News AMD Announces Ryzen 7000X3D Pricing: $449 to $699 Starting Feb 28th

[HideOut]

The much larger factor is the clock speed especially the boost clocks. Now what everyone is waiting to see is how they manage to pull this off. The rumor is the cores that can do the high boost do not have the 3d memory where all the cores on the 7800 have 3d memory.
So now how does the OS or whatever know what application need cores with the memory and which do better with the higher clock. Almost going to need some menu that lets the user define this.

The 7900x3d is only $50 more than the 7900x but the 7900x sells for much less than MSRP

You clearly have no clue what you are talking about. ALL CORES HAVE 3D MEMORY. Thats the point of these chips.

 

[drivinfast247]

You clearly have no clue what you are talking about. ALL CORES HAVE 3D MEMORY. Thats the point of these chips.

Ummmm no they don't! Only one CCD will have 3d cache.

 

[kostas321]

Based on these prices, if the actual 30-40 game tests don't show a massive +40% improvement (proportional to the price increase), then I'm going for a 13700KF or 13900KF.
I only use my home PC for remote work and some lighter gaming, but as long as I'll have to buy this year 5 new computers for my family with the goal to have them last 8-10 years... then these X3Ds are way too overpriced to be worth it. Also, with the launch by Easter or Black Friday the 13900KF might be even cheaper.

Current prices: 13700 KF (380$), 7900X (430$) and from what they say 79003D (600$... even the 13900KF (570$) looks good

Sorry AMD. I totally wanted to move from my 5800x to the new 7900X3D but the prices are absurd (at least here in Switzerland). The motherboard is 80-100$ more expensive than the Intel version, the CPUs are another 50-80$ more expensive and the ram is 40-50$ more expensive (for 64 GB). It just makes no financial sense overpay by 170-230$ for a new build and multiply that by the 5 computers I am replacing. Bring the big US discounts to Europe and we can talk. Otherwise you missed your chance and I will buy intel + nvidia, after 17 years of buying only AMD products for my family.

Lastly, on the professional side, I've been working in IT procurement for +20 years. Every time we end up buying Intel and Nvidia because AMD products are barely available (unreliable supply), for the first year of use the drives are very unstable and price wise in Europe Intel is always significantly cheaper. There is simply no contest when you buy 200 high end work stations and 1000 servers in one purchase, as you can never defend AMD's +20% price bill. I've been hoping that the once every 2 year / blue moon discount offers make it to the corporate side, but they never have in Europe. They offer discounts n the US but that means nothing for me as 80% of our employees are in the EU so we buy the same machines for everyone.

 

[jeremyj_83]

HDD's major selling point is Price per GiB or Price per TiB, maintaining that while closing the gap on Sustained Read/Write speeds and offering value with Optane is what matters.

For a technology that is primarily used for bulk storage, there is very little use (at least on the consumer side) to make it much faster than it already is. If you get the right HDD you can get sustained R/W for sequential data up to 225MB/sec, with higher bursting speeds. Unless you are doing daily full backups to a local drive, that R/W is plenty for anything else the drive will be doing. Most of the files are going to be movies, music, or photos. None of those are going to be significantly improved using Optane for caching. On top of this if you are using the HDD as a network drive you will be network limited before you saturate the drive, especially if it is in a RAID. Really unless you are still using a HDD as your OS drive there is little need for a better caching device right now for HDD for consumers. On the enterprise side this could be useful. Instead of using SSD caching at the outer layer for your NAS you could have it on the drive side. Having a 20TB dual actuator drive with 128GB Optane cache in a RAID 50 array could speed up backups quite a bit. In fact you could easily find yourself network limited if you are only running 10GbE to the NAS. Overall as someone who works with storage and servers professionally I just don't see a need for Optane on the consumer side. It is expensive and there are other options available that are almost as good but much cheaper. That changes completely on the enterprise side where it is the best caching device we have.

 

[Maebius]

I'm not sure what to think about this.
While I'd really like to upgrade my 7700K, prices in all parts have really crept up, while having exploded for graphics cards of course.
The slow demise of the PC market is looking like a self fulfilling prophecy.

 

[Kamen Rider Blade]

For a technology that is primarily used for bulk storage, there is very little use (at least on the consumer side) to make it much faster than it already is. If you get the right HDD you can get sustained R/W for sequential data up to 225MB/sec, with higher bursting speeds. Unless you are doing daily full backups to a local drive, that R/W is plenty for anything else the drive will be doing. Most of the files are going to be movies, music, or photos. None of those are going to be significantly improved using Optane for caching. On top of this if you are using the HDD as a network drive you will be network limited before you saturate the drive, especially if it is in a RAID. Really unless you are still using a HDD as your OS drive there is little need for a better caching device right now for HDD for consumers. On the enterprise side this could be useful. Instead of using SSD caching at the outer layer for your NAS you could have it on the drive side. Having a 20TB dual actuator drive with 128GB Optane cache in a RAID 50 array could speed up backups quite a bit. In fact you could easily find yourself network limited if you are only running 10GbE to the NAS. Overall as someone who works with storage and servers professionally I just don't see a need for Optane on the consumer side. It is expensive and there are other options available that are almost as good but much cheaper. That changes completely on the enterprise side where it is the best caching device we have.

The next major performance gains to be had come from 2x Major Areas:

- I want Multi-Actuator to become standard since the one major aspect of HDD's that has been lacking is significant performance improvements.
Multi-Actuator gets you more performance in RAID 0 or whatever RAID you want to use, WD/Seagate needs to build that into the drive for the average consumer.
We'll start at Dual Actuator and gradually add more stacks of Arms, maybe add in a entire second arm stack, we'll figure it out over time.
Eventually, I want a gradual roadmap to every Actuator arm being fully independent to be combined together to improve Sustained Read/Write Throughput, enough to eventually satisfy SAS-4's 2.8 GB/s per second R/W bandwidth when you get larger drives with many arms and platters with smaller platter drives satisfying lower SAS-3's 1.5 GB/s throughput.

- Optane gives you SSD like performance and a large SLC cache like area, something HDD's don't have. When the average customer (not us enthusiasts) are expecting basic SSD like performance, and they don't understand HDD technology, you need something to play catch up. Optane is the cheapest way to do that IMO. Optane is only expensive if you don't have it in bulk manufacturing and attached to all major HDD's. Optane would make a great giant Buffer cache that once it's filled, it'll be easier to optimize the Read/Writes for Optimized Bulk Sustained Writes and allow high responsive-ness for Random I/O once everything is cached. At the end of the day, I want HDD's to be enhanced in every perfomrance metric as possible. Optane will do that and is proven to work well for my intended goal.
I'm glad that on the Enterprise side, Optane will work great as a built in Cache buffer, so we can find use for it on both Enterprise & Client side.

- The OptiNAND tech from WD is good, that helps alleviate the amount of writes you need to do on the HDD, that should be kept.

 

[Eximo]

I'm not sure what to think about this.
While I'd really like to upgrade my 7700K, prices in all parts have really crept up, while having exploded for graphics cards of course.
The slow demise of the PC market is looking like a self fulfilling prophecy.

Depends on how you look at it.

A i3-12100 outperforms a 7700K for about a third the cost. (Intel 600 series motherboards are getting pretty low in price too) Conceivably you could upgrade platforms, gain that performance, and then wait for used prices for 12th or 13th gen chips to get low enough.

GPUs are a bit off at the moment. But we still have the RX6600->6650XT which are great budget gamers.

Only at the high end and lowest end has performance per dollar seen a huge hit. There is always a premium for high end parts. Aside from Intel, not really any decent low-end GPUs. RX 6500XT is just flawed with its 4x lanes and no encoder.

 

[Maebius]

Depends on how you look at it.

A i3-12100 outperforms a 7700K for about a third the cost.....

Well, when I got the 7700K, I got pretty much all the "reasonably" almost best stuff .

By reasonably best I mean a top tier 1080ti (not a Titan X), Asus Maximus Hero mobo (not Formula that was the best-then), 3600 ram (not 4133 that was hardly worth it for the price) and so on.

In the spirit of really moving forward and not just take a small-ish step up (like getting a 12100 as you suggested), at this point, a similar... goal, is out of the question.

As you said, even the low end has taken a huge hit.If I had a kid that wanted to game now, I'd get him a Xbox/PS and a laptop for homework and stuff. There's no way I'd get a desktop.
The next good attempt with cloud gaming should put the final nail in the PC market's coffin, as all we'll really need is a good screen and a good connection.

 

[Ogotai]

The next good attempt with cloud gaming should put the final nail in the PC market's coffin

ill belive that when i see it. " pc gaming is dead " has been tossed around for years now, yet, its still here. console gaming is ok, it has its place beside PC gaming, but, for me at least, the PC is where i spend 90% of my game time. thats mostly cause consoles, specifically xbox and PS have never had enough games that i want to play on them, to justify the cost to buy them.

 

[cryoburner]

Well, when I got the 7700K, I got pretty much all the "reasonably" almost best stuff .

By reasonably best I mean a top tier 1080ti (not a Titan X), Asus Maximus Hero mobo (not Formula that was the best-then), 3600 ram (not 4133 that was hardly worth it for the price) and so on.

In the spirit of really moving forward and not just take a small-ish step up (like getting a 12100 as you suggested), at this point, a similar... goal, is out of the question.

As you said, even the low end has taken a huge hit.If I had a kid that wanted to game now, I'd get him a Xbox/PS and a laptop for homework and stuff. There's no way I'd get a desktop.
The next good attempt with cloud gaming should put the final nail in the PC market's coffin, as all we'll really need is a good screen and a good connection.

The i7-7700K had a US $350 MSRP at launch. Adjusted for inflation, that would be around $425 today. Currently at US online retailers, you can get an i7-13700K for just under $420, or the KF version without integrated graphics for under $400. So the prices for an i7 haven't really increased over what they were then. And even the i5-13600K/KF tends to outperform the previous generation's i7-12700K, with current street prices of around $320 and $300. And of course, the new hardware offers significantly more cores and threads than the quad-cores that were available on consumer platforms at that time, and each core is notably faster, amounting to multiple times the heavily-multithreaded performance, and probably close to double the lightly-threaded performance. If anything, I'd say the 7700K was a worse value, being the last of a few generations of processors that mostly just incremented the clock rates a little each time. With the increased competition from AMD, Intel has been steadily increasingly core counts and making other improvements in the years since.

Because of all the new product launches, I'm hoping this time next year that I can get a Ryzen 9 5900X for 250 to 270.

If anything, the 5900X should be that price now when you consider that an i5-13600KF outperforms it at most heavily-multithreaded and lightly-threaded workloads by a significant margin. There's no excuse from a performance standpoint for it being priced higher. AMD might not be producing them in any significant quantities anymore though, and the prices of remaining stock might be kept higher for those looking to upgrade an existing system.

Just (16 GiB / 32 GiB / 64 GiB / 128 GiB depending on the size of the HDD) built into every HDD would MASSIVELY help the HDD improve performance and latency along with acting like a huge Read/Write buffer and function as a sudden power down memory buffer that can flush all the Read/Write data in flight into Optane.

Optane would also MASSIVELY help Optical Disc Drives as a form of Virtual Optical Disc Changer where the contents of the Optical Disc could be dumped into Optane memory and Read/Writes would be at the speed of Optane.

Imagine how many Full CD's / DVD's / Blu-Ray's could live in 16 GiB / 32 GiB / 64 GiB of Optane.

It would make no real sense for an optical drive. Things like media discs are designed to be accessed in a mostly sequential manner, and even an optical drive's relatively slow native read speed can keep up with the streams of data that are read off of them. And there would be no reason to cache that data to ultra-fast storage when even a platter-based drive could fulfill that purpose. As for caching hard drives, it might not be bad, but a regular SSD cache could fulfill the same purpose at a fraction of the cost. And since there's little reason to use hard drives for anything but bulk storage these days, even that probably wouldn't be all that beneficial in most usage scenarios. For bulk-storage, the cost-per-GB tends to be more important than performance.

 

[headloser]

AMD announced the availability and pricing for three new Ryzen 7000 processors with the revolutionary 3D V-Cache tech.

AMD Announces Ryzen 7000X3D Pricing: $449 to $699 Starting Feb 28th : Read more

That great but i waiting for the built-in GPU. I using Ryzen 5600G. The built-in GPU is a life saver. I can barely play Destiny 2 but it is payable. WHEN the hell they going to release a built-in GPU version?

 

[jeremyj_83]

That great but i waiting for the built-in GPU. I using Ryzen 5600G. The built-in GPU is a life saver. I can barely play Destiny 2 but it is payable. WHEN the hell they going to release a built-in GPU version?

All the 7000 series have a built in GPU. It only has 2CUs so won't do anything gaming wise, but it does have the ability to run without a dGPU.

 

[jeremyj_83]

The next major performance gains to be had come from 2x Major Areas:

- I want Multi-Actuator to become standard since the one major aspect of HDD's that has been lacking is significant performance improvements.
Multi-Actuator gets you more performance in RAID 0 or whatever RAID you want to use, WD/Seagate needs to build that into the drive for the average consumer.
We'll start at Dual Actuator and gradually add more stacks of Arms, maybe add in a entire second arm stack, we'll figure it out over time.
Eventually, I want a gradual roadmap to every Actuator arm being fully independent to be combined together to improve Sustained Read/Write Throughput, enough to eventually satisfy SAS-4's 2.8 GB/s per second R/W bandwidth when you get larger drives with many arms and platters with smaller platter drives satisfying lower SAS-3's 1.5 GB/s throughput.

- Optane gives you SSD like performance and a large SLC cache like area, something HDD's don't have. When the average customer (not us enthusiasts) are expecting basic SSD like performance, and they don't understand HDD technology, you need something to play catch up. Optane is the cheapest way to do that IMO. Optane is only expensive if you don't have it in bulk manufacturing and attached to all major HDD's. Optane would make a great giant Buffer cache that once it's filled, it'll be easier to optimize the Read/Writes for Optimized Bulk Sustained Writes and allow high responsive-ness for Random I/O once everything is cached. At the end of the day, I want HDD's to be enhanced in every perfomrance metric as possible. Optane will do that and is proven to work well for my intended goal.
I'm glad that on the Enterprise side, Optane will work great as a built in Cache buffer, so we can find use for it on both Enterprise & Client side.

- The OptiNAND tech from WD is good, that helps alleviate the amount of writes you need to do on the HDD, that should be kept.

Again I ask why. WD's dual actuator drives have a lower total writes per year than the single actuator ones (500TB vs 550TB). The additional actuators increase the BOM, more moving parts in the drive to fail, more heat, more noise, and more power. It is true that a dual actuator drive uses less power than two separate drives, but you will still want a redundant setup at least in enterprise. Now you want to make each platter have an independent actuator just doesn't make sense. You will still be slower than an SSD in any level QD IOPS, your seek time will still be higher (even with caching), etc... Not to mention that SSD prices are coming down all the time. How many people who use their computers for just normal desktop use need 14TB of storage? How about 4TB? Answer not that many. A 2TB Western Digital Black SN770, this is a PCIe 4 drive, runs $125 right now on Newegg. For a good 90% of people 2TB is enough storage, that includes most people who use their computers for gaming. I have a 2TB SSD on my desktop with 24 games installed, sure some are older, but I am not even using 1TB. The only really good use for HDDs on the consumer side now is cheap bulk storage. Files that are going to mainly be sequential in nature (iso's of movies or backups) or a bunch of small (less than 100MB) files (MP3s, pictures, documents). For any of these Optane isn't going to be useful because even running at a slow transfer rate of 125MB/sec (that of a 2.5" 5400RPM USB3 drive) is more than enough to stream even 8k video (only needs a 50Mbps connection, granted that will need to be higher for other things but you get the point). That leave us with going back to HDDs for our OS drive. That means laptops will be thicker, heavier, less battery life for same battery size, less endurace, and feel slower. Desktops will use more power, have less endurance, and feel slower. Do you want any of those things? As great of a technology as Optane is, outside of Enterprise solutions there isn't a good use for it. Even those uses in Enterprise are limited to niche write caching scenarios in areas that are going to have a lot of low QD writes.

 

[Kona45primo]

That great but i waiting for the built-in GPU. I using Ryzen 5600G. The built-in GPU is a life saver. I can barely play Destiny 2 but it is payable. WHEN the hell they going to release a built-in GPU version?

7 series including the X3D series have built in GPU's:

Graphics Capabilities
Graphics Model
AMD Radeon™ Graphics
Graphics Core Count
2
Graphics Frequency
2200 MHz

And how did this thread become an optane/storage discussion. Optane is dead because Intel lost billions of dollars trying to make it profitable. Didn't happen.

Focus on what's important, AMD taking the lead again for a period of time.

 

[TerryLaze]

And how did this thread become an optane/storage discussion. Optane is dead because Intel lost billions of dollars trying to make it profitable. Didn't happen.

Focus on what's important, AMD taking the lead again for a period of time.

Because intel is ready to release CPU max which will have inbuilt cache and not a few hundred Mb like ryzen but 64Gb, and it will not only work as cache but will also do the things intel wanted optane for.
Mainly also work as system ram or as a caching device for other storage.
So of course, just like you, everybody had to chip in on how intel "fumbled billions on optane"
The period of time you are talking about is basically already over at least for professional work.
Intel isn't as desperate as AMD to release server tech at desktop prices yet, but it will trickle down eventually.

 

[Kona45primo]

Now that is VERY interesting. Thanks for the info. You gave me something to learn about. Sounds like Intel will be hot on AMD's heels when this trickles down. Gotta love a competitive marketplace

 

[Kamen Rider Blade]

Because intel is ready to release CPU max which will have inbuilt cache and not a few hundred Mb like ryzen but 64Gb, and it will not only work as cache but will also do the things intel wanted optane for.
Mainly also work as system ram or as a caching device for other storage.
So of course, just like you, everybody had to chip in on how intel "fumbled billions on optane"
The period of time you are talking about is basically already over at least for professional work.
Intel isn't as desperate as AMD to release server tech at desktop prices yet, but it will trickle down eventually.

Intel "Bought the HBM" memory from the Memory Vendors and used it.

They didn't manufacture it themselves.

HBM can work as L4$, it can work as RAM, it's user configurable at the end of the day because Intel wanted it for that purpose.

But don't kid yourself on the speed of HBM, which is DRAM vs SRAM.

Them screwing up on Optane is a seperate issue all together from their "Max-Line" of Server CPU's.

Don't kid yourself on thinking they're related.

 

[Kamen Rider Blade]

It would make no real sense for an optical drive. Things like media discs are designed to be accessed in a mostly sequential manner, and even an optical drive's relatively slow native read speed can keep up with the streams of data that are read off of them. And there would be no reason to cache that data to ultra-fast storage when even a platter-based drive could fulfill that purpose. As for caching hard drives, it might not be bad, but a regular SSD cache could fulfill the same purpose at a fraction of the cost. And since there's little reason to use hard drives for anything but bulk storage these days, even that probably wouldn't be all that beneficial in most usage scenarios. For bulk-storage, the cost-per-GB tends to be more important than performance.

Well, I guess I'm at the opposite end of the spectrum and want "Speed" to cover the other aspect of what the slower mediums lack.
Optical Disc Drives & Hard Disk Drives already have the "Bulk-Storage" aspect down.
What they lack is speed or the equivalent of a SLC-cache.

That's something that Optane could bring to HDD's.

A simple Virtual Optical Disc Changer can be done with Optane.

The reason I choose Optane is for it's ultra low latency, high reliability, and performance.

But my objectives are different from yours.

 

[Kamen Rider Blade]

Again I ask why. WD's dual actuator drives have a lower total writes per year than the single actuator ones (500TB vs 550TB). The additional actuators increase the BOM, more moving parts in the drive to fail, more heat, more noise, and more power. It is true that a dual actuator drive uses less power than two separate drives, but you will still want a redundant setup at least in enterprise. Now you want to make each platter have an independent actuator just doesn't make sense. You will still be slower than an SSD in any level QD IOPS, your seek time will still be higher (even with caching), etc...

If I want redundancy, I'd buy another identical model HDD and backup the contents of the first drive onto the second drive.
As for lower total writes, I'm not worried about 550 TB -> 500 TB. That's insignificant given my use case as a average consumer.

I'm not worried about QD IOPS for a HDD. The entire point is to boost Sustained Linear R/W throughput with a HDD is what matters to me.
That's why I want more Actuators, to help boost Sustained Linear Throughput.

Not to mention that SSD prices are coming down all the time. How many people who use their computers for just normal desktop use need 14TB of storage? How about 4TB? Answer not that many. A 2TB Western Digital Black SN770, this is a PCIe 4 drive, runs $125 right now on Newegg. For a good 90% of people 2TB is enough storage, that includes most people who use their computers for gaming. I have a 2TB SSD on my desktop with 24 games installed, sure some are older, but I am not even using 1TB. The only really good use for HDDs on the consumer side now is cheap bulk storage. Files that are going to mainly be sequential in nature (iso's of movies or backups) or a bunch of small (less than 100MB) files (MP3s, pictures, documents).

SSD's are hitting a price floor and there's not much you can do about it. They're already running dangerously close to their BoM cost.
The average person's needs for storage are constantly going up, and I definitely don't fall in the "Average Person's" category when I exceed 32 TB alone of stuff stored.

For any of these Optane isn't going to be useful because even running at a slow transfer rate of 125MB/sec (that of a 2.5" 5400RPM USB3 drive) is more than enough to stream even 8k video (only needs a 50Mbps connection, granted that will need to be higher for other things but you get the point).

I was never worried about a HDD's streaming performance, that was never a concern in my mind.

That leave us with going back to HDDs for our OS drive. That means laptops will be thicker, heavier, less battery life for same battery size, less endurace, and feel slower. Desktops will use more power, have less endurance, and feel slower. Do you want any of those things? As great of a technology as Optane is, outside of Enterprise solutions there isn't a good use for it. Even those uses in Enterprise are limited to niche write caching scenarios in areas that are going to have a lot of low QD writes.

I wasn't even considering HDD's as a OS Drive in this day & age. That should be the sole domain of a pure Optane SSD.
Regular SSD's make sense for Mobile devices.
But for DeskTop computing and Home NAS/Server, cheap bulk storage is king.

Anways...

The entire point of Optane is to function as a large SLC-cache, but for the HDD.
Something HDD doesn't have and could use.

Let's say everybody is hitting the same HDD with writes, the drive controller can reorganize everything for maximum linear throughput.
Same with reads, once the data is loaded into Optane, it can be sent at Optane speeds.
Those are features that a HDD can't do without Optane, and would require a special controller card & Flash Drive to perform.
These are nice "QoL (Quality of Life)" features that the average consumer would benefit from.

As you increase the bit depth from SLC (1-bit per cell) to MLC (2-bit per cell) -> TLC (3-bit per cell) -> QLC (4-bit per cell) -> PLC (5-bit per cell).
The Total P/E (Program/Erase) cycles per cell are going down.

Spoiler: QLC's pathetic P/E cycles compared to it's Predecessors

Spoiler: As Bit-Density per cell goes up, P/E cycles goes down

Once PLC becomes standard, will 500 P/E cycles be acceptable to you as a consumer, what about 100 P/E cycles?
Will disposable SSD's be a thing where you don't care for the environment and buy SSD's to dispose of once you quickly exhaust it's low P/E cycles?

I sure as hell won't have to worry about the P/E cycles on Optane or HDD's happening anytime soon.

It's sad to see QLC drives Linear Throughput fall to the point, where once they're out of SLC-cache, they have HDD levels of performance or worse.
QLC's Write Speeds are already significantly worse than previous generations.

Spoiler: QLC's slow Write speeds relative to it's predecessors

Spoiler: QLC SSD's Sustained Steady State Write falls off the cliff once SLC-cache is exhausted

QLC Drive's Steady state write speeds fall to 80 MB/s after the SLC-cache is exhausted, is that acceptable to you?
If you find having a SLC-cache to hide the real performance of QLC drives acceptable, why can't HDD's have Optane to hide HDD's weaknesses?

QLC Drive's Read Latency are getting worse over time as you increase the density of the # of bits per cell.

Spoiler: SSD's Latency goes up with denser bits per cell

I wonder how bad things will get once you hit PLC and above.

If using SLC-cache to hide any weaknesses of QLC and more dense SSD technologies is acceptable to you, then why can't HDD use Optane in bulk to hide it's weaknesses?

Spoiler: SLC cache size in GB vs SSD Capacity

If modern SSD's can use up to 280 GB of SLC cache, why can't we add in up to 256 GB of Optane on the largest HDD's to mask any weaknesses?

 

[TerryLaze]

Intel "Bought the HBM" memory from the Memory Vendors and used it.

They didn't manufacture it themselves.

What does it matter if they bought it or made it themselves?
Is HBM under IP? Will they be able to use their future fabs to make hbm, if they want to, or not?

HBM can work as L4$, it can work as RAM, it's user configurable at the end of the day because Intel wanted it for that purpose.

And that's the great thing about these new cpus, was there ever any other CPU that supported this?

But don't kid yourself on the speed of HBM, which is DRAM vs SRAM.

What, is it slower than optane ram dimms?! No?! Then it's still a better choice than optane.

Them screwing up on Optane is a seperate issue all together from their "Max-Line" of Server CPU's.

Don't kid yourself on thinking they're related.

They could be unrelated, I don't know it but you don't know that either.
But it is clear that everything that intel wanted from optane has been achieved in an even better way.
Optane also was never a huge seller, it never took off, you just didn't buy enough disks when you had the chance.

 

[Kamen Rider Blade]

What does it matter if they bought it or made it themselves?

You keep trying to imply that Intel will manufacture HBM themselves, which they clearly won't.
Intel left the Memory Business "A LONG TIME AGO".

Is HBM under IP? Will they be able to use their future fabs to make hbm, if they want to, or not?

Somebody owns the IP, not Intel, that's for sure.
Doubtful, since Intel isn't in the Memory Business anymore.

And that's the great thing about these new cpus, was there ever any other CPU that supported this?

IBM & Intel has had L4$ before, so it's not a new concept.

What, is it slower than optane ram dimms?! No?! Then it's still a better choice than optane.

It's faster than Optane RAM DIMMs & Higher Cost & Less Capacity & doesn't retain data when you turn off the power.
It's a trade-off based on your workload.
Certain workloads really benefitted from Optanes larger capacity.

They could be unrelated, I don't know it but you don't know that either.

I'm pretty damn sure given how the technology works and their end goals for each.

But it is clear that everything that intel wanted from optane has been achieved in an even better way.
Optane also was never a huge seller, it never took off, you just didn't buy enough disks when you had the chance.

I doubt you understood what the real purpose for Optane was if you think Intel achieved their goals in a better way with HBM.

I think you're missing the point about Optane and what it's real benefits are.

 

[TerryLaze]

You keep trying to imply that Intel will manufacture HBM themselves, which they clearly won't.
Intel left the Memory Business "A LONG TIME AGO".


Somebody owns the IP, not Intel, that's for sure.
Doubtful, since Intel isn't in the Memory Business anymore.

What is this obsession with intel having to be in the memory business to make their own CPUs?
Intel wasn't in the GPU business for 20 years but still was the biggest GPU manufacturer during that time because of all of their CPUs having igpus.

It's faster than Optane RAM DIMMs & Higher Cost & Less Capacity & doesn't retain data when you turn off the power.
It's a trade-off based on your workload.

So it is faster.
Higher cost but since smaller capacity probably a wash in the grand scheme.
Doesn't retain data but since it's smaller capacity and has a direct line to a super fast nvme drive repopulating the ram should be much faster than using the slower optane ram in the first place.

Certain workloads really benefitted from Optanes larger capacity.

Sure, and we will have to look at benches to see how much those get affected.

I'm pretty damn sure given how the technology works and their end goals for each.

So am I.

I doubt you understood what the real purpose for Optane was if you think Intel achieved their goals in a better way with HBM.

I think you're missing the point about Optane and what it's real benefits are.

Hey, I'm sure a lot of people don't know about that so if you have a link share it with the class.

Until then, I go with this.
And I'm no expert, but I think that cache right on the CPU package will have the least amount of latency, even compared to optane on ram dimms.
Sure, the amount the accessible mem is much less but if oneAPI through pcie nvme can leverage that it will be a win.
https://www.intel.com/content/www/u...cing-bandwidth-and-latency-article-brief.html

 

[Kamen Rider Blade]

What is this obsession with intel having to be in the memory business to make their own CPUs?
Intel wasn't in the GPU business for 20 years but still was the biggest GPU manufacturer during that time because of all of their CPUs having igpus.

I thought you were talking about Intel manufacturing Memory to function as cache, since that's what you were talking about several posts ago.
Intel hasn't been in that business for several decades, they left because there was a serious lack of profits in that division.
There's no reason for IMEC to join that business if the profits aren't there.
And as far as the GPU business, Intel has still outsourced that production to TSMC for Arc Graphics.
So the consumer end hasn't been done in-house yet.

So it is faster.
Higher cost but since smaller capacity probably a wash in the grand scheme.
Doesn't retain data but since it's smaller capacity and has a direct line to a super fast nvme drive repopulating the ram should be much faster than using the slower optane ram in the first place.

HBM is still DRAM at the end of the day, slower in latency than traditional DRAM for Desktop, faster than GDDR# Latency.
The bandwidth for HBM still blows GDDR# & DDR# out of the water, at the downside of:

• BoM Cost
• Installation location physical limits
• Trace length limits
• PCB design complexity
• Can't keep data when the power gets shut-down like Optane (but most DRAM types suffer this flaw, but the massive amounts of data that HBM has makes it THAT MUCH HARDER to dump once you have a sudden power outage (You need a MUCH bigger Capacitor/Battery), slower Optane RAM wouldn't need to dump it to NVMe due to it's inherent properties of retaining data once you lose power, it would literally sit in Optane untouched, and ready to go on next boot. That's a incredibly hard to beat feature. The best use of energy is to not need to exert any, especially in a sudden power outage situation.)

Sure, and we will have to look at benches to see how much those get affected.

I'm sure they'll get tested, the "Serve the Home" folks made great use of Optane DIMMs and loves them.

So am I.

Ok.

Hey, I'm sure a lot of people don't know about that so if you have a link share it with the class.

Until then, I go with this.
And I'm no expert, but I think that cache right on the CPU package will have the least amount of latency, even compared to optane on ram dimms.
Sure, the amount the accessible mem is much less but if oneAPI through pcie nvme can leverage that it will be a win.
https://www.intel.com/content/www/u...cing-bandwidth-and-latency-article-brief.html

From everything I can research about HBM's latency:
- HBM2's Latency is ~(73-107) ns
And that's directly sitting next to the Memory Controller, right outside the package.
There's no more room to move HBM closer, it's as close as you can physically get without being in the die directly.

Regular DRAM is faster in terms of latency:

Spoiler: Regular DRAM latency along with various System Level Cache Latencies

Regular Main System Memory Latency is ~(59-68) ns

GDDR6 Latency is ~(200-600) ns

With that type of latency for Optane DIMMs, it would literally be the perfect OS Drive.
Imagine how fast your daily OS usage experience would be if Optane DIMMs were the "Standard OS Drive" physical format.
We could re-use the (Less used) Mini-DIMM physical format and use that as a "Dedicated" OS Drive slot to mount the Optane DIMM as dedicated memory for your OS.

 

[atomicWAR]

My problem with these cache heavy X3D CPU is that benefits are extremely game specific.

There are games that are helped tremendously by more cache because they have data intensive algorithms that are CPU bound.

But there are are also many games that don't benefit from this at all.

And since I have no idea if the games I currently play, or will even play in the near future, will benefit from this, it's just a gamble.

While true to a degree, there are some very specific gaming scenarios where the FPS uplift is pretty much a given with 3D V-cache. Those being CPU bottlenecked games which is fairly easy to test for. So if your gamer targeting 30-60FPS...3d V-cache chips are going to be overkill for you as you won't be hitting that bottleneck anytime soon at those lower frame rates. BUT if your targeting max FPS in games and are the type of person to turn down a setting to get more FPS even when you're already over 100-120FPS...you the type of gamer who 3d V-cache is going to benefit.

Now I'll give you that this is a very niche segment of gamers, though I'd argue its one of the fast growing sects of gamers due to all the cheap(ish) high refresh rate panels on the market which is why CPUs like this are gaining traction. Prior to the RTX 4090 I'd a said if your a 4K gamer these new CPUs are pointless...but as a owner of a RTX 4090 with a 7950X...I can say quite definitively that there are now a good number of games hitting a CPU bottleneck even at 4K with a RTX 4090. Which is why I'll be picking one of these up sometime in the next 6-10 months.

ALL of that said you do have a point for most gamers. These chips are really aimed at the high refresh gaming community. Anything else is likely over kill and extremely hit or miss on FPS gains.