:

Armorian

Banned
Ok, so if the Storage is CPU bound (which should change via Direct Storage) then whatever performance and bandwidth gains these new processors have due to PCIe 5.0 should be significant.

It has more to do how data has to go trough CPU to GPU, from one memory to the other etc. DXStorage will make data go directly to the GPU. PCIE 4.0 will be more than enough for DXS to work properly, I'm not sure about 3.0

Oh and dev support, they usually don't give a shit about new apis/features for years (DX12 is from 2015, only recently new AAA games that use it are in majority)...

Its not CPU limited, its software limited. Software doesn't allocate all cores to the load proces otherwise people with 5950x would see a massive gain from people with a 4/4 thread cpu. which isn't the case.
It's the same why games don't allocate memory to the ram to speed things up, because devs give zero shits.

3rd party's only started to care when sony started to push things forwards, because nobody was complaining about them anyway on PC. i would argue connections are bigger issue's then load times in games.

Still nvme drivers have there uses over sata with microstutter. Less chance of data getting strung up specially if you do more things at the same time with your gaming.

Yeah it's CPU limited, mostly by single core/IPC but in many games you can see that during loading they are using as many cores as possible, even if in gameplay games are using 4 or 8 threads. Decompressing data is heavy task for CPU, you can see how Stem fucks up CPU if you have fast enough internet.
 

Md Ray

Member
Nobody knows the gains because direct storage isnt out yet and nobody knows when it is coming. We havent even seen it demonstrated in a game yet.
We kind of see it demonstrated by NVIDIA using some software in alpha state. This is from Ampere whitepaper from Sept. 2020, btw.

Level load time: 38 sec w/ HDD vs. 1.5 sec w/ NVMe.
U0SesRJ.png


This is the same kind of boost we see on PS5 today:

Y5Sq1gK.jpg
 

Kenpachii

Member
It is also CPU-limited according to Microsoft and NVIDIA. Fast loading also involves asset decompression, this asset decompression at NVMe's multi GB/s speeds require a lot of cores. This is why they're introducing RTX I/O to offload the decompression work to GPU.


geforce-rtx-30-series-rtx-io-announcing-rtx-io.jpg


RTX I/O is more efficient and takes barely a hit on your CPU cores. 0,5 cores is nothing and that's for 14gbps, which funny enough is pci-e 5.0. its also the reason i never cared for 4.0 after seeing it the slide. They are very much going to focus on pci-e 5.0 and a new series of gpu's that will be focused around it.

Still even without this, asa u can see CPU cores can easily compress a nvme drive with 3,5gbps without effort. Sure it costs cpu cores, but what are you using those cpu cores for with load screens anyway? u can slam them all to 100%.

The reason why they don't do it, is because nobody even cares about loading speeds until sony moved forwards with it. The reason sony cares = because they are heavily relaying next gen on this shit.PC? not so sure about that. As i didn't hear anybody complain about loading speeds ever ( obviously game design could change entirely so yea there is that ofcourse) .

That is if directstorage is getting ignored largely on the PC platform. It could be a revolution or just more of the APU bullshit we heard last generation. We still have no proof how much that SSD in sony is actually doing and nobody seems to provide any insight on this because that sony SSD is most likely not doing very much at the end of the day. So this could also mean that rtx i/o will largely be a minor improvement at the end of the day. Could obviously be a big one if it gets used a lot.

We will see tho.

It has more to do how data has to go trough CPU to GPU, from one memory to the other etc. DXStorage will make data go directly to the GPU. PCIE 4.0 will be more than enough for DXS to work properly, I'm not sure about 3.0

Oh and dev support, they usually don't give a shit about new apis/features for years (DX12 is from 2015, only recently new AAA games that use it are in majority)...



Yeah it's CPU limited, mostly by single core/IPC but in many games you can see that during loading they are using as many cores as possible, even if in gameplay games are using 4 or 8 threads. Decompressing data is heavy task for CPU, you can see how Stem fucks up CPU if you have fast enough internet.

Maybe u play different games then i do, but the only game i noticed it was ac unity from back in the day after that i really don't see it happening. Not even sure if newer ac games do it for example.
 
Last edited:

Md Ray

Member
geforce-rtx-30-series-rtx-io-announcing-rtx-io.jpg


RTX I/O is more efficient and takes barely a hit on your CPU cores. 0,5 cores is nothing and that's for 14gbps, which funny enough is pci-e 5.0. its also the reason i never cared for 4.0 after seeing it the slide. They are very much going to focus on pci-e 5.0 and a new series of gpu's that will be focused around it. the 3000's feels a bit like a old gen on steroids card, the v-ram kinda also helps with that vision.

Still even without this, asa u can see CPU cores can easily compress a nvme drive with 3,5gbps without effort. Sure it costs cpu cores, but what are you using those cpu cores for with load screens anyway? u can slam them all to 100%.

The reason why they don't do it, is because nobody even cares about loading speeds until sony moved forwards with it. The reason sony cares = because they are heavily relaying next gen on this shit.PC? not so sure about that. As i didn't hear anybody complain about loading speeds ever ( obviously game design could change entirely so yea there is that ofcourse) .

That is if directstorage is getting ignored largely on the PC platform. It could be a revolution or just more of the APU bullshit we heard last generation. We still have no proof how much that SSD in sony is actually doing and nobody seems to provide any insight on this because that sony SSD is most likely not doing very much at the end of the day. So this could also mean that rtx i/o will largely be a minor improvement at the end of the day. Could obviously be a big one if it gets used a lot.

We will see tho.
But you can see when reading compressed data, it's taking up all of 24 cores of a Threadripper CPU. Games are stored in a compressed format in your drive, as a result, a typical gaming PC with 6-8 cores and fast NVMe drives with DirectStorage's optimization in place is going to fully saturate all those CPU cores without a dedicated hardware decompressor (RTX IO).
 
Last edited:

Kenpachii

Member
But you can see when reading compressed data, it's taking up all of 24 cores of a Threadripper CPU. Games are stored in a compressed format in your drive, as a result, a typical gaming PC with 6-8 cores and fast NVMe drives with DirectStorage's optimization in place is going to fully saturate all those CPU cores without a dedicated hardware decompressor (RTX IO).

My point is a 8 core cpu can load faster then a 4 core cpu. If the software was taking it into account. That's the whole thing. I see no difference in loading speeds on a 4 core versus a 8 core. Unless some games came out after i last checked that do exactly that, and then i am talking about 95% of the games not the out outlier that nobody cares about like a game like godfall ( no clue if godfall does it by the way ).
 
Last edited:

Md Ray

Member
My point is a 8 core cpu can load faster then a 4 core cpu. If the software was taking it into account. That's the whole thing. I see no difference in loading speeds on a 4 core versus a 8 core. Unless some games came out after i last checked that do exactly that.
I'm not talking about current games. None of the current games tap into NVMe's full bandwidth which is why you see identical load times between SATA and Gen4 SSD.

To leverage these untapped NVMe speeds, DirecStorage is required (a software component i.e. storage API). Once you get DStorage in place to utilize these NVMe drives to their full potential, it is sort of like "opening the floodgates" which results in overwhelming the CPU cores as you can see in the graph you linked (reading compressed data at NVMe speeds overwhelms all of 24 cores).

This is what we mean by "CPU-limited" and why GPU decompression is introduced in the PC space.
 
Last edited:

Kenpachii

Member
I'm not talking about current games. None of the current games tap into NVMe's full bandwidth which is why you see identical load times between SATA and Gen4 SSD.

To leverage these untapped NVMe speeds, DirecStorage is required (a software component i.e. storage API). Once you get DStorage in place to utilize these NVMe drives, it is sort of like "opening the floodgates" which results in overwhelming the CPU cores as you can see in the graph you linked (reading compressed data at NVMe speeds overwhelms all of 24 cores).

This is what we mean by "CPU-limited" and why GPU decompression is introduced in the PC space.

I agree with that.
 
After seeing their talks about programming for new cores & what "assumptions should be discarded" I'm more convinced than ever that Alder Lake will disappoint, at least for gaming. I think we're going to see some particularly awful results with older games. Alas.

Stack cache or GTFO.

Intel already has more than one kind of miss and match dies and stacking working that goes beyond just adding cache.
 

A.Romero

Member
Whatever they can bring to the table to push tech over is welcomed. Many times sockets, lanes and all that are pushed to market before they are needed but an installed base is required so manufacturers actually build devices for it.

If it was up to some people we would be still using the first commercially available USB ports or something...

I'll probably upgrade in 2022 and I rather go for an Intel/Nvidia combo so this sounds good.
 

Melon Husk

Member
"Intel 7"

"Guys we don't need to move to 7nm if we just start calling our 10nm process 7nm! Nobody will ever notice!"

veronica mars wink GIF


Also I hope they change the product naming scheme itself. I mean..

Intel® Core™ i5‑1145G7E6

Really rolls off the tongue!
I thought you were joking, but no, it's actually called that.


The Intel i5-numpad smash.
Intel 7 i5 (on 10 nanometers) Inside™.
 

supernova8

Banned
I thought you were joking, but no, it's actually called that.


The Intel i5-numpad smash.
Intel 7 i5 (on 10 nanometers) Inside™.

Yeah AMD has done a great job of (mostly streamlining the names).

Of course it might be confusing next gen if they have 6900 XT CPUs alongside this gen's 6900 XT GPUs. GPU and CPU naming schemes are a bit too close imo. Then again.. better than "Radeon 7" or "Vega 64" which sounded stupid.
 
It has more to do how data has to go trough CPU to GPU, from one memory to the other etc. DXStorage will make data go directly to the GPU. PCIE 4.0 will be more than enough for DXS to work properly, I'm not sure about 3.0

Oh and dev support, they usually don't give a shit about new apis/features for years (DX12 is from 2015, only recently new AAA games that use it are in majority)...



Yeah it's CPU limited, mostly by single core/IPC but in many games you can see that during loading they are using as many cores as possible, even if in gameplay games are using 4 or 8 threads. Decompressing data is heavy task for CPU, you can see how Stem fucks up CPU if you have fast enough internet.

I read how we can't even saturate PCIE 4.0 let alone 5.0. I know the prospects for 5.0 are looking sick af. To put it plainly. I almost regret not waiting, but I'm tired of waiting. I waited long enough and while I wasn't able to get ahold of a 3090 at a reasonable price... I feel alright with what I got. I had to pull the trigger at some point and I changed my PC part picker profile numerous times. I had to lay down the fucking law and just DO IT. (also Fuck Nike)
 

Intel Alder Lake CPU Power Requirements Unveiled - 12th Gen CPUs More Power Hungry Than 10th & 11th Gen Processors​

The 12th Gen Intel Alder Lake CPUs will be featuring the brand new 'Intel 7' process node (formerly 10nm Enhanced SuperFin). This will be a major departure from the existing 14nm process that has been in use since Skylake. Based on the latest numbers from FCPOWERUP, Intel has disclosed the power requirements of its Alder Lake Desktop CPUs and it looks like the power requirements have gone up.

iUFh05T.gif
 
Top Bottom