I think that Windows is written in Visual C/C++ in addition to other languages and that it would be a fair amount of work to get it running under Objective C. Microsoft could also pay Apple to port it. I don't think that Microsoft is likely to port to Apple Silicon unless they get the right to build their own machines with the chips.
Microsoft future in in Linux! I think they can do a version in Unix too, I think that what Apple OSes are from Next Computer. Apple would make easy to do if Microsoft wanted to do it. They could share all their Mac and iOS drivers which would cut the time in half since it been tested already. Apple could also be a part of testing the Silicon Windows too. Apple has very good quality operating system testing staff.
Microsoft future in in Linux! I think they can do a version in Unix too, I think that what Apple OSes are from Next Computer. Apple would make easy to do if Microsoft wanted to do it. They could share all their Mac and iOS drivers which would cut the time in half since it been tested already. Apple could also be a part of testing the Silicon Windows too. Apple has very good quality operating system testing staff.
Ugh, it is to late for Microsoft! Apple started doing their own silicon with a purchased of a silcon design company in 2007. Even AMD, Nvidia and Intel are to far behind Apple and Apple will just accelerate their silicon chip power. This is Apple opening M1. Apple is already tested their M4. They still have M2-M3 to be release into the marketplace, fully done.
Ugh, it is to late for Microsoft! Apple started doing their own silicon with a purchased of a silcon design company in 2007. Even AMD, Nvidia and Intel are to far behind Apple and Apple will just accelerate their silicon chip power. This is Apple opening M1. Apple is already tested their M4. They still have M2-M3 to be release into the marketplace, fully done.
I'd agree that they already have working silicon given the way chips are designed these days. It seems a fairly easy thing to just add cores to CPU given that we have 1, 2, 4, 6, 8, 10, 16, 18, 32, 64 core chips these days. I don't like the configurations that they offer but what they have released is the biggest bang for the buck out there in terms of absolute performance and performance per watt and usability without a lot of RAM.
It's relatively easy for AMD thanks to the chiplet design as long as you have the space on the motherboard to accommodate the giant Threadripper and Epyc class CPUs. Meanwhile, Intel is literally going down in core count in order to attempt to stay relevant. It doesn't seem that trivial to me; it just so happens that AMD designed their CPUs with that in mind, providing more useable chips per wafer by making smaller parts. You have to physically put the extra cores somewhere after all.
Having more software scale with core count seems to be the challenge, as well as not having performance tank on CPUs with high core count the way it happened with 2nd gen Threadripper especially on Windows' scheduler compared to Linux's (mostly fixed with 3rd gen), but that's another topic that is of little concern to most consumers.
It's relatively easy for AMD thanks to the chiplet design as long as you have the space on the motherboard to accommodate the giant Threadripper and Epyc class CPUs. Meanwhile, Intel is literally going down in core count in order to attempt to stay relevant. It doesn't seem that trivial to me; it just so happens that AMD designed their CPUs with that in mind, providing more useable chips per wafer by making smaller parts.
Having more software scale with core count seems to be the challenge, as well as not having performance tank on CPUs with high core count the way it happened with 2nd gen Threadripper especially on Windows (mostly fixed with 3rd gen), but that's another topic that is little concern to most consumers.
Intel has high core counts on their Xeon and Xeon Phi chips.
It's up to the software guys, sure, but the development tools are a lot better these days for that. My trading programs work fine with more cores but I have a gazillion charts running and they can just throw threads or cores at those charts in parallel. Same with email, web browsing, etc.
The packaging of M1 is that you get a great CPU but you have to get Apple packaging and the packaging is never cheap. But the price/performance/thermals/performance per watt is crazy good for the M1 models. They just cost a small fortune when you add RAM and SSD beyond the minimum.
In all honesty, those chips might as well not exist. Their 56 core CPU is about as tangible as mist, and even their HTPC 18 cores 10980XE, which is little more than a repackaged 9980XE, which already was essentially a better binned 7980XE, has never been particularly available. Something that the most cynical folks may say makes sense considering it performs worse than a Ryzen 3950X and isn't worth the wafer allocation; especially as it was released at a time where Intel struggled to cope with 14nm demand. Even their 10900K 10 core CPU is hard enough to find, and seemingly to bin for, that Intel themselves released a 10850K instead with slightly lower clock speeds.
For CPUs you can actually purchase, only AMD offers those absurd core counts.
In all honesty, those chips might as well not exist. Their 56 core CPU is about as tangible as mist, and even their HTPC 18 cores 10980XE, which is little more than a repackaged 9980XE, which already was essentially a better binned 7980XE, has never been particularly available. Something that the most cynical folks may say makes sense considering it performs worse than a Ryzen 3950X and isn't worth the wafer allocation; especially as it was released at a time where Intel struggled to cope with 14nm demand. Even their 10900K 10 core CPU is hard enough to find, and seemingly to bin for, that Intel themselves released a 10850K instead with slightly lower clock speeds.
For CPUs you can actually purchase, only AMD offers those absurd core counts.
All that while being slower in all accounts than AMD and being more expensive, when it's not made of unobtainium. For the price of both kidneys and a liver, you could also buy the Xeon Platinum 9282, Intel's 56 core CPU they don't bother giving any pricing for; for a frame of reference, the Xeon Platinum 8280 was announced at $10,000. Realistically the company that can really consistently deliver extremely high core count CPUs that have any kind of relevance performance wise is AMD. Until that changes, I'd disagree with the notion that adding more cores to a CPU is "fairly easy".
I do realise it's pedantry or a small hill to die on though. Intel can't really do it, nor have I heard of any ARM based silicon that can do it. It doesn't take anything away from how good Apple's or anyone's silicon is.
All that while being slower in all accounts than AMD and being more expensive, when it's not made of unobtainium. For the price of both kidneys and a liver, you could also buy the Xeon Platinum 9282, Intel's 56 core CPU they don't bother giving any pricing for; for a frame of reference, the Xeon Platinum 8280 was announced at $10,000. Realistically the company that can really consistently deliver extremely high core count CPUs that have any kind of relevance performance wise is AMD. Until that changes, I'd disagree with the notion that adding more cores to a CPU is "fairly easy".
I do realise it's pedantry or a small hill to die on though. Intel can't really do it, nor have I heard of any ARM based silicon that can do it. It doesn't take anything away from how good Apple's or anyone's silicon is.
It clearly isn't. Otherwise they wouldn't go down in core count on their relatively lower end consumer chips on their 11th gen. Otherwise, the Mac Pro wouldn't be as out of place performance wise compared to AMD at such a price. Otherwise, Apple wouldn't make their own silicon due to Intel's lack of progress. Otherwise, Intel wouldn't make the 10850K as a stop-gap to the 10900K. Otherwise, their 18 cores 10980XE would be more readily available, even if it gets embarrassed by the 16 cores 3950X (nevermind the newer 5950X). Otherwise, their 8280, 28 cores CPU wouldn't be at $10,000 when AMD's 64 cores 3990X retailed for $3,990; all that with their 9290 chip with 56 cores at an "undisclosed" price. I consider that very different to say that Intel makes them and that Apple has a tray of them, and saying that it's easy for Intel to do. It doesn't look like it to me.
It clearly isn't. Otherwise they wouldn't go down in core count on their relatively lower end consumer chips on their 11th gen. Otherwise, the Mac Pro wouldn't be as out of place performance wise compared to AMD at such a price. Otherwise, Apple wouldn't make their own silicon due to Intel's lack of progress. Otherwise, Intel wouldn't make the 10850K as a stop-gap to the 10900K. Otherwise, their 18 cores 10980XE would be more readily available, even if it gets embarrassed by the 16 cores 3950X (nevermind the newer 5950X). Otherwise, their 8280, 28 cores CPU wouldn't be at $10,000 when AMD's 64 cores 3990X retailed for $3,990; all that with their 9290 chip with 56 cores at an "undisclosed" price. I consider that very different to say that Intel makes them and that Apple has a tray of them, and saying that it's easy for Intel to do. It doesn't look like it to me.
Those aren't just qualifiers: they're the reality of the current CPU market, that information is publicly available. "Fairly easy" for me means "readily available at prices not too far removed from the competition". I don't even say "cheap": just available at competitive prices. At least that's what I would call "fairly easy".
That being said, those core counts are a bit far removed from what Apple does at the moment with their M-series CPUs, especially on ARM process when they're on 4 cores at the moment with a different philosophy. It's another topic altogether. On the other hand, Intel is seemingly working on a big.LITTLE architecture for Alder Lake. For mobile devices, it could be interesting.
Those aren't just qualifiers: they're the reality of the current CPU market, that information is publicly available. "Fairly easy" for me means "readily available at prices not too far removed from the competition". I don't even say "cheap": just available at competitive prices. At least that's what I would call "fairly easy".
That being said, those core counts are a bit far removed from what Apple does at the moment with their M-series CPUs, especially on ARM process when they're on 4 cores at the moment with a different philosophy. It's another topic altogether. On the other hand, Intel is seemingly working on a big.LITTLE architecture for Alder Lake. For mobile devices, it could be interesting.
I'm not talking about the current. I'm talking about the ability to make multi-core processors. Something that they've done for a very long time in the past and they will make them for a very long time into the future.
They are currently on 8 cores, four performance, four efficiency. This is their low-end initial offering. I do not see the need to really buy anything from AMD or Intel once they come out with their mid-range and high-end CPUs.
I'm not talking about the current. I'm talking about the ability to make multi-core processors. Something that they've done for a very long time in the past and they will make them for a very long time into the future.
They are currently on 8 cores, four performance, four efficiency. This is their low-end initial offering. I do not see the need to really buy anything from AMD or Intel once they come out with their mid-range and high-end CPUs.
Their ability to make them is limited regardless. You buy what makes sense for you, that I do not dispute. I simply disagree on that making high core count CPUs is fairly easy, as only AMD does so reliably. Intel has shown the ability to make some, but it's clearly not "fairly easy" for them; nowhere near as easy as it is for AMD. I'd be curious to learn how far ARM can scale in that regard actually.
Did I mention it was a small hill to die on? We probably should move on.
Their ability to make them is limited regardless. You buy what makes sense for you, that I do not dispute. I simply disagree on that making high core count CPUs is fairly easy, as only AMD does so reliably. Intel has shown the ability to make some, but it's clearly not "fairly easy" for them; nowhere near as easy as it is for AMD. I'd be curious to learn how far ARM can scale in that regard actually.
Did I mention it was a small hill to die on? We probably should move on.
I don't have Intel insider knowledge to assess that; I see what information is available and I draw conclusions from it. Regardless, it's a package deal: a CPU design that can actually be built. Even their latest 2nd gen Xeon Scalable server processors are mostly a refresh with a price cut. As for Apple, they still uses the same foundries as AMD in TSCM, albeit on a different ARM-based architecture and on a different process node. It's a completely whole other topic, one on which I'll keep the same stance as earlier: I'll remain patient, I'll wait and see what happens.
Anyway and amidst Nvidia acquisition endeavours, seems ARM is trying to get into the data centre market more aggressively. Now, ARM isn't technically making those CPUs, but it's still indication that they believe that in some aspects, they can be competitive with x86_x64 even in the server space.
Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
www.anandtech.com
Microsoft unveiled their Pluton Hardware Security platform in partnership with Qualcomm, AMD and Intel. I'm not too excited into the user depending on Microsoft's whims regarding security. Installing Linux on a Surface device already isn't straight forward, if you're one who likes the hardware but not the software.
Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
www.anandtech.com
AMD's is introducing PBO2, which comes with Adaptive Undervolting. Unfortunately, it's restricted to 5000 series desktop CPUs and 500 series motherboards, with 400 series motherboards support coming "later".
I don't have enough insider knowledge to assess what specifically Intel does well or not: I see what information is available and I draw conclusions from that. Regardless, it's a package deal: a design that can actually be built. That's their 10nm problem: can't make it reliably. Even their latest 2nd gen Xeon Scalable server processors are mostly a refresh with a price cut. I wish Intel was more competitive, but it's not the case. As for Apple, they still uses the same foundries as AMD in TSCM, albeit on a different ARM-based architecture on a different node. It's a completely whole other topic, one on which I'll keep the same stance as earlier: I'll remain patient, I'll wait and see what happens.
Anyway and amidst Nvidia acquisition endeavours, seems ARM is trying to get into the data centre market more aggressively. Now, ARM isn't technically making those CPUs, but it's still indication that they believe that in some aspects, they can be competitive with x86_x64 even in the server space.
Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
www.anandtech.com
Microsoft unveiled their Pluton Hardware Security platform in partnership with Qualcomm, AMD and Intel. I'm not too excited into the user depending on Microsoft's whims regarding security. Installing Linux on a Surface device already isn't straight forward, if you're one who likes the hardware but not the software.
Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
www.anandtech.com
AMD's is introducing PBO2, which comes with Adaptive Undervolting. Unfortunately, it's restricted to 5000 series desktop CPUs and 500 series motherboards, with 400 series motherboards support coming "later".
Intel has designed multicore products that worked. They have big process issues. That doesn't mean that they don't know how to design and make them..
IBM makes up to 240-core System Z chips and those are used in the real world.
A lot of people are upset with Apple that they released their low-end chip first. But that makes sense from a financial perspective.
The thing about Apple is that they have an incredible amount of cash and they can just buy any talent that they need. From AMD, Intel or IBM. And they're taking their time and choosing their battles carefully.
Apple M1X:-12 Cores.- 8 performance cores.- 4 high efficiency cores.- Coming first on a MacBook Pro 16” unveiling as a press release.- According to a source who used a prototype, “if you think M1 is fast, you haven’t seen M1X”.-Name isn’t final though. pic.twitter.com/tpBhXpDCadNovember 22, 2020
The information comes from LeaksApplePro on Twitter, an account with a decent track record for all things Apple. Citing an insider who used a prototype, the next chipset from Cupertino could be called the Apple M1X. It will supposedly debut on the next 16-inch MacBook Pro in 2021. More interestingly, it is claimed to move to a 12-core configuration with eight high-performance cores and four for efficiency. The leak also mentions “if you think M1 is fast, you haven’t seen M1X”.
It's not where I'm coming from. What I see on the market is what I talked about: Intel struggles to make have high-core count chips, much less competitive ones or anywhere near the core count of AMD. This was the information I use to say that it's not "fairly easy" for them. This was the tiny, little hill. What reason they have for it is another topic.
Honestly last time I heard of them is when they bought Red Hat. They seem to mostly invest in cloud and AI at the moment, but as far as the CPU space goes, I really haven't heard of them in a long while. I had to look up their latest POWER CPUs: 16 cores with 15 available due to yield issues on Samsung 7nm process. It is the same process node NVidia uses for their latest Ampere GPUs, and fairly ironically NVidia also struggles to cope with demand for those GPUs, not expecting to meet demand until 2021. IBM seems to be focused on providing services to their (fairly captive) user base moreso than making strides in the larger CPU market. Good on them to keep alive the RISC-based processors; honestly their 8 threads per core design looks pretty cool.
Once that said, I don't like comparing different architectures like between them since they target different things and you don't get to see them often in apples to apples comparison. Apple is a bit different in that they're not only transitioning from one into another, but they seem to care a bit about the desktop market so wait and see how it scales. It could be the first really relevant ARM desktop CPU. We'll see in a few years.
For those looking to build a desktop gaming and/or production PC during the end of year stretch despite the very real shortage issues, here are Gamers Nexus CPU recommendations for 2020, with a few (fairly well-deserved) shots here and there.
When your entry level wipes out your flag ship from just a month before.
movdqa going Hitler bunker with how much less heat.
Their M1 vs Intel mini with 2080 external GPU was like 12 watts versus 250 for same performance. Is it going to be possible for flagship M1 with inegrated graphics to beat 3080?
When your entry level wipes out your flag ship from just a month before.
movdqa going Hitler bunker with how much less heat.
Their M1 vs Intel mini with 2080 external GPU was like 12 watts versus 250 for same performance. Is it going to be possible for flagship M1 with inegrated graphics to beat 3080?
When your entry level wipes out your flag ship from just a month before.
movdqa going Hitler bunker with how much less heat.
Their M1 vs Intel mini with 2080 external GPU was like 12 watts versus 250 for same performance. Is it going to be possible for flagship M1 with inegrated graphics to beat 3080?
No on the graphics. M1 Geekbench 5 OpenCL is 1bout 18,000. They could double performance to 36,000 but then it's just comparable to a mid-range external card. The discussions are saying that they are either going to permit internal or external GPUs or build their own dGPU for the MacBook Pro 16, iMac Pro and Mac Pro. There currently isn't support for an external GPU in Big Sur for Apple Silicon but it could always be added. Apple has been keeping a much tighter lid on leaks for Apple Silicon than for past product launches.
I was going to wipe the disk of a Late 2009 iMac 27 inch using a Firewire cable (no video problem), using Target Disk Mode and then the iMac booted and ran just fine. This has a Core 2 Duo, 4 GB RAM and a 1 TB HDD. It's old but performance is surprisingly good. The display is fantastic as is the audio system. I am using it as a television set and general basement system. I ordered 8 GB of RAM to get it to 12 GB of RAM which should be plenty. It was swapping about 1 GB and you don't want to swap with a HDD. I'm giving some consideration to replacing the HDD with an SSD though I don't think that it really needs it.
This means that I have a spare MacBook Pro and a spare QHD 25 inch monitor as I was using those before. It might be getting time to sell my 2014 MacBook Pro 15.
I am going to wait on an Apple Silicon system as I don't really need one given what I currently have.
I also repurposed an old iPad Mini 2 which somewhat decreases the need for Macs around the house.
I finally did order the 2 TB NVMe for the new desktop. So this desktop has almost $600 in SSD storage. I'm not sure if I will replace the current SATA3 SSD with the NVMe or leave it in there. There's still quite a bit of space remaining which makes it handy as a NAS but I'm trying to anticipate future needs.
Something that I just noticed today is that one of my trading programs is using significant GPU resources. The peak usage is 60% under the category: 3D. GPU RAM usage is running around 75%. I was looking around on Amazon for something in the upper-midrange and I was surprised to see that the supply on mid-range and high-end graphics cards is near non-existent. It's not really an issue for now but something that I'll think about for a month or two. Hopefully supply gets better in January. I could always just add a second video card to offload some of the work from the first.
I'm not sure what suddenly caused the greater GPU usage but I would guess that it was the big Windows Update this past Tuesday. Microsoft may have optimized some of the graphics libraries to use GPU compute from scalar CPU computer. The program is written in Java so it's possible that Java is doing more GPU compute (unlikely I think) too.
So RAM for the iMac arriving in the next week, NVMe arriving for the new desktop in the next week, thinking about a new GPU.
Having an SSD increase speeds at boot time and overall system in general. SATA 3 SSD are at give away prices now especially when people are moving to M2 form factor and competition is trying to move product during holidays. I think you should upgrade if you plan on keeping this PC since it going speed everything up by 3X+!
Having an SSD increase speeds at boot time and overall system in general. SATA 3 SSD are at give away prices now especially when people are moving to M2 form factor and competition is trying to move product during holidays. I think you should upgrade if you plan on keeping this PC since it going speed everything up by 3X+!
I have several spare SATA3 SSDs so I could just swap one in. But it's a significant amount of work and one mistake and it's a brick. 2009-era hardware was SATA2 so it would be SATA2 speeds at best. In some cases, using a SATA3 drive with a SATA2 controller will result in SATA1 speeds. I have this problem on an old laptop and it was something about the particular Intel chipset that resulted in this result. I leave the system up all the time so boot time isn't an issue and I'm adding 8 GB of RAM so that it has 12 GB. I only use about 5 GB of RAM so the other 7 GB can be used for disk cache.
That's one reason I like to have twice as much RAM as what I actually need. Almost all of the stuff you need from the HDD or SSD is in RAM and super-fast.
One small bonus to using the 2009 iMac - I found two Mac IR Remotes. These are really small and thin, and came with iMacs and some MacBook Pros prior to 2012. I still had two of these in a desk drawer. I tried them both out and they still both work - the batteries on them are original. One is from a 2007 MacBook Pro.
Still waiting for my iMac RAM and the 2 TB NVMe SSD. I'm starting to think that the RAM was lost. It's supposed to be here Tuesday but there's no shipping information on it available. I would have been better off getting it from Best Buy.
I had a look at upgrading the HDD to an SSD and the only sticking point is the temperature sensor. The HDDs have a temperature sensor and you just connect a cable from the drive to the motherboard. The one solution I found is to buy a $35 cable from OWC which also has a temperature sensor in it and stick it onto the SSD. There is another solution and that's to get a DVD-ROM drive to SSD caddy and replace the DVD-ROM with the SSD. I'm going to just go with more RAM for now. I suspect that 12-16 GB of RAM will solve all of the performance issues on this system.
It has gotten slower when I leave it up for a long time. My guess is that Firefox RAM climbs over time and probably settles around 4 GB.
Overall system responsiveness is about the same with SATA or PCIe NVMe SSDs, or even bottom of the barrel DRAMless SSDs, so I suggest to most people not to overspend on their SSDs for the latest and shiniest. However for people frequently transferring files or installing programs, especially in local networking applications, it can save heaps of time. That being said, I would never recommend DRAMless SSDs to anyone as their main drive, or cheap QLC drives. When these things run out of cache, the performance can be absolutely dreadful.
It's one thing Apple did decently back then: pushing for nice SSDs even if it meant losing storage space while most laptops were still using painfully slow spinning drives. Then they soldered the damn thing to the board and that was a d*ck move.
EDIT: Especially since they had a nice thing in that you could basically unplug an SSD from a Macbook and move it to an identical one seamlessly. Was it economical? Nope, but at least it was easy. While Win10 is a lot more tolerant about plugin-in a system drive into a different computer, it's pretty bad practice. Shame Apple killed that; nevermind having your data locked onto a non-working device.
My RAM is stuck in PA. Apparently there are a lot of packages stuck in the USPS system that are in limbo. I found a lot of recent complaints on forums. It appears that a lot of packages get lost, destroyed or damaged from that particular post office area too. If it doesn't arrive on Tuesday, then I'll try to cancel the order and buy it at the local Best Buy. That is if they have it in stock. There's only one in right now and their next shipment is scheduled to arrive on Thursday.
I wish I could post charts here. Micron (they own Crucial and they're also the source of other memory brands) stock is up almost 50% in the past month. People are buying RAM and SSDs like mad!
Overall system responsiveness is about the same with SATA or PCIe NVMe SSDs, or even bottom of the barrel DRAMless SSDs, so I suggest to most people not to overspend on their SSDs for the latest and shiniest. However for people frequently transferring files or installing programs, especially in local networking applications, it can save heaps of time. That being said, I would never recommend DRAMless SSDs to anyone as their main drive, or cheap QLC drives. When these things run out of cache, the performance can be absolutely dreadful.
It's one thing Apple did decently back then: pushing for nice SSDs even if it meant losing storage space while most laptops were still using painfully slow spinning drives. Then they soldered the damn thing to the board and that was a d*ck move.
EDIT: Especially since they had a nice thing in that you could basically unplug an SSD from a Macbook and move it to an identical one seamlessly. Was it economical? Nope, but at least it was easy. While Win10 is a lot more tolerant about plugin-in a system drive into a different computer, it's pretty bad practice. Shame Apple killed that; nevermind having your data locked onto a non-working device.
If you want to upgrade an late model Apple Mac, just get a graduate electrical engineering student to do it for you! They have all the tools in their labs to remove the soldered SSD, RAM and replace it with the upgrade, solder it back if you want it soldered in. It look like it came directly from Apple assembly line soldered in.
Here comes Intel's 11th gen mobile processor, the i7-1165G7 as compared to Ryzen 4000 systems. Less of an incremental update and ore of an overhaul in a bid to be competitive against AMD's mobile offerings despite topping out at 4 cores and 8 threads. However Intel's strong single-core performance compared to Ryzen, as well as their improvements on their iGPUs compared to previous generation ones, may make Intel's 11th gen a relevant option in certain thin and light designs that simply don't do any multi-core workloads at all. At least until AMD's 5000 series mobile CPUs are released, seemingly in early 2021.
If you want to upgrade an late model Apple Mac, just get a graduate electrical engineering student to do it for you! They have all the tools in their labs to remove the soldered SSD, RAM and replace it with the upgrade, solder it back if you want it soldered in. It look like it came directly from Apple assembly line soldered in.
Here comes Intel's 11th gen mobile processor, the i7-1165G7 as compared to Ryzen 4000 systems. Less of an incremental update and ore of an overhaul in a bid to be competitive against AMD's mobile offerings despite topping out at 4 cores and 8 threads. However Intel's strong single-core performance compared to Ryzen, as well as their improvements on their iGPUs compared to previous generation ones, may make Intel's 11th gen a relevant option in certain thin and light designs that simply don't do any multi-core workloads at all. At least until AMD's 5000 series mobile CPUs are released, seemingly in early 2021.
That it's only 4 cores is definitely a problem in the performance category but their Xe graphics scores about 21K in Geekbench 5 OpenCL so it's a LOT better than UHD 630.
Apple may launch three new computers tomorrow. They have added three new Mac SKUs and they are doing hardware tomorrow.
There was an article in Bloomberg about their upcoming plans:
The current M1 has four performance and four efficiency cores.
Plans are for models with 8, 12 and 16 and 32 performance cores for Mini, MacBook Pro, iMac and Mac Pro.
The current M1 has 8 GPU cores and gets an OpenCL score of about 19,000.
Upcoming plans are for version with 16 and 32 GPU cores for integrated graphics. If true, this would put their integrated graphics around the mid-high end of the midrange discrete market.
A discrete GPU with 64-128 cores to overtake AMD and nVidia current discrete offerings.
GPU comparison: I have a GTX 1050 Ti which has a TDP of 75 Watts and it's comparable to the GPU in the M1 at about 10 Watts.
I do not expect Apple to ship an M1X system tomorrow but they would be sold out in about ten minutes if they did. I am expecting minor modifications like more ports, Gigabit Ethernet, more RAM offerings.
I'm prepping my 2014 and 2015 models to sell them. I plan to move applications from the MacBook Pros to my Windows desktop over the next week and then put them up for sale.
Apple is working on a new ARM-based processor with as many as 32 high-performance CPU cores that could appear in a Mac in late 2021, according to a new report from Bloomberg. The processor could also appear in a new “half-sized Mac Pro” in 2022.Alongside it, the company is also reportedly developing CPU designs with up to 16 high-performance cores and four power-efficient cores which are destined for new versions of the MacBook Pro and iMac. The new processors could arrive as early as spring 2021. Future Apple Silicon designs could also feature GPUs with up to 128 dedicated cores.
News of the upcoming processors comes as Apple has just released its first Macs powered by its own chips. The new MacBook Air, MacBook Pro, and Mac mini each used the company’s M1 chip, which features a CPU with four high-performance cores and four high-efficiency cores. However, the company’s more powerful machines like the Mac Procontinue to use Intel chips. Apple has said it intends to transition its whole Mac lineup onto its own chips over the course of two years.
As well as the increased CPU core count, Bloomberg reports that Apple is also working on chips with more GPU cores. While the current M1 chip comes with either seven or eight GPU cores, Apple is currently testing models with 16 and 32-cores, and is working on chips with as many as 128 for late 2021 or 2022.
The M1 chip has up to eight dedicated GPU cores.Image: Apple
Although Apple is working on a processor design with 16 high-performance CPU cores, Bloomberg notes that it might choose to release it with only eight or 12 cores enabled, depending on how mass production goes.
Prior to Apple’s announcement of its switch to ARM-based processors, Bloomberg reportedthat it was working on a processor with a CPU featuring eight high-performance cores and four energy-efficient cores. A processor with this exact combination of cores is yet to be officially announced.
Considering Apple’s ambitious plans to move its entire Mac lineup onto its own silicon in the next two years, it makes sense that it has more powerful chips in development. Its first ARM-based Macs have impressed thanks to their combination of power-efficiency and performance, but matching the capabilities of its more powerful Intel-based machines like the Mac Pro is likely to be a far bigger challenge.
Apple M1 GPU is 7 or 8 cores in the MacBook Air. It's 8 in the other models. 8 GPU cores is about 20K OpenCL. 32 would be more than 1 1660. I think that the latest from nVidia and AMD are in the upper 160K-200K.
8 GB of RAM is in. I installed it and the iMac shows 12 GB.
This means that I can run more programs at the same time and not have to worry about pauses moving from one program to the next. I have to say that this old machine is quite pleasant to use even though it's 2009 tech. I use my much faster Windows system for work but the iMac is really great for email, writing, browsing and watching videos. Performance is fine for what I'm using it for as well.
I was going to Hackintosh the 435mt as it's far faster than the iMac but I started reading on OpenCore and decided to just keep it as a backup Windows system.
I might order a Mini or MacBook Air/Pro if the configurations are interesting tomorrow. My guess is that I wouldn't get it until February though. The BTO models of the existing system are really getting out there in time.
I wonder how many channels of memory that makes on an Apple paltform. Cheap PC enthusiasts used to add 4GB of RAM to 8GB systems (be it in a 4GB+8GB or 3*4GB config), but there was either no performance benefit, or in some cases a performance penalty. The extra 4GB tend to be slower to access. But to be honest, it was mostly relevant to memory sensitive applications (like gaming), so who cares for a backup system. Better than nothing.
I wonder how many channels of memory that makes on an Apple paltform. Cheap PC enthusiasts used to add 4GB of RAM to 8GB systems (be it in a 4GB+8GB or 3*4GB config), but there was either no performance benefit, or in some cases a performance penalty. The extra 4GB tend to be slower to access. But to be honest, it was mostly relevant to memory sensitive applications (like gaming), so who cares for a backup system. Better than nothing.
The RAM in this system is competing with a HDD so it's a really easy win.
Most of what I'm running is in the cloud. Mail and Notes are local and in the cloud so there's an advantage to caching all of my notes and emails in RAM instead of having to go to the HDD.
My 2008 435mt is a much faster system with a Core i7-920 Bloomfield and triple-channel memory compared to the Core 2 Duo that came in the Late 2009 iMac as the default CPU option. They had Core i5 and i7 options which would have much better performance as they're a later generation of CPU and they are quad-core over dual-core. The limits for the Core i5 and i7 options on memory are 32 GB compared to 16 GB for the Core 2 Duo.
The Geekbench 5 scores are 378 for single core and 668 for multicore for the iMac and 491 for single core and 1,926 for multicore so my 2008 Dell XPS Studio 435mt smokes the iMac for raw performance. But I find that the old iMac is more comfortable for email, writing, working out, watching videos, etc. I would guess that it's the consistency of macOS vs the wide variety of UI paradigms that we see in the Windows world.
In general, I prefer macOS to Windows but my trading programs run better on Windows. If we get an M1 variant with eight performance cores, though, I'll switch as it would have about 80% more horsepower than my desktop at probably one quarter to one third the power consumption. If we get an M1 variant with twelve performance cores, then it will be absolutely no contest. Intel won't have anything that can compete. AMD has the 5050X which would be in the ballpark but Apple would have a huge power advantage.
This is, of course, assuming that I can get at least 32 GB of RAM and that it supports 3x4k.
Even with this as a work rig, though, I think that I'd still use the iMac when I'm not in work mode.
The NVMe SSD just arrived and I'll pop it in tonight or tomorrow. I'm still undecided about what to do with the existing SATA3 SSD that's already in there. I may just leave it in there. I have spare 500, 180 and 120 SSDs as well which I can put in the system but it's a lot easier to just have a large, flat space for storage.
I've been watching Memory statistics on the iMac and it's going as I expected.
Memory used is 4.97 GB and cached files are an additional 3.07 GB. I'll leave it running for a few days and see if it creeps up anywhere near 12 GB. If it does, then I guess I could consider going up to 16 GB. I rather doubt that it will do that though.
Update: I opened several YouTube tabs on Firefox and memory usage ran up to 5.83 GB. That's about as many tabs as I'll have open at once though. There are nine YouTube tabs open.
In dual channel, you use matched sticks on both channels. You want to make full use of the bandwidth on the two channels.
In my case, I had 2+2, so one 2 on one channel and one 2 on the other channel. You could put them both on the same channel but you'd be wasting the bandwidth on the other channel.
In the new configuration, there's 2+4 on one channel and 2+4 on the second channel.
Why Is Apple’s M1 Chip So Fast? Real-world experience with the new M1 Macs has started ticking in. They are fast. Real fast. But why? What is the magic?
On YouTube, I watched a Mac user who had bought an iMac last year. It was maxed out with 40 GB of RAM costing him about $4,000. He watched in disbelief how his hyperexpensive iMac was being demolished by his new M1 Mac Mini, which he had paid a measly $700 for.
In real-world test after test, the M1 Macs are not merely inching past top-of-the-line Intel Macs, they are destroying them. In disbelief, people have started asking how on earth this is possible?
Real world experience with the new Macs has sunk in. They are fast. Real fast. But why? What is the magic?
debugger.medium.com
This is a article on computer architecture but the author tries to dumb it down quite a bit so that it's understandable if you don't have an EECS degree. There are two main reasons why Apple's M1 chip is so fast and why it is quite unlikely for Intel and AMD to overcome these two reasons.
1) Apple is putting functions that have traditionally been done in software in hardware where they can run faster. So they can do video encoding/decoding in hardware instead of software. They can look at what their customers are doing that is CPU intensive and then add hardware functions to do those same things so that they run a lot faster. Intel and AMD do this for integrated GPU and various other things like USB or Thunderbolt. But Apple is doing a lot more of it. A lot of Apple customers do video editing, audio work and this stuff typically requires a big CPU and a lot of RAM - but Apple has put moved the functionality in hardware so that the little M1 can beat systems costing two to five times as much. It would be quite difficult to overcome this because Intel and AMD make General Purpose CPUs that will do a wide variety of tasks. The problem with adding a bunch of tasks to their chips is determining which tasks to add to their chips. Apple can simply decide because they are one company whereas Intel, AMD, nVidia, Dell, HP, Lenovo and Microsoft could all disagree on what should go into the chip. Someone would have to do the support and maintenance as well.
2) Intel and AMD use four decoders to translate x86 instructions into MicroOps and then these MicroOps are sorted to maximize execution parallelism. Apple uses eight decoders and so Apple can execute far more instructions per clock compared to Intel and AMD. The reason AMD and Intel can only have four decoders is because x86 instructions are variable-length while ARM instructions are fixed length. So an ARM decoder knows where the next eight (or 20) instructions start and it's simple to do the decoder. x86 instructions are variable length so you don't know where the next instruction starts until you look at the current one. AMD and Intel solve this by looking at all potential starting points and then tossing out the ones that are wrong. This obviously makes their decoders far more complex.
The article notes that one of AMD's cores is faster than the M1. But that's because it is running at 60% higher clockspeed. Apple could also run their chips at higher clockspeeds but their design goal is power efficiency so they're just going to add more cores for customers that want more power.