-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
So when is the new Ice Lake and Tiger Lake Desktop expected release time frame?
- Thread starter c627627
- Start date
- Joined
- Mar 7, 2008
What did Intel say when/where? In case I missed something. I only recall them saying that 10nm was not cancelled on desktop contrary to rumours going around at the time, but no more detail beyond that. What we kinda know is mobile is obviously already shipping, server is supposed to be 1H 2020 but no more said about consumer desktop. I can only speculate they might offer a cut down version of 10nm server later in the year and offer that has a desktop solution, but we might have a game of consumer vs HEDT there. Outside that, isn't the rumour of the day that Intel have a 10 core 14nm CPU?
OP
- Joined
- Feb 18, 2002
- Thread Starter
- #4
Spring for Ice Lake. Do you expect PCIe Gen 4 support, Woomack?
I was surprised to learn just how much more attractive Tiger Lake will be over Ice Lake, do you see it being that much better than Ice Lake and will it be one year's wait for Tiger Lake after Ice Lake's release? Educated guess.
I only had info about motherboard availability but I'm not sure if there will be PCIE 4 or anything else. Honestly, I wouldn't count on really significant changes. Maybe a bit faster and maybe with refreshed controllers. For example, Intel released AX WiFi at the beginning of this year and all X570 motherboards have it (as long as support WiFi) but Intel motherboards not. The first will be refreshed X299 which is entering the market right now.
I was expecting to see DDR5 next year but clearly it will be delayed as next 2 generations still have DDR4 in plans.
Btw. 10nm was on laptops last year but it was only in low power i3. The only problem was that performance wasn't really better than the 14nm version and these chips were heating up a lot causing problems in small laptops. Clearly some design flaws. Back then 10nm was expected to hit the desktop market but Intel clearly had problems with 10nm. Also about then they had problems with production in general.
I don't think that the lower process is helping much. We already have problems with the cooling of higher chips that have small die.
I was expecting to see DDR5 next year but clearly it will be delayed as next 2 generations still have DDR4 in plans.
Btw. 10nm was on laptops last year but it was only in low power i3. The only problem was that performance wasn't really better than the 14nm version and these chips were heating up a lot causing problems in small laptops. Clearly some design flaws. Back then 10nm was expected to hit the desktop market but Intel clearly had problems with 10nm. Also about then they had problems with production in general.
I don't think that the lower process is helping much. We already have problems with the cooling of higher chips that have small die.
OP
- Joined
- Feb 18, 2002
- Thread Starter
- #6
Tiger Lake mobile will have DDR5 it sounds like, when you say next two generations, do you mean Ice Lake and Tiger Lake, in other words, do you see DDR5 on Desktops only after Tiger Lake?
EDIT: Yea, Sapphire Rapids, which comes after Tiger Lake will have DDR 5, and since Ice Lake is 2020, it'll be two or three years, so no DDR5 in the nearest future.
The question then goes back to what Tiger Lake would bring over Ice Lake. PCIe 4 and significant graphics improvements.
Question: Will we be able to buy an Ice Lake motherboard and just swap in a Tiger Lake CPU later?
EDIT: Yea, Sapphire Rapids, which comes after Tiger Lake will have DDR 5, and since Ice Lake is 2020, it'll be two or three years, so no DDR5 in the nearest future.
The question then goes back to what Tiger Lake would bring over Ice Lake. PCIe 4 and significant graphics improvements.
Question: Will we be able to buy an Ice Lake motherboard and just swap in a Tiger Lake CPU later?
Last edited:
OP
- Joined
- Feb 18, 2002
- Thread Starter
- #7
I guess I was blissfully unaware of AMD's advances. It sounds like that in 2020, AMD Zen (3) Ryzen 4000 Series is probably the best way to go, if you can't wait for 2021 Zen (4) Ryzen 5000 with DDR5. Intel doesn't seem to be able to compete with them now, is that correct?
They compete with them... on clockspeeds and overclock capabilities. IPC is about the same (with perhaps AMD slightly in the lead), but the AMD CPUs really can't overclock much at all. To the point where unless you can utilize (not use) all the cores and threads, it is best to use their "PBO" instead.
It just depends on your needs, but in general, the AMD CPU is the better buy these days. Feel free to look at reviews covering this on the front page.
It just depends on your needs, but in general, the AMD CPU is the better buy these days. Feel free to look at reviews covering this on the front page.
- Joined
- Mar 7, 2008
If I had nothing and buying a new general purpose home system, with generous but not unlimited budget, I'd probably pick a Zen 2 CPU as 1st choice, then a current Intel, with older Ryzen a distant third. There's two parts to this, absolute performance, and performance per $. Well, there's more, you could argue about power efficiency and platform considerations, but let's keep it simple. If we take Skylake architecture as a reference point, which hasn't significantly changed through to Coffee Lake, it's still good but feeling a little dated. With Zen 2 on the scene, it matches or beats Skylake in most areas, and generally speaking, is cheaper like for like. Win-win. Older Ryzen, before Zen 2, was still good, but there were a lot more "but..." areas.
In specific areas, one or other CPU may be better, but right now Zen 2 offers decent overall performance, at a decent price, with some degree of platform longevity that isn't known on Intel side (the next gen desktop is rumoured to have yet another socket change).
Intel CPUs are still as good as they were, just that they have lost the performance crown and I think that perception will hurt them more than any technical reason. There is no indication of a radical change on desktop with next generation, so any hopes will be on the one after that. We know they have a better architecture, known as Sunny Cove and found in Ice Lake mobile CPUs, but desktop will have a wait. They had said 10nm on desktop was not cancelled, but the feeling is they will do some kind of limited release to say they done it, than any mass market range.
BTW when overclocking Zen 2, the struggle is that different workloads load the CPU differently. The old school method of fixed voltage and clock doesn't work well with that and IMO is relegated to competitive benching where you optimise for a given specific workload. PBO essentially turns off the CPU power limiter and allows it to run as fast as AMD think it can. At stock, it has a fixed power limiter enabled which is a bit above TDP e.g. 88W PPT for 65W TDP parts. I actually go the other way now. There is an ECO mode that sets PPT at or slightly below TDP. You don't really lose much clock from single or multi-thread loads from that, but you move into a much more power efficient operating area.
In specific areas, one or other CPU may be better, but right now Zen 2 offers decent overall performance, at a decent price, with some degree of platform longevity that isn't known on Intel side (the next gen desktop is rumoured to have yet another socket change).
Intel CPUs are still as good as they were, just that they have lost the performance crown and I think that perception will hurt them more than any technical reason. There is no indication of a radical change on desktop with next generation, so any hopes will be on the one after that. We know they have a better architecture, known as Sunny Cove and found in Ice Lake mobile CPUs, but desktop will have a wait. They had said 10nm on desktop was not cancelled, but the feeling is they will do some kind of limited release to say they done it, than any mass market range.
BTW when overclocking Zen 2, the struggle is that different workloads load the CPU differently. The old school method of fixed voltage and clock doesn't work well with that and IMO is relegated to competitive benching where you optimise for a given specific workload. PBO essentially turns off the CPU power limiter and allows it to run as fast as AMD think it can. At stock, it has a fixed power limiter enabled which is a bit above TDP e.g. 88W PPT for 65W TDP parts. I actually go the other way now. There is an ECO mode that sets PPT at or slightly below TDP. You don't really lose much clock from single or multi-thread loads from that, but you move into a much more power efficient operating area.
OP
- Joined
- Feb 18, 2002
- Thread Starter
- #10
If the topics are 2020/2021 and Desktops (not mobile) and overclocking (not stock).
• It appears that things on the AMD side are clear cut: AMD Zen 3 Ryzen 4000 in 2020 followed by Zen 4 Ryzen 5000 in 2021.
• This morning news is that Intel Z490 motherboard chipset + Comet Lake-S CPUs are scheduled for April 2020.
So what would you say is a revolutionary feature coming in the next two years, that's more than just a clock for clock 18% performance increase and the like? Would you say it's PCIe4 and PCIe5 and DDR5?
Intel has none of that on the horizon, whereas Zen 3 Ryzen 4000 appears on track for PCIe4 and Zen 4 Ryzen 5000 will have DDR5 and PCIe5.
The curve ball here is the post above about AMDs not overclocking, I didn't know that, so I'm trying to gauge if current state of overclocking potential is any guide, what will 2020/2021 system comparisons be like when you compare overclocked versions of AMD and Intel, not stock?
• It appears that things on the AMD side are clear cut: AMD Zen 3 Ryzen 4000 in 2020 followed by Zen 4 Ryzen 5000 in 2021.
• This morning news is that Intel Z490 motherboard chipset + Comet Lake-S CPUs are scheduled for April 2020.
So what would you say is a revolutionary feature coming in the next two years, that's more than just a clock for clock 18% performance increase and the like? Would you say it's PCIe4 and PCIe5 and DDR5?
Intel has none of that on the horizon, whereas Zen 3 Ryzen 4000 appears on track for PCIe4 and Zen 4 Ryzen 5000 will have DDR5 and PCIe5.
The curve ball here is the post above about AMDs not overclocking, I didn't know that, so I'm trying to gauge if current state of overclocking potential is any guide, what will 2020/2021 system comparisons be like when you compare overclocked versions of AMD and Intel, not stock?
Those features are fairly useless as of now (2020). No cards that exist can really use PCIe 4 bandwidth (2x PCIE 3.0, as 3.0 was to 2.0). DDR5 isn't going to be a game-changer either. Faster (but to what end... memory doesn't do much now unless you are AMD due to CCXs), slightly lower power (what 10W to 8W?)... etc. Zen 2/Ryzen 3000/Z570 already have PCIe 4.0... PCIe 5.0 is a ways away and also likely useless upon launch. Remember, only today with a 2080 Ti is PCIe 3.0 bandwidth not enough......... and the difference is only ~1-2%... as it has been since that was tested (2.0-3.0 etc).
As far as overclocking on next gen/2 gens away... that is anyone's guess. However the writing on the wall is that out of the box both Intel and AMD have less headroom (AMD with almost literally none - the can barely get past their boost rate). You'll have to read between the lines on rumors and wait for an actual review.
- Joined
- Mar 7, 2008
On (lack of) overclocking, I think it can be seen either way. AMD are selling you a CPU that does pretty close to what it can without you having to tinker with it. To not fall behind, Intel are kinda forced to do similar.
To my personal uses, faster PCIe is of limited benefit until devices exist that can take genuine advantage of it. About the only use case I can see of more aggregate PCIe bandwidth is allowing more NVMe SSDs in a given system. Even then that's not a necessity, I'm running many SATA SSDs and the limit is more in low QD random reads than anything sequential.
Faster ram is where I'd disagree with ED. Intel really needs it. AMD less so. It is challenging to feed cores, and since the core wars broke out, the compute potential is growing much faster than ram bandwidth is to feed it. For now, Intel CPUs have smaller L3 caches, and in the case of AVX-512, a lot more FP64 potential per core than AMD. They will have to either do something like Zen 2 with big L3 caches, or throw much more bandwidth at it. I think this is an opportunity for Intel to redraw the lines on market segments. With Intel HEDT price cuts, I think they can move that down to higher end enthusiast level.
On Zen 2, because of the big caches, it helps mitigate the effect of ram bandwidth in many cases. So the main driver for fast ram wasn't so much the ram bandwidth, but the IF bandwidth. Since Zen 2, that can be set async so slower ram, fast IF is possible. I love how 3200 speed is now considered slow as people aim for the 3600 sweet spot.
To me, DDR5 can't come fast enough.
To my personal uses, faster PCIe is of limited benefit until devices exist that can take genuine advantage of it. About the only use case I can see of more aggregate PCIe bandwidth is allowing more NVMe SSDs in a given system. Even then that's not a necessity, I'm running many SATA SSDs and the limit is more in low QD random reads than anything sequential.
Faster ram is where I'd disagree with ED. Intel really needs it. AMD less so. It is challenging to feed cores, and since the core wars broke out, the compute potential is growing much faster than ram bandwidth is to feed it. For now, Intel CPUs have smaller L3 caches, and in the case of AVX-512, a lot more FP64 potential per core than AMD. They will have to either do something like Zen 2 with big L3 caches, or throw much more bandwidth at it. I think this is an opportunity for Intel to redraw the lines on market segments. With Intel HEDT price cuts, I think they can move that down to higher end enthusiast level.
On Zen 2, because of the big caches, it helps mitigate the effect of ram bandwidth in many cases. So the main driver for fast ram wasn't so much the ram bandwidth, but the IF bandwidth. Since Zen 2, that can be set async so slower ram, fast IF is possible. I love how 3200 speed is now considered slow as people aim for the 3600 sweet spot.
To me, DDR5 can't come fast enough.
- Joined
- Mar 7, 2008
I'm not alone in this, it also affects HPC and some types of server use cases so they will be the driving force for improvement. Consumers like us are relatively insignificant. AMD have more cores, but they also have way more L3 cache (per core, or total). That is a mitigation strategy against lack of ram bandwidth. It doesn't work in all scenarios though. I'm liking Zen 3 rumours which if true will make the smallest logical unit a CCD, not a CCX as currently and that more unified cache will enable better scaling to bigger things.
In many areas of computing we have far more performance than we need. Doesn't mean we shouldn't have them. Doesn't mean no one cares. Ram is just the one I'm more focused on.
Can't comment on that TPU link just yet, and I'm not sure I can be bothered to find something more along the lines of what I was thinking. Others have also introduced the variable of slower ram, faster IF, and while ram speed still plays a part, it is complicated by IF speed.
In many areas of computing we have far more performance than we need. Doesn't mean we shouldn't have them. Doesn't mean no one cares. Ram is just the one I'm more focused on.
Can't comment on that TPU link just yet, and I'm not sure I can be bothered to find something more along the lines of what I was thinking. Others have also introduced the variable of slower ram, faster IF, and while ram speed still plays a part, it is complicated by IF speed.
I agree... but this is the consumer space... not HPC/servers. Look at it from the perspective of the OP asking the question. For the OP, I think the "18% IPC improvement" will matter more than using DDR5 or PCIe 4.0/5.0. Honestly, it is tough to discuss benefits/facts now considering these things are not out and we do not know the relative performance increase for such things. We can only go by what we have now. And in general, for a home users, memory speeds do not matter much in most cases, but tend to matter more for AMD in that same space.
Anyway, I digress.
Last edited:
Alaric
New Member
- Joined
- Dec 4, 2011
- Location
- Satan's Colon, US
They compete with them... on clockspeeds and overclock capabilities. IPC is about the same (with perhaps AMD slightly in the lead), but the AMD CPUs really can't overclock much at all. To the point where unless you can utilize (not use) all the cores and threads, it is best to use their "PBO" instead.
It just depends on your needs, but in general, the AMD CPU is the better buy these days. Feel free to look at reviews covering this on the front page.
Over in the Cinebench R20 thread Kenrou, who has 25% more cores than I do, got a 33% scoring advantage over my Skylake. That was with his stock 4.2 GHz Boost compared to my 4703 MHz all core with HT enabled. I'd say more than a "little" IPC advantage. He trounced my old 'Lake. Slapped it around, took its lunch money and made it cry. LOL
https://www.overclockers.com/forums...ebench-R20-scores-here!?p=8131586#post8131586
Edit: The Ryzen 5 3600 is $195 on newegg right now.
Last edited:
They hold a significant advantage in SMT efficiency over Intel.
You'll also note he has 50% more actual cores (6 vs 4 - 2 more than 4 is 50%) and threads (12 vs 8 - 4 more than 8 is 50%), not 25%.
Last edited:
Alaric
New Member
- Joined
- Dec 4, 2011
- Location
- Satan's Colon, US
Well, I have 66% of his core count, while Kenrou has a 33% scoring advantage, so I just fell in love with my obsolete Skylake all over again. Except for his speed disadvantage-if he held all six cores at 4.2 GHz I have a 500 MHz "advantage", but I don't think his 3600 can do that. Would SMT provide that much help? Not being skeptical, I really don't know.
A few % IPC difference in ours... is that little (seems to be in an Intel context ;p)? Maybe not?
Here is our IPC testing and SMT efficiency testing...
https://www.overclockers.com/amd-ryzen-9-3900x-and-ryzen-7-3700x-cpu-review/
Our SMT tests show double digits...
Moar links for performance/IPC - https://www.tomshardware.com/reviews/ryzen-9-3900x-7-3700x-review,6214-3.html
https://www.guru3d.com/articles-pages/amd-ryzen-7-3700x-ryzen-9-3900x-review,9.html
Also only(?) a few percent.
Anyway, not entirely sure if all of this useful to the OP or not, so, I'm out like I said earlier.
Here is our IPC testing and SMT efficiency testing...
https://www.overclockers.com/amd-ryzen-9-3900x-and-ryzen-7-3700x-cpu-review/
Our SMT tests show double digits...
Moar links for performance/IPC - https://www.tomshardware.com/reviews/ryzen-9-3900x-7-3700x-review,6214-3.html
https://www.guru3d.com/articles-pages/amd-ryzen-7-3700x-ryzen-9-3900x-review,9.html
Also only(?) a few percent.
Anyway, not entirely sure if all of this useful to the OP or not, so, I'm out like I said earlier.
Last edited:
OP
- Joined
- Feb 18, 2002
- Thread Starter
- #20
I upgrade rarely. Only two times since AMD Athlon XP. Therefore a system capable of taking all the advantages coming in the next half decade or so is what I will be looking at.
I am on the first Skylake motherboard right now.
So it'll be between the new AMD socket, or Tiger Lake since Ice Lake will not be revolutionary enough.
I am on the first Skylake motherboard right now.
So it'll be between the new AMD socket, or Tiger Lake since Ice Lake will not be revolutionary enough.
Similar threads
