What's new
Carbonite

South Africa's Top Online Tech Classifieds!
Register a free account today to become a member! (No Under 18's)
Home of C.U.D.

12th Gen/DDR5/Z690 Stuff...

Gouhan

Forum Addict
TheOverClocker.com
Rating - 100%
96   0   0
Joined
Apr 22, 2010
Messages
3,311
Reaction score
977
Points
5,265
Credit: From this Thread on TechPowerup

eQcOrnRPXxpYwzYE.jpg


Core i5 12400 @4.4GHz (DDR5 4800 CL40?)

V.S.

Core i9 11900K @5.4GHz (DDR4 4800 CL19)

11900K_5400.jpg

A decent improvement in IPC, especially after the many years of incremental performance gains. + 1GHz~

CPU-Z for confirmation

DGJPwPTLF2Rp1gdm.jpg



Core i5 12400 @4.4GHz (DDR5 4800 CL40?)

V.S.

Core i9 11900K @5.4GHz (DDR4 4800 CL19)


CPUZ_11900K.jpg


* For those wondering if their coolers will work, some vendors have dual mounting options/spacing on their boards so LGA1200 coolers work just as well on such boards.

* Looks like 06 again. :)
 

Attachments

  • CPUZ_11900K.jpg
    CPUZ_11900K.jpg
    101.4 KB · Views: 104
Last edited:
Info about Alder lake is pretty much everywhere. Huge list of boards already out.

FB6A99vUYAE0zFC

Most of the DDR5 boards will be expensive it seems or cost more than their DDR4 counterparts. Thinking many will opt for DDR5 versions 1st given we already have plenty of cheap & fast DDR4. Which should be faster than DDR5 in games, right now because of the much lower latency.
27th is less than 10 days away, then we should be able to see/show all sort so boards, DRAM, CPUs etc.
 
This is actually kinda saddening. For most enthusiasts this is a dead platform already. Why? Once DDR5 has matured, Z690 and Alder Lake will be past its prime already. We’ll probably be on Meteor Lake by that point.

Due to manufacturers’ infatuation with DDR5, pretty much all of the enthusiast grade boards are DDR5 - something that basically no one will opt for over something like good Micron RevB, Micron RevE, Hynix DJR, or Samsung B-Die.

All this is going to do is leave people who prefer Intel for their use cases on the 10900K, and people who want the best experience to opt for X570S + Zen 3D and likely a DDR4-4000 B-Die kit as I presume Zen 3D will further improve on the IMC and IF slightly.
 
That's fair I guess to some degree.
I don't think it's a dead platform though, unless there's something I'm not considering.

Meteor-lake will work just fine on Z690 boards, so at the very least, the boards will be relevant into 2023. Don't know about the CPU's after that however. That said, it's PCIe 5.0, which we don't even have GPUs for and lots of Gen4 M.2 possibilities. Platform for platform, there isn't one as forward looking as this one as even Zen3 with 3DVcache is still DDR4/PCIe 4.0.

Best option I think is high end or as high end as available DDR4 Z690, who knows we may see even more high end variants with DDR4 in future.
 
That's fair I guess to some degree.
I don't think it's a dead platform though, unless there's something I'm not considering.

Meteor-lake will work just fine on Z690 boards, so at the very least, the boards will be relevant into 2023. Don't know about the CPU's after that however. That said, it's PCIe 5.0, which we don't even have GPUs for and lots of Gen4 M.2 possibilities. Platform for platform, there isn't one as forward looking as this one as even Zen3 with 3DVcache is still DDR4/PCIe 4.0.

Best option I think is high end or as high end as available DDR4 Z690, who knows we may see even more high end variants with DDR4 in future.
It’s just such an odd decision. I guess in that regard with the motherboard thing, there is some level of hope. I’ve never been a fan of buying an expensive board to keep for years and years when you could just pick up some thrown-out Z490 ROG Maximus Extreme or Z490 Aorus Master at 50% of their retail price and then wait for Z790 or Z890, along with like 13th gen and some actually decent and mature DDR5.

I just feel like we as the consumer have been subjected to beta testing. I just don’t see why we would need higher density modules when little to no one even uses 32GB by this point, and few games and applications require more than maybe 4400Mbps to run well. It can be argued that you’re better off running lower latency with a 1:1 memory:controller ratio.

I’m just so confused and disappointed. I’m not trying to be that guy who consistently finds issues, I just can’t believe it’s so ‘not what I was expecting.’ If that makes sense.
 
I just can’t believe it’s so ‘not what I was expecting.’ If that makes sense.
So what WERE you expecting? This is Intel taking the reigns for the first time in almost a decade.
 
So what WERE you expecting? This is Intel taking the reigns for the first time in almost a decade.
I was hoping for DDR5 at reasonable speeds. Like maybe the average entry-level kit at 4800 22-22-22, not as many quirks with the little cores, and no weird DRM issues that have been going around over the last couple of days. The CPU stuff with the little core scheduling, DRM issues, and the reintroduction of a FIVR is neither here nor there. I'm sure Intel's genius engineers had valid reasons and considerations; why this godawful DDR5 forced adoption though?

I understand that they're introducing all this wonderful tech, but absolutely no one needs PCIe 5.0 and DDR5 at the consumer level. Most people can't even get their hands on a PCIe 4.0 capable card, let alone one fast enough to justify the necessity of said PCI spec. I currently know one person aside from myself who even daily's a 4.0 drive, and the only reason he does is that it was cheap and I parted out his build for him.

Again, not trying to be that guy who thinks he knows more than engineers, but it doesn't take a genius to realize that DDR5-4800Mbps @ 40-40-40-77-117 is gonna get absolutely hammered by even that lovely Adata Spectrix 4400 19-26-26 kit you stock on Progenix. I just wish they had foregone DDR5.

Now I will admit, I was one of those chops talking about how excited I was and that I was gonna jump on it ASAP. Now that I've seen what DDR5 truly is, I've changed my mind. I see absolutely no justification for why anyone should spend their money on a R15000-16000 fake 16-core, a R10000-12000 motherboard, and R3000 worth of 4800 40-40-40. It just seems pointless.

I am just gonna add: the only reason I've been saying this sort of stuff is because this was a thread dedicated to early leaks and leaked benchmarks, I'm not going to try and convince people not to buy Alder Lake stuff on official review and sales threads.
 
I was hoping for DDR5 at reasonable speeds. Like maybe the average entry-level kit at 4800 22-22-22, not as many quirks with the little cores, and no weird DRM issues that have been going around over the last couple of days.
* With each DRAM generation it has more ticks, but higher overall bandwidth scaling. This isn't new and has been the case with DDR - > DDR2 - > DDR3 - DDR4 - > DDR5

* You're not likely to remember all these transitions many do.
Avexir was selling DDR3 3000 and 3100 kits before DDR4 came out and that came out at DDR4 2400 and the performance kits were DDR4 2800 for a while. Today you'd not ever argue that DDR3 is faster than DDR4.
Same as with DDR5, the only difference is that the performance scaling will be much steeper or quicker if you will. When DDR5 hits it's stride, you may change your mind. :)

* DRM issues are Denuvo related as that copy protection recognizes the CPU as two separate systems. That is, it's not an Intel or OS issue but a Denuvo issue, which they are releasing patches for.

* You don't have to worry about core scheduling anymore than you worry about it now on any Intel 11/10/9th Gen or AMD Ryzen 2000/3000/5000 system. Thread director is completely hardware bound and merely reports info to the OS schedular on top of allocating loads as it sees fit. Windows schedular was inevitably going to need to be aware of the difference performance cores, especially if it has aspirations of being usable on many SOCs that use BIG.Little config. In what you use, this is unlikely to be an issue.
 
Eh? I really don't understand most of these points:

I was hoping for DDR5 at reasonable speeds. Like maybe the average entry-level kit at 4800 22-22-22,
We will get there. JEDEC spec speeds has always sucked, and RAM has always launched at or near JEDEC spec speeds. An example, DDR4's original JEDEC speed was 2133 MHz 15-15-15, while I have a kit of DDR3-2800 12-14-14. That kit of DDR3 has looooong since been left in the dust by DDR4.

not as many quirks with the little cores,
That's a software issue that affects older operating systems - much like Windows 98 didn't have a clue what to do with HyperThreading on the P4.

and no weird DRM issues that have been going around over the last couple of days.
Specific to one form of ALREADY hated DRM (Denuvo) - referring to the Pentium 4 again, we had plenty software back in the day that was also confused by HyperThreading even on an SMT aware OS such as Windows 2000/XP, where performance would suffer greatly with HT enabled. That was neither the fault of the CPU nor operating system.

The CPU stuff with the little core scheduling, DRM issues, and the reintroduction of a FIVR is neither here nor there.
You've just mentioned two of these, and FIVR has no down sides to the average user.

I'm sure Intel's genius engineers had valid reasons and considerations; why this godawful DDR5 forced adoption though?
It's NOT - have a look at upcoming boards, plenty of them are DDR4. In fact, suppliers have already indicated to me that they might bring in DDR4 boards exclusively in the beginning, depending on the price and availability of DDR5 (no point selling DDR5 boards if there's no DDR5 available in the country - several vendors have told SA there will not be launch allocation for our market due to short supply).

I understand that they're introducing all this wonderful tech, but absolutely no one needs PCIe 5.0
The same has been said about PCIe 2, 3, 4... And I disagree. PCIe lanes have always been a bottleneck at a consumer level. Look at Intel's current flagship Core i9-11900K with a massive (drumrooooooooll) 20 lanes. PCIe 5.0 would allow you to use four SSDs with a single lane each, with each drive getting enough bandwidth for the same performance as a Gen3x4 drive.

NVMe storage is really taking off. Recently (the last 4-6 months) I sell more NVMe drives and NVMe AICs for 2 or 4 NVMe drives than I sell SATA SSDs. People running two NVMe drives, a graphics card and any other AIC (be it a sound card, 10Gbit ethernet card (yes, consumers are starting to want them), a video capture card or whatever the case may be) and you've saturated the bandwidth available from both the CPU and the chipset. If we can't have more lanes, make them faster.

and DDR5 at the consumer level.
The same could be said about DDR4 - nobody NEEDS it, but there's performance potential that's missed by DDR3. The DDR5 roadmap goes all the way past DDR5-10000, and once we have Samsung releasing semi decent ICs as they have for almost every generation (DDR and DDR4 really stand out here, with TCCD for DDR and B-Die for DDR4), DDR3 will become a decent memory.

Most people can't even get their hands on a PCIe 4.0 capable card,
I do weekly sales reports for NVIDIA, so I have the stats at hand. Less than 0.15% of cards sold this year have been PCIe 3.0, the rest have all been PCIe 4.0.

let alone one fast enough to justify the necessity of said PCI spec.
Disagree due to a point I made above. Once you've run out of lanes and your card drops to 8 lanes or even 4, the bandwidth is needed even if it is a midrange card.

I currently know one person aside from myself who even daily's a 4.0 drive, and the only reason he does is that it was cheap and I parted out his build for him.
These aren't things I keep track of the way I do with graphics cards as I don't do reporting on their sales, but looking at my 50 top selling items, the Aorus Gen4 and Samsung 980 Pro make appearances. You're quite young and your friends are likely students working with student budgets, which might explain the situation you see.

Again, not trying to be that guy who thinks he knows more than engineers, but it doesn't take a genius to realize that DDR5-4800Mbps @ 40-40-40-77-117 is gonna get absolutely hammered by even that lovely Adata Spectrix 4400 19-26-26 kit you stock on Progenix.
DDR2 launch spec was DDR2-533 4-4-4 and DDR2-667 5-5-5, both of which get destroyed by DDR-400 1.5-2-2, DDR500 2.5-4-4, etc

DDR3 launch spec was DDR3-1066 7-7-7, which gets ripped a new one by DDR2-1200 5-5-5, etc.

DDR4 launch spec was DDR4-2133 15-15-15, which gets annihilated by DDR3-2800 12-14-14, DDR3-3000 12-14-14, etc.

You can't compare some of the highest speed of one generation to entry level of the next.

I just wish they had foregone DDR5.
So we'd be on DDR4 forever? DDR5 will not progress until launched as there will be no reason for progression (why better a product that doesn't exist - and how would you test it if there isn't a single platform for validation?).

Now I will admit, I was one of those chops talking about how excited I was and that I was gonna jump on it ASAP. Now that I've seen what DDR5 truly is, I've changed my mind.
This is a generation where you will be left behind. People have joked about every generation for the last few years being a rebranded Core i7-6700K with a few extra cores slapped on, but we finally have REAL progression. The performance jump over 11th Gen even with slow DDR5 or DDR4 is something that a year or two ago I would have laughed at as the last time Intel made real strides was almost a quarter of my life ago.

I see absolutely no justification for why anyone should spend their money on a R15000-16000 fake 16-core
You're quick to call it "fake" but you're running a "fake" 16 core - E cores are a lot less "fake" than HyperThreaded cores. Their performance is incredible by any standard, and they sip power. We'll be seeing performance similar to what we have now at 1/4 the power draw, or much better performance within the same power envelope we currently use.

a R10000-12000 motherboard
You don't need a R 10k to R 12k motherboard. Hell, you probably don't NEED the board you have right now.

and R3000 worth of 4800 40-40-40.
So get faster RAM - looking at the SKU sheets I have available, several vendors haven't announced anything that slow.

Reading your post for the third time, it kind of reminds me of people who were expecting Ryzen 2000 to be 50% faster at gaming than Intel while costing half as much. People set an impossible bar and were disappointed when the impossible didn't happen - Alder Lake isn't the impossible happening, but based on Intel's track record for the last few years it sure feels like it is.
 
Eh? I really don't understand most of these points:


We will get there. JEDEC spec speeds has always sucked, and RAM has always launched at or near JEDEC spec speeds. An example, DDR4's original JEDEC speed was 2133 MHz 15-15-15, while I have a kit of DDR3-2800 12-14-14. That kit of DDR3 has looooong since been left in the dust by DDR4.


That's a software issue that affects older operating systems - much like Windows 98 didn't have a clue what to do with HyperThreading on the P4.


Specific to one form of ALREADY hated DRM (Denuvo) - referring to the Pentium 4 again, we had plenty software back in the day that was also confused by HyperThreading even on an SMT aware OS such as Windows 2000/XP, where performance would suffer greatly with HT enabled. That was neither the fault of the CPU nor operating system.


You've just mentioned two of these, and FIVR has no down sides to the average user.


It's NOT - have a look at upcoming boards, plenty of them are DDR4. In fact, suppliers have already indicated to me that they might bring in DDR4 boards exclusively in the beginning, depending on the price and availability of DDR5 (no point selling DDR5 boards if there's no DDR5 available in the country - several vendors have told SA there will not be launch allocation for our market due to short supply).


The same has been said about PCIe 2, 3, 4... And I disagree. PCIe lanes have always been a bottleneck at a consumer level. Look at Intel's current flagship Core i9-11900K with a massive (drumrooooooooll) 20 lanes. PCIe 5.0 would allow you to use four SSDs with a single lane each, with each drive getting enough bandwidth for the same performance as a Gen3x4 drive.

NVMe storage is really taking off. Recently (the last 4-6 months) I sell more NVMe drives and NVMe AICs for 2 or 4 NVMe drives than I sell SATA SSDs. People running two NVMe drives, a graphics card and any other AIC (be it a sound card, 10Gbit ethernet card (yes, consumers are starting to want them), a video capture card or whatever the case may be) and you've saturated the bandwidth available from both the CPU and the chipset. If we can't have more lanes, make them faster.


The same could be said about DDR4 - nobody NEEDS it, but there's performance potential that's missed by DDR3. The DDR5 roadmap goes all the way past DDR5-10000, and once we have Samsung releasing semi decent ICs as they have for almost every generation (DDR and DDR4 really stand out here, with TCCD for DDR and B-Die for DDR4), DDR3 will become a decent memory.


I do weekly sales reports for NVIDIA, so I have the stats at hand. Less than 0.15% of cards sold this year have been PCIe 3.0, the rest have all been PCIe 4.0.


Disagree due to a point I made above. Once you've run out of lanes and your card drops to 8 lanes or even 4, the bandwidth is needed even if it is a midrange card.


These aren't things I keep track of the way I do with graphics cards as I don't do reporting on their sales, but looking at my 50 top selling items, the Aorus Gen4 and Samsung 980 Pro make appearances. You're quite young and your friends are likely students working with student budgets, which might explain the situation you see.


DDR2 launch spec was DDR2-533 4-4-4 and DDR2-667 5-5-5, both of which get destroyed by DDR-400 1.5-2-2, DDR500 2.5-4-4, etc

DDR3 launch spec was DDR3-1066 7-7-7, which gets ripped a new one by DDR2-1200 5-5-5, etc.

DDR4 launch spec was DDR4-2133 15-15-15, which gets annihilated by DDR3-2800 12-14-14, DDR3-3000 12-14-14, etc.

You can't compare some of the highest speed of one generation to entry level of the next.


So we'd be on DDR4 forever? DDR5 will not progress until launched as there will be no reason for progression (why better a product that doesn't exist - and how would you test it if there isn't a single platform for validation?).


This is a generation where you will be left behind. People have joked about every generation for the last few years being a rebranded Core i7-6700K with a few extra cores slapped on, but we finally have REAL progression. The performance jump over 11th Gen even with slow DDR5 or DDR4 is something that a year or two ago I would have laughed at as the last time Intel made real strides was almost a quarter of my life ago.


You're quick to call it "fake" but you're running a "fake" 16 core - E cores are a lot less "fake" than HyperThreaded cores. Their performance is incredible by any standard, and they sip power. We'll be seeing performance similar to what we have now at 1/4 the power draw, or much better performance within the same power envelope we currently use.


You don't need a R 10k to R 12k motherboard. Hell, you probably don't NEED the board you have right now.


So get faster RAM - looking at the SKU sheets I have available, several vendors haven't announced anything that slow.

Reading your post for the third time, it kind of reminds me of people who were expecting Ryzen 2000 to be 50% faster at gaming than Intel while costing half as much. People set an impossible bar and were disappointed when the impossible didn't happen - Alder Lake isn't the impossible happening, but based on Intel's track record for the last few years it sure feels like it is.
I’m not saying that Alder Lake is objectively a bad architecture and again, I’m not on any official sales or review threads trying to sabotage people’s livelihoods or anything like that. I’m merely sharing my opinion on why I don’t like what’s in front of us and I feel like that’s totally reasonable. If my comments are compromising future sales, and that’s the reason for this backlash, then please let me know and I’ll certainly cut it out.

If you’re correct then you’re correct with regards to the stuff you seem to have evidence for. The only things I do have to disagree with are the things with the NVMe and the lanes.

A relatively small portion of my friends are students, and fewer of them are students who do PC gaming. Most of my friends who game are actually either done with education, or are competitive gamers.

At this point I only have two friends that own PCIe 4.0 drives. One is a student and another is a Carbie. Pretty much everyone else that I’ve ever interacted with regarding PCs or that I remotely regard as a friend is on 3.0

This extends to every single HOF review score I’ve looked at, at least 75% of SSDs I’ve seen mentioned on signatures; I’m sure you selling a decent amount of 4.0 drives, but I don’t feel like it’s fair to say that they’re widely adopted.

The same can be said with regard to Ampere. I’m sure you’ve sold many units in ZA, but overall sales figures for Ampere and RDNA simply can’t have the two at the highest %. Most gamers are still using older hardware, and previous generation stuff will continue to top charts for a while because most gamers buy budget stuff and used stuff.

PCIe 5.0 follows a similar trend. I’m sure you have clients and friends who do very intricate tasks and require terabytes upon terabytes of storage and are absolutely drowning in terms of how limited PCIe 4.0 is. Although those individuals definitely deserve to be catered to, they’re definitely not in the majority of PC users and I can’t even begin to imagine what running out of PCIe 4.0 lanes would look like. I’ve genuinely only ever come across one individual with multiple PCIe 4.0 devices, and that was @Moemfie_ZA.

The reason we buy higher end motherboard is because if you want to have decent aesthetics, if you want to sync up RGB, if you want to be able to OC without worrying, and if you want a good memory topology, you simply purchase can’t low end. There’s a very valid and reasonable argument for not buying MSI A-Pro boards with 4 layer PCBs and no heat syncs :p. I’m not going to put you on the spot, but Z490 and Z590 weren’t as expensive as Z690, yet they landed here at rather high prices.

All I’m doing is inferring. If the Z690 Aorus Master sells for R8000 I’d be not only surprised, but downright confused. That’s not to mention what the likely figures are for 12th gen CPUs and DDR5. TeamGroup’s 4800 40-40-40 has in fact been on shelves already, and there are multiple vendors with that JEDEC spec up for grabs. Currently Samsung and Adata have some higher end stuff that looks interesting.

The TG was going for $300 I believe, and obviously we’d have to pay VAT and other fees. The Royal Elite 4000C14 should be $280 here, but it sells for R5200. Obviously it might not be totally transferable but I’ve never seen South Africans pay less than what should be expected when fees are added. R7000 to me for a baseline 32GB seems like it would just make historical sense. I can’t see us paying less, what of VAT, distro markups and import fees then?

If you disagree with my price estimations then fair enough. I’m not a site owner and I don’t know any distro’s - I certainly don’t write stuff for NVIDIA 😅. I can’t see my estimations on the pricing being that far off though, if they are, then are you saying this is the first platform that’ll be priced fairly?🤔

Look man, if I’m completely off the mark with regards to everything then fair enough. Perhaps there are millions of PC owners that own 2-3 980 Pro’s, who am I to say otherwise? If we’re getting Z690 boards at 1:1 NA MSRP, we’re paying a solid and fixed $600 for the 12700K, and we’re getting much more competitive DDR5 than the kit I mentioned that had been up for sale, then fair enough. Again, I know absolutely nothing in terms of being an industry insider. It makes discourse with you especially difficult, but I still don’t see why it shouldn’t be allowed to be disappointed.

I really do feel like this sort of apprehension needs to be reserved for individuals who shout silly things on your future sales threads, similarly to how individuals were trying to argue about GPU prices and such. That’s obviously a big no-no and I would never try and interfere with sales threads.

I actually had to back-track after you first replied to me because I thought that I perhaps posted stupid shit on one of Gouhan’s official review threads or a press release. Would’ve deleted them and apologised immediately were that the case. To be completely honest I tried to go back and delete my comments regardless but I see I can’t do that for some reason, as I can imagine there’s probably something distasteful about what I was commenting and I do apologise for that.
 
Last edited:
:)
Dude calm down, it's all good. You've said nothing wrong, what are you apologizing for?
Let's not conflate market/business aspects with the technology.
Much of what you say would not be an issue if Z690 boards suddenly became $120, CPU at $150and DRAM $90. So that speaks to it being a price issue rather than a technology issue. Pricing sucks, for a variety of reasons including the newness of it all.

You can't be disappointed by what hasn't released. The 1st Desktop BIG.little platform is here. For that we needed a new OS as well. We have DDR5 the largest change in DDR since well, the advent of DDR DRAM. Then we have PCIe 5.0
=================================================================================
Snip

Anyway, I think we will all have a few headaches with DDR5 but we will love it more than any other DDR
Just some info around DDR5 which should apply to AMD's Zen4 as well.

* DDR5 modules as we know have an additional intelligent* chip called the PMIC/PMC.
- DDR5 modules get a single 5V line from board, which the PMIC then splits into three lines VDDC, VDDQ and I forgot the other. These three will be what we adjust for our DRAM OC, tuning etc.

* DDR5 has the ability to not only change DRAM frequency on the fly as in the OS, but timings and voltage. The repercussions or opportunities this presents should be obvious. DDR5 can do the usual DRAM training at POST within the OS for the 1st time (Not sure which parts of the training are skipped however)

* X.M.P has been significantly expanded upon to the users benefit. One of the most exciting features is what we can do with our DRAM now in terms of simplifying our tuning, but also just managing our overclocks. It's made much much easier. Can't say much more without making trouble.

* DDR5 official 1.1V doesn't mean we won't see memory go to 1.6V or there abouts. DDR5 frequency will go well beyond the 10Gbps mark and don't be surprised to see DDR5 12000+/14000+ later on with some meaningful timings as well. Remember, absolute latency is both a function of frequency and cas latency.

I think many of us will be blown away by what's on offer, even with DDR4. The platform is better than I could have imagined it to be, but a lot more intimidating for enthusiasts.
 
While I'm not at a level where the tech has any impact on me, my issue isn't absolute price or the absolute performance or even price per performance or even the relative scaling.

Everything you already have is sunk cost so upgrading is going to be significantly more. Socket and CPU upgrades are normal for intel users and even for amd we knew it was coming. Adding new ram in which at this point will provide no upgrade over cheaper better performing options. My understanding of the point affxct is making is at the enthusiast end where the nice mobos are which are going to cost more than current ones won't have ddr4 support, least without inside info to the contrary which just adds useless cost. Same as where the PCIe 5.0 falls for the significantly large majority of people. PCIe 5.0 costs more to implement than 4.0. Like intel has been on 4.0 for how long now?

The issue with the tech isn't the future performance, all that looks like it's going to be significant. It's the problems that are going to crop up on release and the tech as we "the general public" know about the advancement to cost to achieve it doesn't make sense at this point in time.
 
While I'm not at a level where the tech has any impact on me, my issue isn't absolute price or the absolute performance or even price per performance or even the relative scaling.

Everything you already have is sunk cost so upgrading is going to be significantly more. Socket and CPU upgrades are normal for intel users and even for amd we knew it was coming. Adding new ram in which at this point will provide no upgrade over cheaper better performing options. My understanding of the point affxct is making is at the enthusiast end where the nice mobos are which are going to cost more than current ones won't have ddr4 support, least without inside info to the contrary which just adds useless cost. Same as where the PCIe 5.0 falls for the significantly large majority of people. PCIe 5.0 costs more to implement than 4.0. Like intel has been on 4.0 for how long now?

The issue with the tech isn't the future performance, all that looks like it's going to be significant. It's the problems that are going to crop up on release and the tech as we "the general public" know about the advancement to cost to achieve it doesn't make sense at this point in time.
Someone has to be 1st.
Costs, performance and all which you take for granted now wasn't always like that.
I remember the pains of 1st gen Ryzen and Threadripper, we remember the pains of X99 and 1st DDR4 platforms.

There is no time where it's all mastered and ready to go out the box. You wont' get faster DDR5 if there isn't any DDR5 on the market. Same as we couldn't have started with 52X CD-ROM drives, we had to start in the late 80's early 90s with 1X 150Kib/s SCSI drives. They were expensive, slow and prone to all sorts of issues, years later, we had 56X drives,

We couldn't have started at DDR5 12600 out the box. And as @Oj0 said. We have 6600 CL36 now from G.SKILL, which in terms of absolute latency, matches DDR4 3200 CL16.
For us to get to CL16 3200 DDR4, how long did it take? Was more than 6 months, but now even before launch from standard 4800 CL40 spec, we have 6600 CL36.

In their own works... "12th Gen Core is the most significant change to x86, since Core was introduced."

I've my own issues with the platform LAUNCH, not the platform itself, but despite all of that I think it'll be better than imagine it to be, by some margin :)
 
Right, now I have time to reply.

I’m not saying that Alder Lake is objectively a bad architecture and again, I’m not on any official sales or review threads trying to sabotage people’s livelihoods or anything like that. I’m merely sharing my opinion on why I don’t like what’s in front of us and I feel like that’s totally reasonable. If my comments are compromising future sales, and that’s the reason for this backlash, then please let me know and I’ll certainly cut it out.
What don't you like about it? Don't want DDR5? Grab a DDR4 board. Don't like PCIe 5.0? Grab a PCIe 4.0 or older SSD/graphics card.

Objectively it is better in every day, other than rumoured pricing (and we all know how rumours go - heck, other than a lot of what I'm told being under embargo which I respect, a lot of vendors and/or suppliers give out a bit of patently fake information (unique to each person or company) so when it becomes "news" they know exactly where the leak occurred.

If objectively it's better in every way, what is there that is subjectively bad?

If you’re correct then you’re correct with regards to the stuff you seem to have evidence for. The only things I do have to disagree with are the things with the NVMe and the lanes.

A relatively small portion of my friends are students, and fewer of them are students who do PC gaming. Most of my friends who game are actually either done with education, or are competitive gamers.

At this point I only have two friends that own PCIe 4.0 drives. One is a student and another is a Carbie. Pretty much everyone else that I’ve ever interacted with regarding PCs or that I remotely regard as a friend is on 3.0

This extends to every single HOF review score I’ve looked at, at least 75% of SSDs I’ve seen mentioned on signatures; I’m sure you selling a decent amount of 4.0 drives, but I don’t feel like it’s fair to say that they’re widely adopted.

The same can be said with regard to Ampere. I’m sure you’ve sold many units in ZA, but overall sales figures for Ampere and RDNA simply can’t have the two at the highest %. Most gamers are still using older hardware, and previous generation stuff will continue to top charts for a while because most gamers buy budget stuff and used stuff.
You are completely missing the point here. It's not about what people are ALREADY RUNNING, it's about what people are BUYING. The GTX 1060 is still the top card on Steam, it doesn't mean people are rushing out in doves to buy them.

Offices, people trying to get a little more life out of an outdated machine, and those trying to do repairs on the cheap are about the only ones buying Gen3 SSDs these days. GT 1030s are being bought exclusively by those who don't need it to begin with, but don't have integrated graphics.

The "average" gamer these days is buying an RTX 3060 - partially due to the terrible value offered by anything older, and partially due to availability.

The "average" person is buying either an ADATA Swordfish, about the cheapest and slowest NVMe drive on the market, clearly as a first dabble into NVMe or to upgrade/repair on an extremely tight budget, or they're going Gen4. That's not to say nobody is buying any other Gen3 SSD, but they make up a TINY number.

Look through those 75% of sigs running Gen3 drives and ask them when they were purchased - I'll bet that the very vast majority were more than six months ago.

PCIe 5.0 follows a similar trend. I’m sure you have clients and friends who do very intricate tasks and require terabytes upon terabytes of storage and are absolutely drowning in terms of how limited PCIe 4.0 is.
It doesn't take that to hit the limit. Run a graphics card along with ANY other AIC on an Intel platform and you're halving your available lanes per device. Scale it up (and not by much at all) and suddenly you're nowhere near the potential available. PCIe lanes have been a bottleneck for many users for a long time - I dare say since PCI was phased out altogether.

Although those individuals definitely deserve to be catered to, they’re definitely not in the majority of PC users and I can’t even begin to imagine what running out of PCIe 4.0 lanes would look like. I’ve genuinely only ever come across one individual with multiple PCIe 4.0 devices, and that was @Moemfie_ZA.
They don't need to be Gen4. A PCIe 3.0 (or even 2.0) AIC will still halve the lanes available. As NVMe begins replacing SATA (it's already happening, with boards having more and more M.2 slots and less usable SATA ports, with ports being disabled when using an NVMe drive or the board simply offering less SATA ports) and that 7GB/s SSD is suddenly maxing out at the speed of a slightly above entry level NVMe drive.

The reason we buy higher end motherboard is because if you want to have decent aesthetics, if you want to sync up RGB, if you want to be able to OC without worrying, and if you want a good memory topology, you simply purchase can’t low end. There’s a very valid and reasonable argument for not buying MSI A-Pro boards with 4 layer PCBs and no heat syncs :p.
The Z690 boards arguably have even nicer aesthetics, so your subjective points are potentially out the window. The price you're paying to get the absolute most out of RAM exceeds the gains, so value is out the window. Running air/water cooling, you're got nothing to worry about with pretty much any board available. I've sent out a Gigabyte Z390 UD, an extremely entry level board, with the CPU running at 4.9 GHz without issues to date.

I’m not going to put you on the spot, but Z490 and Z590 weren’t as expensive as Z690, yet they landed here at rather high prices.
Neither was released during severe component shortages. Things such as passive component shortages and sound/network controller shortages started since the release of either. I don't think there is a single component that isn't up in price since the release of Z590. CPUs, RAM, storage, graphics cards, PSUs and even cases have all gone up.

All I’m doing is inferring. If the Z690 Aorus Master sells for R8000 I’d be not only surprised, but downright confused. That’s not to mention what the likely figures are for 12th gen CPUs and DDR5. TeamGroup’s 4800 40-40-40 has in fact been on shelves already, and there are multiple vendors with that JEDEC spec up for grabs. Currently Samsung and Adata have some higher end stuff that looks interesting.
4800 is the lowest JEDEC spec for DDR5. The lowest for DDR4 was 1600. That's 3x the speed when comparing entry level to entry level. Of all the vendor SKU sheets I have, nobody, and I mean nobody, doesn't have something faster than 4800. Be it 5200 CL38 or 6600 CL36 (or higher :)), everyone has something faster and/or tighter.

The TG was going for $300 I believe, and obviously we’d have to pay VAT and other fees. The Royal Elite 4000C14 should be $280 here, but it sells for R5200. Obviously it might not be totally transferable but I’ve never seen South Africans pay less than what should be expected when fees are added. R7000 to me for a baseline 32GB seems like it would just make historical sense. I can’t see us paying less, what of VAT, distro markups and import fees then?

If you disagree with my price estimations then fair enough. I’m not a site owner and I don’t know any distro’s - I certainly don’t write stuff for NVIDIA 😅. I can’t see my estimations on the pricing being that far off though, if they are, then are you saying this is the first platform that’ll be priced fairly?🤔

Look man, if I’m completely off the mark with regards to everything then fair enough. Perhaps there are millions of PC owners that own 2-3 980 Pro’s, who am I to say otherwise? If we’re getting Z690 boards at 1:1 NA MSRP, we’re paying a solid and fixed $600 for the 12700K, and we’re getting much more competitive DDR5 than the kit I mentioned that had been up for sale, then fair enough. Again, I know absolutely nothing in terms of being an industry inside. It makes discourse with you especially difficult, but I still don’t see why it shouldn’t be allowed to be disappointed.

I really do feel like this sort of apprehension needs to be reserved for individuals who shout silly things on your future sales threads, similarly to how individuals were trying to argue about GPU prices and such. That’s obviously a big no-no and I would never try and interfere with sales threads.
If you don't like the price of DDR5, go DDR4. If you don't like the price of PCIe Gen5, go Gen4 (or Gen3). You will need a new motherboard, with anticipated prices that definitely higher but not terrible, and a new CPU - but you don't need to go like-for-like (eg you don't need a 12700K to get better performance than you already have). CPU prices are rumoured across the internet to be insanely high, but I can't remember a generation launch that didn't have such rumours, with MSRP being much more reasonable.

I'm TRULY not sure why you're against Alder Lake, when it is proving to be better in every way and with more technological leaps than we've ever seen before - it's a first ever for PCIe Gen5, it's a first ever for DDR5, it's a first ever for BIG.little cores, it's a first for Intel in a long time that we have a meaningful IPC increase. Ryzen 5000 was exciting, but the last time I've been THIS excited about a CPU launch was probably Sandy Bridge a DECADE (yes, slightly more than 10 years) ago.

The expected evolutionary jumps (DDR5, PCIe5) are expected, but BIG.little has the possibility to revolutionise computing as we know it. We can have mega low power draw when idle/browsing the net/typing a document/watching a video, with an absolute powerhouse available within milliseconds when we need it.
 
Right, now I have time to reply.


What don't you like about it? Don't want DDR5? Grab a DDR4 board. Don't like PCIe 5.0? Grab a PCIe 4.0 or older SSD/graphics card.

Objectively it is better in every day, other than rumoured pricing (and we all know how rumours go - heck, other than a lot of what I'm told being under embargo which I respect, a lot of vendors and/or suppliers give out a bit of patently fake information (unique to each person or company) so when it becomes "news" they know exactly where the leak occurred.

If objectively it's better in every way, what is there that is subjectively bad?


You are completely missing the point here. It's not about what people are ALREADY RUNNING, it's about what people are BUYING. The GTX 1060 is still the top card on Steam, it doesn't mean people are rushing out in doves to buy them.

Offices, people trying to get a little more life out of an outdated machine, and those trying to do repairs on the cheap are about the only ones buying Gen3 SSDs these days. GT 1030s are being bought exclusively by those who don't need it to begin with, but don't have integrated graphics.

The "average" gamer these days is buying an RTX 3060 - partially due to the terrible value offered by anything older, and partially due to availability.

The "average" person is buying either an ADATA Swordfish, about the cheapest and slowest NVMe drive on the market, clearly as a first dabble into NVMe or to upgrade/repair on an extremely tight budget, or they're going Gen4. That's not to say nobody is buying any other Gen3 SSD, but they make up a TINY number.

Look through those 75% of sigs running Gen3 drives and ask them when they were purchased - I'll bet that the very vast majority were more than six months ago.


It doesn't take that to hit the limit. Run a graphics card along with ANY other AIC on an Intel platform and you're halving your available lanes per device. Scale it up (and not by much at all) and suddenly you're nowhere near the potential available. PCIe lanes have been a bottleneck for many users for a long time - I dare say since PCI was phased out altogether.


They don't need to be Gen4. A PCIe 3.0 (or even 2.0) AIC will still halve the lanes available. As NVMe begins replacing SATA (it's already happening, with boards having more and more M.2 slots and less usable SATA ports, with ports being disabled when using an NVMe drive or the board simply offering less SATA ports) and that 7GB/s SSD is suddenly maxing out at the speed of a slightly above entry level NVMe drive.


The Z690 boards arguably have even nicer aesthetics, so your subjective points are potentially out the window. The price you're paying to get the absolute most out of RAM exceeds the gains, so value is out the window. Running air/water cooling, you're got nothing to worry about with pretty much any board available. I've sent out a Gigabyte Z390 UD, an extremely entry level board, with the CPU running at 4.9 GHz without issues to date.


Neither was released during severe component shortages. Things such as passive component shortages and sound/network controller shortages started since the release of either. I don't think there is a single component that isn't up in price since the release of Z590. CPUs, RAM, storage, graphics cards, PSUs and even cases have all gone up.


4800 is the lowest JEDEC spec for DDR5. The lowest for DDR4 was 1600. That's 3x the speed when comparing entry level to entry level. Of all the vendor SKU sheets I have, nobody, and I mean nobody, doesn't have something faster than 4800. Be it 5200 CL38 or 6600 CL36 (or higher :)), everyone has something faster and/or tighter.


If you don't like the price of DDR5, go DDR4. If you don't like the price of PCIe Gen5, go Gen4 (or Gen3). You will need a new motherboard, with anticipated prices that definitely higher but not terrible, and a new CPU - but you don't need to go like-for-like (eg you don't need a 12700K to get better performance than you already have). CPU prices are rumoured across the internet to be insanely high, but I can't remember a generation launch that didn't have such rumours, with MSRP being much more reasonable.

I'm TRULY not sure why you're against Alder Lake, when it is proving to be better in every way and with more technological leaps than we've ever seen before - it's a first ever for PCIe Gen5, it's a first ever for DDR5, it's a first ever for BIG.little cores, it's a first for Intel in a long time that we have a meaningful IPC increase. Ryzen 5000 was exciting, but the last time I've been THIS excited about a CPU launch was probably Sandy Bridge a DECADE (yes, slightly more than 10 years) ago.

The expected evolutionary jumps (DDR5, PCIe5) are expected, but BIG.little has the possibility to revolutionise computing as we know it. We can have mega low power draw when idle/browsing the net/typing a document/watching a video, with an absolute powerhouse available within milliseconds when we need it.
Okay look fair enough, I get that you’re excited and I have no right to try and put down genuine innovation. I’ll say one thing though, and that is that while doing tasks on MS Word and browsing on MS Edge, my CPU doesn’t draw more than 25W and it’s locked at 1.25V with all power saving features disabled; not trying to facetious or anything like that. Generally the P=V.I rule saves overclockers in that regard because unless the load is heavy, power and heat isn’t that much if an issue.
 
Last edited:
What’s the latency on that? Can we calculate the response time based on the technical specs? I thought

I thought saw some calculations around it recently.
11ns but the sticks access data differently so I guess we can't really draw comparisons.
 
11ns but the sticks access data differently so I guess we can't really draw comparisons.
You're right it is 11ns. But because of the changes to caches etc, the higher than normal absolute latency doesn't hurt as much as one would thik.

That said, for obvious NDA reasons, can't say much but I'm still in shock as to what I'm witnessing. Can't
wait to share all the stuff with you all. Let's see what we can get away with showing on the 27th, where we can do unboxing's etc.
 

Preview is out. Not much longer to go now just next thursday.
There's a lot to this architecture and CPUs, therefore I'll not do CPU review until end of Nov, next week will be motherboards, DDR5 etc.

This is at present my favorite Z690 board for obvious reasons. Overclocking on this board and the HERO for that matter is rather interesting. SImplified from Z590/490 and that's also due to necessity. There are significantly more tuning options available which is rather intimidating TBH.

That said, lots of neat features on these boards including the mechanical button which you press to pop out your GPU instead of using a sharp/dangerous object.

Power delivery is again the most advanced we've seen on desktop, this generation. The Apex is no different and the specs speak for themselves. (105A teamed Power Stages, 24 Phases etc. ) Presence of Gracemont cores on the CPUs I can imagine provide further power complications. As such it makes sense why power delivery is so much more robust this generation (At least on paper, we'll wait for Buildzoid et al. for the true test)

Thus far pretty good specifications for the mid to high end boards. Budget ones may cut too many corners sometimes. The shift to X8 DMI link and a PCEe 4.0 capable PCH has shown some interesting configurations, like boards with x5 M.2 sockets with four at PCIe 4.0


Apex_690.jpg
 
Sigh....

I should have known this would happen. 12th Gen is rather detailed when it comes to overclocking and you can easily screw it up, when you're unsure of what you're doing (board descriptions of features are partly to blame).

Look at this.
From Guru3d - Core i9 12900K processor review

For this score, this OC had a monumental power draw of 469W ........ Why? Please compare with the score I got the power draw that matches it below

index.php



Higher Score - 11078 - Lower Power 236W - Lower All core Clock 5.1GHz

How.jpg


I knew this would happen. It's easy far too easy to n00b Z690. It was very easy with Rocket Lake (I did on numerous occasions) as is and Z690 makes it worse. It's clear that reviews right now are a bit all over the place.

SMH

469W vs 240W with a higher score
 
Whats up with the cas latency on ddr5 being so high compared to ddr4, will that not affect response time and hurt performance??
 
Whats up with the cas latency on ddr5 being so high compared to ddr4, will that not affect response time and hurt performance??
It’s not just CAS. tRCD, tRP, tRAS, tRC, tRTP, tRFC, tCWL, and tWR will all be sky high. Not terribly sure about some of the others. The communication between the DDR5 IMC and DDR5 ICs is more efficient and thus the timings are less of a factor. They definitely do make a significant difference to latency benchmarks and many games, especially stuff like Tomb Raider, WZ and StarCraft according to one review I watched.
 
Whats up with the cas latency on ddr5 being so high compared to ddr4, will that not affect response time and hurt performance??
What you lose out on in cas latency it makes up for in other areas. It'll eventually come down either through increased frequency (absolute latency is determined not only by CAS ticks, but by frequency in combination with the number of CAS ticks) or the timings themselves improving due to better ICs.

We've seen CL30 DDR5 6000 already ;)
 

Users who are viewing this thread

Back
Top Bottom