Intel Plans To Battle AMD Ryzen 4000 In 2020 With Mass of Hyper-Threaded Processors Including 5.3GHz

imsirovic5

[H]Lite
Joined
Jun 21, 2011
Messages
75
It's actually very logical.

PCIe 3.0 is barely used over PCIe 2.0, thus, PCIe 4.0 will be even less used over PCIe 3.0.

From Anandtech:

Samsung is now talking about three major improvements they bring over earlier SSDs in addition to the raw performance increases enabled by PCIe 4.0.

Link: https://www.anandtech.com/show/14884/samsungs-pcie-gen-4-enterprise-ssds-get-reliability-performance-boost

Given this, how will SSD market look in few years? Sorry I am not smart enough to follow your logic.....

Also:

https://www.tomshardware.com/news/intel-has-pcie-40-optane-ssds-ready-but-nothing-to-plug-them-in-to

Now you see why I would like PCIe 4 to plug these into?
 
Last edited:

Mega6

2[H]4U
Joined
Aug 13, 2017
Messages
2,745
Intel appears to be enabling Hyperthreading on lower end CPU's like the Core i3 and Core i5. CPU's which have traditionally had HT disabled.
Enabling HT on low end cpus might be a "new marketing feature" for Intel, but certainly not innovative as their competitor has been doing it for years. Innovative - as in new technology, which HT is of course, not.
 
Last edited:

IdiotInCharge

[H]ardForum Junkie
Joined
Jun 13, 2003
Messages
13,265
Now you see why I would like PCIe 4 to plug these into?
Nope; we've had "PCIe 4.0" drives for a while, and Optane too.

Increasing the bus speed won't fix the storage performance limitations we're currently seeing, which are almost exclusively software performance bound.

And those SSDs will work great in PCIe 3.0 slots. You have to find a very specific workload to show an improvement with PCIe 4.0 over PCIe 3.0, and you're not likely to be doing that on a consumer platform.
 

imsirovic5

[H]Lite
Joined
Jun 21, 2011
Messages
75
Nope; we've had "PCIe 4.0" drives for a while, and Optane too.

Increasing the bus speed won't fix the storage performance limitations we're currently seeing, which are almost exclusively software performance bound.

And those SSDs will work great in PCIe 3.0 slots. You have to find a very specific workload to show an improvement with PCIe 4.0 over PCIe 3.0, and you're not likely to be doing that on a consumer platform.

Again, I was not referring to current PCIe 4.0 SSD advantages, but rather the developments over next few years. The links I shared earlier suggest industry leaders such as Intel and Samsung see benefit of PCIe 4.0 when it comes to SSD speed and performance. In fact, per link above, Intel is specifically targeting PCIe 4.0 for their upcoming Optane SSDs (not the existing optane tech but ones with second-gen 3D XPoint) for the consumer market. I imagine Intel is doing this to unlock the additional potential of their tech in the future.

Further, we also have a healthy history of continuous increase of system bus throughput. I am sure back in 2003 when PCIe 1.0 was introduced with maximum transfer rate of 2.5 GT/s, many thought that this would be more than enough and that we would never have to move away from this standard, the same you believe that current 8.0 GT/s is more than we will ever need. In fact, I am sure back in ISA-8 and ISA-16 days many thought we would not need more. Or once PCI 32/33 was introduced in 1993 I am sure there were individuals who thought this was all we need.

And you may be right, both the Samsung and Intel and other leaders in SSD space may be wrong for targeting perceived benefits of PCIe 4.0 for the next gen SSD. But as a computer enthusiast, I am excited about potential of upcoming technologies and I refuse to think that the progress will stop. And this is why I would like to have PCIe 4.0 on my next system – just in case Samsung and Intel succeed in their current efforts to leverage PCIe 4.0 for faster SSD drives.

With this, I will refrain from further posts on this topic, as I do not want to hijack this thread that should be about Hyper-Threaded Processors and not about perceived advantages of PCIe 4.0 when it comes to future generations of SSD drives. So my apologies to those in this thread for temporarily hijacking the subject.
 
Last edited:

OrangeKhrush

[H]ard|Gawd
Joined
Dec 15, 2016
Messages
1,614
These new I9's will need more robust cooling, now is a good time to build a swimming pool
 

IdiotInCharge

[H]ardForum Junkie
Joined
Jun 13, 2003
Messages
13,265
I imagine Intel is doing this to unlock the additional potential of their tech in the future.
Intel upgrades controllers because that's what they do. Even Optane cannot make good use of PCIe 3.0 today, and yet, it's the most performant non-volatile storage solution outside of pure sequential reads and writes. And that increase in purely sequential performance is all PCIe 4.0 provides over PCIe 3.0, aside from an increase in cost.

Further, we also have a healthy history of continuous increase of system bus throughput. I am sure back in 2003 when PCIe 1.0 was introduced with maximum transfer rate of 2.5 GT/s, many thought that this would be more than enough and that we would never have to move away from this standard, the same you believe that current 8.0 GT/s is more than we will ever need. In fact, I am sure back in ISA-8 and ISA-16 days many thought we would not need more. Or once PCI 32/33 was introduced in 1993 I am sure there were individuals who thought this was all we need.
I'm sure 'many' believed many different things, and that's not what I'm talking about here.

Our problem is not that devices and buses are getting faster -- it's that the software that directs the work to be done isn't making use of them. Only in extremely niche enterprise environments does NVMe show a real advantage over SAS, for example, and on the desktop, NVMe and SATA are almost indistinguishable to the end user. One has to run benchmarks to make sure the performance is there!

And you may be right, both the Samsung and Intel and other leaders in SSD space may be wrong for targeting perceived benefits of PCIe 4.0 for the next gen SSD.
It's not about 'right' or 'wrong'. I'm not going to complain about more speed; that's not the point. The point is that, as it stands, there is very little utility for PCIe 4.0 over PCIe 3.0, and further, due to the limited utility, no sound reason to prefer one over the other. It's like buying tires rated for 160MPH over tires rated for 120MPH for a car that will only ever be going close to 60MPH, unless the 'better' tires come with no drawbacks and are the same price, as the extra speed rating simply won't be of any use.

. But as a computer enthusiast, I am excited about potential of upcoming technologies and I refuse to think that the progress will stop.
So am I!

...for technologies that will actually improve end-user experience. Right now, that would be cheaper Optane drives regardless of the speed of the NVMe slot they plug in to, if we're talking about storage.

And this is why I would like to have PCIe 4.0 on my next system – just in case Samsung and Intel succeed in their current efforts to leverage PCIe 4.0 for faster SSD drives.
They'll make faster SSD drives, no doubt, but outside of drive-specific benchmarks, you won't be able to show a difference in application performance.

With this, I will refrain from further posts on this topic, as I do not want to hijack this thread that should be about Hyper-Threaded Processors and not about perceived advantages of PCIe 4.0 when it comes to future generations of SSD drives. So my apologies to those in this thread for temporarily hijacking the subject.
I've continued because there is a further point: while AMD did tout PCIe 4.0 in their marketing, the real advance with X570 was the increase in the number of PCIe lanes. The extra bandwidth is near useless, and in fact, increases cost, but the increase in the number of lanes makes a difference in terms of motherboard layout flexibility that benefits end-users.

By the time PCIe 3.0 is limiting to the point that PCIe 4.0 would actually be useful, we'll be looking at replacing PCIe 5.0 with PCIe 6.0. But, for those that want or need to plug in more peripherals to make better use of the increasing core counts available in newer consumer CPUs, more PCIe lanes on consumer platforms is really a huge win!
 

NightReaver

[H]Lite
Joined
Apr 20, 2017
Messages
103
Intel upgrades controllers because that's what they do. Even Optane cannot make good use of PCIe 3.0 today, and yet, it's the most performant non-volatile storage solution outside of pure sequential reads and writes. And that increase in purely sequential performance is all PCIe 4.0 provides over PCIe 3.0, aside from an increase in cost.



I'm sure 'many' believed many different things, and that's not what I'm talking about here.

Our problem is not that devices and buses are getting faster -- it's that the software that directs the work to be done isn't making use of them. Only in extremely niche enterprise environments does NVMe show a real advantage over SAS, for example, and on the desktop, NVMe and SATA are almost indistinguishable to the end user. One has to run benchmarks to make sure the performance is there!



It's not about 'right' or 'wrong'. I'm not going to complain about more speed; that's not the point. The point is that, as it stands, there is very little utility for PCIe 4.0 over PCIe 3.0, and further, due to the limited utility, no sound reason to prefer one over the other. It's like buying tires rated for 160MPH over tires rated for 120MPH for a car that will only ever be going close to 60MPH, unless the 'better' tires come with no drawbacks and are the same price, as the extra speed rating simply won't be of any use.



So am I!

...for technologies that will actually improve end-user experience. Right now, that would be cheaper Optane drives regardless of the speed of the NVMe slot they plug in to, if we're talking about storage.



They'll make faster SSD drives, no doubt, but outside of drive-specific benchmarks, you won't be able to show a difference in application performance.



I've continued because there is a further point: while AMD did tout PCIe 4.0 in their marketing, the real advance with X570 was the increase in the number of PCIe lanes. The extra bandwidth is near useless, and in fact, increases cost, but the increase in the number of lanes makes a difference in terms of motherboard layout flexibility that benefits end-users.

By the time PCIe 3.0 is limiting to the point that PCIe 4.0 would actually be useful, we'll be looking at replacing PCIe 5.0 with PCIe 6.0. But, for those that want or need to plug in more peripherals to make better use of the increasing core counts available in newer consumer CPUs, more PCIe lanes on consumer platforms is really a huge win!
That's why I went x570. Forget pcie4, I'm way more happy that I get triple m.2 slots that are full bandwidth.
 

vick1000

[H]ard|Gawd
Joined
Sep 15, 2007
Messages
1,974
PCIe 4.0 is legit, it provides more overall lanes and bandwidth for more devices. 10Gb LAN, multi-NVMe, USB 3.1+, all simultaniously, as long as the CPU can handle all of them.
 
  • Like
Reactions: noko
like this

Jandor

Gawd
Joined
Dec 30, 2018
Messages
527
I hope they put an iGPU on their 10 core since this will be the only market where AMD won't have a better solution and with more cores even after July.
 

N4CR

[H]ardness Supreme
Joined
Oct 17, 2011
Messages
4,449
LOL! 'Future proofing is bad, guys' (when it's AMD)
Intel releases PCie 5 in 2030: 'BEST THING EVA JUST BUY IT'

I can smell the shareholder desperation from here.
 

Lepardi

Limp Gawd
Joined
Nov 8, 2017
Messages
234
By the time PCIe 3.0 is limiting to the point that PCIe 4.0 would actually be useful, we'll be looking at replacing PCIe 5.0 with PCIe 6.0.
We have already exceeded PCI-E 3.0 x8. Getting PCI-E 4.0 is legit if you build your system with say a 5 year life span, in a few years you can upgrade your GPU and not be limited.

That is unless you buy a Gigabyte board, and end up with non functional PCI-E 3.0 setting after a BIOS update like I did lol.
 

drescherjm

[H]ardForum Junkie
Joined
Nov 19, 2008
Messages
14,644
Intel releases PCie 5 in 2030: 'BEST THING EVA JUST BUY IT'
I thought PCIe 5.0 and DDR5 were in the next 2 years. Although I would expect that Intel will do this once they have 7nm EUV silicon.
 

IdiotInCharge

[H]ardForum Junkie
Joined
Jun 13, 2003
Messages
13,265
We have already exceeded PCI-E 3.0 x8. Getting PCI-E 4.0 is legit if you build your system with say a 5 year life span, in a few years you can upgrade your GPU and not be limited.
Barely.

But the upgrade argument doesn't really hold water, as you'd need to upgrade the CPU significantly as well, and that would require a platform change regardless. PCIe has become the example of a bus with the cart before the horse; that's nothing to complain about, but it means that it's also nothing to worry about. We have plenty of other bottlenecks to worry about first before PCIe lane bandwidth becomes an issue, in particular, just having enough PCIe lanes in the first place.
 

Lepardi

Limp Gawd
Joined
Nov 8, 2017
Messages
234
,as you'd need to upgrade the CPU significantly as well, and that would require a platform change regardless.
Not really, my 6-year old 4790K is still enough to drive a 1080 Ti for example. Though the security patches did hit me quite badly.

But considering AMD is now breaking the Intel monopoly and trying to raise the annual improvement up to 10%, you might be right.
 
  • Like
Reactions: N4CR
like this

IdiotInCharge

[H]ardForum Junkie
Joined
Jun 13, 2003
Messages
13,265
Not really, my 6-year old 4790K is still enough to drive a 1080 Ti for example.
In your opinion ;)

But considering AMD is now breaking the Intel monopoly and trying to raise the annual improvement up to 10%, you might be right.
Five years is a lot of time. AMD will likely find themselves facing down much improved Intel products in five years, as well as Intel with their fabs in full swing, unlike today.

Also, AMD exists because Intel allows them -- AMD has no chance at breaking your so-called 'Intel Monopoly', they simply don't have what it takes to get on top before Intel recovers from their own mistakes. At best, AMD can hope to still be approaching parity in the CPU arena, on occasion, like they do in the GPU arena.

Still, that does mean that in five years we should see enough advancement in CPU performance and GPU performance that an upgrade to either would necessitate an upgrade to both. And since a CPU upgrade will come with a new socket and new memory, well, that's a platform upgrade too.
 

Jandor

Gawd
Joined
Dec 30, 2018
Messages
527
In your opinion ;)



Five years is a lot of time. AMD will likely find themselves facing down much improved Intel products in five years, as well as Intel with their fabs in full swing, unlike today.

Also, AMD exists because Intel allows them -- AMD has no chance at breaking your so-called 'Intel Monopoly', they simply don't have what it takes to get on top before Intel recovers from their own mistakes. At best, AMD can hope to still be approaching parity in the CPU arena, on occasion, like they do in the GPU arena.

Still, that does mean that in five years we should see enough advancement in CPU performance and GPU performance that an upgrade to either would necessitate an upgrade to both. And since a CPU upgrade will come with a new socket and new memory, well, that's a platform upgrade too.
This is exactly what anyone would have said about IBM in the 80s.
 

N4CR

[H]ardness Supreme
Joined
Oct 17, 2011
Messages
4,449
I thought PCIe 5.0 and DDR5 were in the next 2 years. Although I would expect that Intel will do this once they have 7nm EUV silicon.
It was a bit of sarcasm but yeah 2-3 years would be my expectation as well.
 

cyclone3d

[H]ardForum Junkie
Joined
Aug 16, 2004
Messages
13,274
Not really, my 6-year old 4790K is still enough to drive a 1080 Ti for example. Though the security patches did hit me quite badly.
Actually... you would be surprised how much a faster CPU could help in your case.

I am currently running a XEON E5-1680 V2 at 4.6Ghz (8-core 16-thread with 25MB L3 cache) - 4th gen just like yours.

With an RTX 2080, turning v-sync off in The Division 2 makes my CPU run at 85-100% load at all times.

At 75fps steady my CPU runs at 45% and higher CPU load.

So yeah, a 1080Ti is being limited by your current CPU unless you are limiting your fps to somewhere in between 60-75fps... and even that may be pushing it.
 

UnknownSouljer

[H]ardness Supreme
Joined
Sep 24, 2001
Messages
6,172
And WHERE IS AMD's 5GHZ PART???? Yeah that is what I thought!
Can't tell if this is /s.
I don't care what the clockspeed is. I care what is faster. That's like caring about a 1000HP car that is only capable of going 115mph. Do I care that the alternative is "only" 800HP but goes to 140?
Or to take the analogy another way you could come up with the example of a 1000HP car having a 0-60 of 4.5 secs vs an "only" 800HP car doing it in 3.9.
Or said another way, there are more variables than just clock speed. Just like there are more variables than horsepower. If clock speed makes you feel better than I guess buy Intel? For everyone else that can think past one layer of abstraction buy AMD?
 
  • Like
Reactions: N4CR
like this

kac77

2[H]4U
Joined
Dec 13, 2008
Messages
2,335
I didn't misread -- PCIe 2.0 speeds for SSDs still exceed what is actually useful on the desktop, and PCIe 3.0 still isn't fully utilized by GPUs.

PCIe 4.0 is fairly useless, outside of being more expensive and a nice bullet point.
Huh. It is very very easy to saturate a pcie 3.0 8x slot. One HBA can do that.


They only come in 8x form factors so if you want more bandwidth the only way to do it is to move to pcie4 or add in another card which isn't always an option.

I get it you like Intel but saying stuff like pcie 3.0 isn't saturated when it most definitely is just gives people the wrong information. It's not accurate at all.
 

Lastan010

Limp Gawd
Joined
Mar 2, 2017
Messages
154
Intel is doing their best to put a happy face on while staring at process trouble. Is what it is. However, Intel isn't out and won't be for a long time. Enjoy your AMD processors while they are ahead. It's what I did during the Athlon years and I'm doing now with Ryzen. I'll jump back to Intel when they take the lead again. It's how every enthusiast should be looking at things.
Athlon clearly destroyed Intel cpu's in gaming back then, that is not the case today with i9 and amd's latest cpu, so idk why anyone would switch to AMD if they wanted the best gaming cpu.
 

Lastan010

Limp Gawd
Joined
Mar 2, 2017
Messages
154
Not really, my 6-year old 4790K is still enough to drive a 1080 Ti for example.
its hard to read that magnitude of pure ignorance.

I wont even waste time, been there and I know the outcome from people who claim stuff like that.
 

Shoganai

Gawd
Joined
Dec 5, 2018
Messages
525
Athlon clearly destroyed Intel cpu's in gaming back then, that is not the case today with i9 and amd's latest cpu, so idk why anyone would switch to AMD if they wanted the best gaming cpu.
Because the numbers are negligible above 1080p. So for gaming, AMD is roughly on par with Intel for above-1080p gaming, and completely decimates Intel with multi-core work. It’s not confusing or complicated.
 

kac77

2[H]4U
Joined
Dec 13, 2008
Messages
2,335
I'm guessing that its a lot harder to saturate in the home and gaming space. Enterprise level data transfers is a totally different game.
You forget the workstation market especially for content creators. With such high core counts the desktop platform can be easily used for that duty as well.

I can see that happening quite easily just to save money.
 

IdiotInCharge

[H]ardForum Junkie
Joined
Jun 13, 2003
Messages
13,265
What is the ignorance? 4790K is basically a 7700K.
:ROFLMAO:

You forget the workstation market especially for content creators. With such high core counts the desktop platform can be easily used for that duty as well.
It can be, but most would prefer a NAS to aggregate spinning storage and keep flash storage for local work.

I can see that happening quite easily just to save money.
This is the kind of 'saving' that bites one in the ass. It's not recommended to shove a bunch of functions into one system for a reason.
 

kac77

2[H]4U
Joined
Dec 13, 2008
Messages
2,335
:ROFLMAO:


It can be, but most would prefer a NAS to aggregate spinning storage and keep flash storage for local work.
Think low cost content creation machine. Something like Photoshop. You'll want a scratch disk. Most people will stick two ssds together in raid 1 for that alone. You've got your OS drive which could be a ssd as well.

Then your local storage for assets that aren't complete. You're not going to want to send that over the network constantly. It's too much data. So local raid 6/5 for that and just like that the slot is saturated. It's really not hard to do. Workstation cases have all of those drive bays for exactly that type of need.
 

IdiotInCharge

[H]ardForum Junkie
Joined
Jun 13, 2003
Messages
13,265
Then your local storage for assets that aren't complete. You're not going to want to send that over the network constantly. It's too much data. So local raid 6/5 for that and just like that the slot is saturated. It's really not hard to do. Workstation cases have all of those drive bays for exactly that type of need.
You have built-in ports, and you'll want to use Storage Spaces or ZFS depending on host OS for those for portability and recovery purposes -- and if you need more than six spinners, yeah, you'll want to use a NAS, and with 10Gbit being affordable, that's not really an issue.
 

kac77

2[H]4U
Joined
Dec 13, 2008
Messages
2,335
You have built-in ports, and you'll want to use Storage Spaces or ZFS depending on host OS for those for portability and recovery purposes -- and if you need more than six spinners, yeah, you'll want to use a NAS, and with 10Gbit being affordable, that's not really an issue.
Yeah if you have only 1 creator. That's really rare. You're going to have more than 1. You also have QoS you have to maintain.

Your typical high resolution image is many gigs. Typically creators sync back to the network at the end of the day. But until that time comes that data is local. You need to protect it. So those machines will have raid. There's no getting around that. That's why raid is quite common for those types of motherboards anyway.
 

noko

[H]ardness Supreme
Joined
Apr 14, 2010
Messages
4,663
Pcie 3 is holding up SSD performance, PCIe 4 use case is video editing high resolution video in real time. If one is going to use multiple GPU's for like rendering (many Cuda driven programs like VRay where folks use 2x 2080Ti's), gaming etc. on a consumer level part, being stuck with pcie 3 8x on one card is showing limitations now. Having pcie 4 where the cards can run at PCIe 4 8x would alleviate that except Nvidia cards are still pcie 3 at this time. Then the other is chipset bandwidth where AMD is 4x pcie 4 where Intel is 4x pcie 3 -> AMD has twice the bandwidth for all the stuff placed on the chipset which can include SSDs and other type cards, Intel to me is becoming very restrictive for future growth on their consumer level mainboards chipsets and CPUs.

I for one if upgrading would want the fastest possible pcie lanes for future potential not being locked out. Even now pcie 3 is limiting in different types of scenarios.
 

crazycuz20

Limp Gawd
Joined
Oct 3, 2017
Messages
139
I'd agree. if you're an @home content creator, then OS drive (SSD), scratch drive (nvme SSD), plus local backup (spinners), then external back up (either a NAS/external HD).

- my freenas box runs a TR4 1950X, 2 onboard NVME 500 gb in Raid, 4 NVME 1TB in Raid on a PCI-E card, and 2 SSD's as freenas cache drives on top of 5 spinners with 10GBe link aggregation
- please don't tell me there's not a reason or need for more PCI-E lanes or more PCI-E bandwith
- Yes, it may not be applicable for many users but there are power users and this is [H]ard|Forum

Sooner 40/100GBe fiber will become more affordable as it trickles down from enterprise hardware, where hopefully local storage isn't required.
- Though my dream is what linus did:
 

cyclone3d

[H]ardForum Junkie
Joined
Aug 16, 2004
Messages
13,274
I'd agree. if you're an @home content creator, then OS drive (SSD), scratch drive (nvme SSD), plus local backup (spinners), then external back up (either a NAS/external HD).

- my freenas box runs a TR4 1950X, 2 onboard NVME 500 gb in Raid, 4 NVME 1TB in Raid on a PCI-E card, and 2 SSD's as freenas cache drives on top of 5 spinners with 10GBe link aggregation
- please don't tell me there's not a reason or need for more PCI-E lanes or more PCI-E bandwith
- Yes, it may not be applicable for many users but there are power users and this is [H]ard|Forum

Sooner 40/100GBe fiber will become more affordable as it trickles down from enterprise hardware, where hopefully local storage isn't required.
- Though my dream is what linus did:
Local storage is always going to be king as far as throughput and latency goes. No way around that unless you have a server / servers with dedicated connections to each client and dedicated storage devices on said servers for each client. And then what is the point?
 

mikeo

Gawd
Joined
May 17, 2006
Messages
624
40g IB at home is the new 10g, the mellanox 3 40g IB cards can be had for 50 bucks each off ebay.

6-8 years from now when the market is flooded with used 100g pcie4 intel cards and my current threadripper 3970x gets turned into the new ZFS server I'll be glad it has PCIE4 to make use of 100g over one of these:

https://cdrdv2.intel.com/v1/dl/getContent/615503

Also of note, intel likes to hide that this 100g pcie4 card uses pcie4 since their platforms don't support it yet.
 
Top