Old PC died - Building new one

[dgianstefani]

I absolutely agree with you on the 7900X3D in particular though. I have no idea why that CPU exists.

Both the 7900/X and the 7900X3D exist because AMD needs to get rid of defect eight core dies and consumers for the most part don't know about CCD/CCX etc. The logical performance solution would be to make an 8+4, so you have none of the 6+6 disadvantages compared to 8+0 or 8+8, but there can't be many dies that fail four of their cores. Plenty that fail one or two though.

There's some weird server chips that have just two cores per CCD active, but the full 3DVCache, so you get massive cache per core.

Actually while the thread has taken a slight detour it has allowed me to get a better look at the AMD chips. I started doing more research on CCD/CCX. In the mean time I'm got a docking station which allows me to connect my laptop to my Dell U4919W monitor.
While I'm more of an AMD fan the laptop is an Intel i7-13800H which is 14 cores and 20 threads.
It has 64Gb
And a 2Tb NVMe

So I'm using it as my development system at home for now while I do more research.
This PC lasted 7 years
My prior PC was 10 years
I try to get the mid-range
7950X while this is a nice CPU, too expensive
And the 7900X was 12 cores - 24 threads which I felt was something that could last me 5-7 years again
I also changed the memory to 64Gb, instead of 32
Still flip flopping on whether to do
3 x 1tb
or
1 1tb and 1 2tb

I also recently purchased a NAS storage system, Terra-master F2-212 with 2 8tb drives in RAID
The drives arrived the same day my PC died and the Terra-master unit came the next day.
Luckily I was able to extract the data from my drives since I had an external reader.

Good work going with 64 GB.

 

[Chrispy_]

Both the 7900/X and the 7900X3D exist because AMD needs to get rid of defect eight core dies and consumers for the most part don't know about CCD/CCX etc. The logical performance solution would be to make an 8+4, so you have none of the 6+6 disadvantages compared to 8+0 or 8+8, but there can't be many dies that fail four of their cores. Plenty that fail one or two though.

There's some weird server chips that have just two cores per CCD active, but the full 3DVCache, so you get massive cache per core.


Good work going with 64 GB.

One thing to note - if you're buying 2x32GB now and plan to double it to 4x32 down the line, don't get bleeding fast DDR5-6000-CL30 because you'll have no chance of running 128GB of that stuff. Whilst it will cost you a few percent in performance today, the safest speed to buy for half-now, half later DDR5 on Ryzen is going to be an EXPO-specific DDR5-5200-CL36 which is the fastest I've reliably had two 64GB kits just plug-and-play together at EXPO/XMP speeds for 128GB at >4800 JEDEC speeds.

As for the 7800X3D vs 7900/7900X - these discussions are all good and true, if you're reading benchmark reviews where an RTX4090 is used to eliminate the GPU from the equation - and that's the correct way to benchmark CPUs.

In the real world, and with a 3060Ti, there's absolutely no point getting a 7800X3D and losing four cores just to get additional 3D-vCache for gaming, because the 3060Ti will be the bottleneck for any CPU at normal resolutions and framerates. Whatever CPU you have, it's not going to be maxed out until you spend north of $1000 on gaming-specific graphics cards.

 

[Princess Garnet]

It's faster than a regular 7900X in games and slightly worse in productivity. It serves the role it was made for, it's very niche but it has it's purpose.

That's true, so I guess I see it as a little too niche and it shows.

I'm looking at it against more than the 7900/X itself. If I wanted better game performance, I'd choose a 7800X3D, and if I wanted better multi-threaded performance, the 7900/X. I don't think I'd ever want to compromise but that might be because it has less v-cache cores than the 7800X3D. I do think there's room in the market for a hex core 3D cache model, but as a budget Ryzen 5 instead.

 

[Chrispy_]

In the TPU review, W1zzard noted the 2nd CCD (or whatever) ran at marginally lower clocks. So there is a performance hit on the parts that operate this way, compared to those that don't. It's why I bought an 8 core Ryzen - to avoid such complex interoperability.

And, FTR, given recent thread developments, please don't shadow bait other members (your final sentence of your post).

Unless you're going to push PBO way beyond the stock 170W cTDP (230W PPT) - you are only going to engage cores in the slower CCD when you are running something all-core, at which point the 230W PPT isn't really enough to push all cores up to the max boost clock of the preferred cores on the better CCD.

Realistically, it matters very little because the stuff that needs peak speeds is running on a couple of preferred cores in the faster CCD and you're power-limited by the time all cores are engaged and you're using the slower CCD too. Likely the Ryzen 9 CPUs get one highly-binned CCD that's good for "better than Ryzen 7" boost clocks, and one standard-bin CCD that's as good as a Ryzen 7 would be anyway. At any given voltage, the higher-binned CCD will run faster, so it makes sense that there's a speed imbalance fully-loaded. If both CCDs were binned equally you'd just see average clocks somewhere between the fast and slow CCD clocks in a Ryzen 9 and the benchmark scores would be indistinguishable.

 

[Princess Garnet]

One thing to note - if you're buying 2x32GB now and plan to double it to 4x32 down the line, don't get bleeding fast DDR5-6000-CL30 because you'll have no chance of running 128GB of that stuff. Whilst it will cost you a few percent in performance today, the safest speed to buy for half-now, half later DDR5 on Ryzen is going to be an EXPO-specific DDR5-5200-CL36 which is the fastest I've reliably had two 64GB kits just plug-and-play together at EXPO/XMP speeds for 128GB at >4800 JEDEC speeds.

Uh! Personal remarks, but thanks for sharing the information on this, even if it sounds disappointing.

I'm on AM4 with 64 GB (four DIMMs of dual rank at 3,600 MHz even) and I'm considering going to AM5 with the 9800X3D release. What I was going to do about RAM is in the air since I'm familiar with how difficult it is to drive four DIMMs at higher speed. I could certainly live with with 64 GB for now (~48 GB is as low as I could tolerate but I'm not going backwards) but I'd like to go up, and 64 GB DIMMs are (I think?) not out just yet so I can't find out yet how those will do. I suppose 2x 48 GB might be an option but I'm not sure if those will help improve the likelihood of getting above 64 GB at higher speeds.

I've been hearing Zen 5 likely won't see any new chipsets release, though supposedly they will see 6,400 MHz instead of 6,000 MHz be recommended as the sweet spot so I suppose that means the IMC improved slightly. Still, the board plays a role in how well it handles a given memory configuration too so... I'm lost as what I'll be likely to get away with. Perhaps a question for another day since the 9800X3D is likely up to a year out from now.

I know the X3D CPUs "lose out on performance less" than non-X3D CPUs with slower RAM but I'd still like to match the sweet spot of the IF clock if I can.

 

[Chrispy_]

That's true, so I guess I see it as a little too niche and it shows.

I'm looking at it against more than the 7900/X itself. If I wanted better game performance, I'd choose a 7800X3D, and if I wanted better multi-threaded performance, the 7900/X. I don't think I'd ever want to compromise but that might be because it has less v-cache cores than the 7800X3D. I do think there's room in the market for a hex core 3D cache model, but as a budget Ryzen 5 instead.

As I said earlier, the 7800X3D is a no-compromise gaming solution, but you can't leverage that advantage without a 4090.

If you can afford a $2000 4090, then why are we discussing the differences between two different $400 CPUs? Just get a 7950X3D and then there's no need to compromise. It's faster than the 7800X3D in gaming since the second CCD's cores are parked, making it a higher-binned 7800X3D, and when it's productivity time you have 16C/32T

 

[Princess Garnet]

As I said earlier, the 7800X3D is a no-compromise gaming solution, but you can't leverage that advantage without a 4090.

That's not entirely true. The GPU is simply more important for gaming performance on average, but you don't need to spend two grand on a GPU first to justify a 7000 series X3D. Most of the market agrees because most are choosing the 7800X3D over the 7950X3D (and the former seems to often be on par or faster anyway). Most people will take the same performance for less money. Most people don't have high multi-threaded needs.

I know Minecraft will certainly benefit from it too. I have a 5800X3D and if you start pushing render distances and want it to keep up with terrain generation or even just loading, especially while flying, then a faster CPU can certainly be very welcome here. I know most people probably don't think of that game but you don't always need an RTX 4090 to justify a 7000 series X3D.

 

[dgianstefani]

As I said earlier, the 7800X3D is a no-compromise gaming solution, but you can't leverage that advantage without a 4090.

If you can afford a $2000 4090, then why are we discussing the differences between two different $400 CPUs? Just get a 7950X3D and then there's no need to compromise. It's faster than the 7800X3D in gaming since the second CCD's cores are parked, making it a higher-binned 7800X3D, and when it's productivity time you have 16C/32T

You absolutely benefit from a 7800X3D with any tier of GPU.

7950X3D is slower in gaming by about 5%, not faster.

 

[AleXXX666]

Don't buy a 7900X it's a six core in disguise.

Either go for a 7800X3D or a 7950X.

Don't buy three 1 TB drives, get two or one 2 TB M.2

Also 32 GB is small for a workstation.

and 7800x3d is definitely "workstation" LMFAO....

 

[Why_Me]

So the MSI MAG B650 Tomahawk Wifi is about $212
where the MSI PRO X670-P WiFi ProSeries is $228

Or this.

PCPartPicker Part List

CPU: *Intel Core i7-13700F 2.1 GHz 16-Core Processor ($289.99 @ Best Buy)
Motherboard: *MSI B760 GAMING PLUS WIFI ATX LGA1700 Motherboard ($159.99 @ B&H)
Memory: *Silicon Power Value Gaming 32 GB (2 x 16 GB) DDR5-6000 CL30 Memory ($94.97 @ Amazon)
Total: $544.95
Prices include shipping, taxes, and discounts when available
*Lowest price parts chosen from parametric criteria
Generated by PCPartPicker 2024-02-06 00:11 EST-0500

Intel® Core™ i7-13700F Processor (30M Cache, up to 5.20 GHz) - Product Specifications | Intel

Intel® Core™ i7-13700F Processor (30M Cache, up to 5.20 GHz) quick reference with specifications, features, and technologies.

www.intel.com

MSI B760 GAMING PLUS WIFI, ATX - Intel 13th/12th Gen - 12 Phases, DDR5, PCIe 4.0, 2.5G LAN, Wi-Fi 6E

B760 GAMING PLUS WIFI is designed with tons of connectivity, flexible tools, and convenient Wi-Fi solution with DDR5 memory version for gamers who want all.

www.msi.com

 

[joepacelli]

Or this.

PCPartPicker Part List

CPU: *Intel Core i7-13700F 2.1 GHz 16-Core Processor ($289.99 @ Best Buy)
Motherboard: *MSI B760 GAMING PLUS WIFI ATX LGA1700 Motherboard ($159.99 @ B&H)
Memory: *Silicon Power Value Gaming 32 GB (2 x 16 GB) DDR5-6000 CL30 Memory ($94.97 @ Amazon)
Total: $544.95
Prices include shipping, taxes, and discounts when available
*Lowest price parts chosen from parametric criteria
Generated by PCPartPicker 2024-02-06 00:11 EST-0500

Intel® Core™ i7-13700F Processor (30M Cache, up to 5.20 GHz) - Product Specifications | Intel

Intel® Core™ i7-13700F Processor (30M Cache, up to 5.20 GHz) quick reference with specifications, features, and technologies.

www.intel.com

MSI B760 GAMING PLUS WIFI, ATX - Intel 13th/12th Gen - 12 Phases, DDR5, PCIe 4.0, 2.5G LAN, Wi-Fi 6E

B760 GAMING PLUS WIFI is designed with tons of connectivity, flexible tools, and convenient Wi-Fi solution with DDR5 memory version for gamers who want all.

www.msi.com

So I've never been a big intel fan. I've always built AMD system's except for one time when I built an Intel system and the migrated back to AMD with the release of Ryzen.
My question though when I was looking at the Core i7-13700K
It stated 16 cores/24 Threads.
Where the Ryzen is also X Cores and the threads are 2X
Why are there only 24 threads and not 32 for the Intel?
On the Amazon when you look at the Images it shows
Up to 16 Cores: 8 Performance-cores, 8 Efficient-cores
Up to 24 Threads: 16 on Performance-cores, 8 on Efficient-cores

would like to understand this

 

[Princess Garnet]

So I've never been a big intel fan. I've always built AMD system's except for one time when I built an Intel system and the migrated back to AMD with the release of Ryzen.
My question though when I was looking at the Core i7-13700K
It stated 16 cores/24 Threads.
Where the Ryzen is also X Cores and the threads are 2X
Why are there only 24 threads and not 32 for the Intel?
On the Amazon when you look at the Images it shows
Up to 16 Cores: 8 Performance-cores, 8 Efficient-cores
Up to 24 Threads: 16 on Performance-cores, 8 on Efficient-cores

would like to understand this

As someone who moved (back) to AMD CPUs lately, and then upgraded the CPU, and is planning to most likely have an AMD platform after this, don't lock yourself to AMD. Don't lock yourself to anything if you can help it. There's reasons a brand might be better for you but "never say never".

The reason for Intel's thread count not being double the core count is this. Intel has a hybrid architecture. They have performance cores, or "p cores", and they have efficient cores, or "e cores".

The performance cores are the traditional ones and those have Hyper-threading.

The efficient cores do not have any form of threading.

In the case of the 13700 series CPUs, they have 8 performance cores, each with Hyper-threading. So at its base it's an 8 core/16 thread CPU. It then has a further 8 efficient cores. So, 16 cores/24 threads.

The efficient cores aren't great, but they get way more hate than they deserve. The idea is this; the number of things that need a lot of performance cores are low. That's not to say the things that scale linearly are low, no. But anything that scales linearly is better off with efficiency cores. Why? Simple, you get around four of them in the same space as one performance core would take up, and four efficient cores total more performance than one performance core for that type of workload (for "non linear" threaded stuff they are indeed worse). The real hurdle with them is scheduling issues, not that "they're weak", but AMD's multi-CCD CPUs have "similar" issues of their own.

If you're not planning to do a lot of highly threaded tasks, I'd sort of ignore the efficient cores up front for a moment. You'd basically be looking at the Core i7 and Core i9 as octo cores at their foundation, and that's plenty for any games or the like, but then the efficient cores exist on the side too so they won't be overwhelmed by background tasks either.

Intel's fine right now, and that's coming from someone on AMD and planning to move to AMD next. They have higher power draw, and they're certainly losing to the Ryzen 7 X3Ds for gaming price/performance, but without the v-cache helping, Intel still holds the per core performance edge by a bit. Both are viable right now. The CPU market resembles what it did a couple decades ago so I wouldn't skip on considering either brand right now. I wish the GPU market had the same competition because it's feeling more and more like an nVidia monopoly emerging there. So I'd enjoy it on the CPU side while it lasts.

 

[RJARRRPCGP]

probably the motherboard that died

My nephew had an msi b450 tomahawk, died after 4 years

Mine died in only 2 years of ownership!

 

[DAPUNISHER]

Actually while the thread has taken a slight detour it has allowed me to get a better look at the AMD chips. I started doing more research on CCD/CCX. In the mean time I'm got a docking station which allows me to connect my laptop to my Dell U4919W monitor.
While I'm more of an AMD fan the laptop is an Intel i7-13800H which is 14 cores and 20 threads.
It has 64Gb
And a 2Tb NVMe

So I'm using it as my development system at home for now while I do more research.
This PC lasted 7 years
My prior PC was 10 years
I try to get the mid-range
7950X while this is a nice CPU, too expensive
And the 7900X was 12 cores - 24 threads which I felt was something that could last me 5-7 years again
I also changed the memory to 64Gb, instead of 32
Still flip flopping on whether to do
3 x 1tb
or
1 1tb and 1 2tb

I also recently purchased a NAS storage system, Terra-master F2-212 with 2 8tb drives in RAID
The drives arrived the same day my PC died and the Terra-master unit came the next day.
Luckily I was able to extract the data from my drives since I had an external reader.

It's your money spend it how ever you want. I will give you some free advice though. The problem you described sounds like either bad ram or a bad PSU. If that's all it is, consider the following option - Here in the U.S. a used 5900 can be had for under $300. A drop in upgrade for a 1700X that is a massive performance uplift even on that OG Tomahawk. The VRMs should handle heavy loads if the case cooling/air flow is good. It's the best bang for buck and easiest upgrade.

Just providing a cheap and low effort option to a new fairly expensive build. You have been getting by on a 1700X so time is money wasn't a large enough concern to motivate you as a buyer. /free advice

Oh and if you do discover it was the ram or PSU and decide to do the drop in upgrade. You need to re-paste all of the heatsinks on that B350.

 

[Chrispy_]

As someone who moved (back) to AMD CPUs lately, and then upgraded the CPU, and is planning to most likely have an AMD platform after this, don't lock yourself to AMD. Don't lock yourself to anything if you can help it. There's reasons a brand might be better for you but "never say never".

The reason for Intel's thread count not being double the core count is this. Intel has a hybrid architecture. They have performance cores, or "p cores", and they have efficient cores, or "e cores".

The performance cores are the traditional ones and those have Hyper-threading.

The efficient cores do not have any form of threading.

In the case of the 13700 series CPUs, they have 8 performance cores, each with Hyper-threading. So at its base it's an 8 core/16 thread CPU. It then has a further 8 efficient cores. So, 16 cores/24 threads.

The efficient cores aren't great, but they get way more hate than they deserve. The idea is this; the number of things that need a lot of performance cores are low. That's not to say the things that scale linearly are low, no. But anything that scales linearly is better off with efficiency cores. Why? Simple, you get around four of them in the same space as one performance core would take up, and four efficient cores total more performance than one performance core for that type of workload (for "non linear" threaded stuff they are indeed worse). The real hurdle with them is scheduling issues, not that "they're weak", but AMD's multi-CCD CPUs have "similar" issues of their own.

If you're not planning to do a lot of highly threaded tasks, I'd sort of ignore the efficient cores up front for a moment. You'd basically be looking at the Core i7 and Core i9 as octo cores at their foundation, and that's plenty for any games or the like, but then the efficient cores exist on the side too so they won't be overwhelmed by background tasks either.

Intel's fine right now, and that's coming from someone on AMD and planning to move to AMD next. They have higher power draw, and they're certainly losing to the Ryzen 7 X3Ds for gaming price/performance, but without the v-cache helping, Intel still holds the per core performance edge by a bit. Both are viable right now. The CPU market resembles what it did a couple decades ago so I wouldn't skip on considering either brand right now. I wish the GPU market had the same competition because it's feeling more and more like an nVidia monopoly emerging there. So I'd enjoy it on the CPU side while it lasts.

On Intel LGA1700, the i7-14700 or 14700K is a more compelling workstation option than the 13700/K with four extra efficiency cores over the 13th gen part for the same money - but you do have to accept that it's a dead-end platform that will never have an upgrade path, it's ridiculously power-hungry at full-tilt so you're going to want a $100+ AIO at a bare minimum, you also NEED Windows 11 to get any sensible performance out of it (Win10's outdated scheduler isn't really a viable option) and if you buy a 14700K you also need a Z790 board which is considerably more expensive than a B650 board.

Given those caveats, It's a solid choice, but those caveats aren't trivial. They're a little less of an issue if you go for the non-K variant and B760 board which is similar in price to an AM5 B650 board.

 

[ThrashZone]

So I've never been a big intel fan. I've always built AMD system's except for one time when I built an Intel system and the migrated back to AMD with the release of Ryzen.
My question though when I was looking at the Core i7-13700K
It stated 16 cores/24 Threads.
Where the Ryzen is also X Cores and the threads are 2X
Why are there only 24 threads and not 32 for the Intel?
On the Amazon when you look at the Images it shows
Up to 16 Cores: 8 Performance-cores, 8 Efficient-cores
Up to 24 Threads: 16 on Performance-cores, 8 on Efficient-cores

would like to understand this

Hi,
Better to think of e cores as thermal defects or limited clocking "best on mobile maybe" because they do indeed clock a lot lower than traditional p cores can "which make no real sense on desktop" frankly.

Intel just had enough socket room "cheapest route I guess" without total redesign to add e cores instead of a much larger chip like amd uses plus when amd dropped 5k x3d intel went into panic mode to get more performance because they have been just throwing more power to keep up with amd's performance.
7k x3d took the intel gaming crown away and intel is holding their balls only thing intel has now is people buying what ever they put out.

All amd cores are p cores and clock/ cool equally
I just bought my first amd system and I doubt I'll go back to intel anytime soon.

 

[dgianstefani]

Hi,
Better to think of e cores as thermal defects or limited clocking "best on mobile maybe" because they do indeed clock a lot lower than traditional p cores can "which make no real sense on desktop" frankly.

Intel just had enough socket room "cheapest route I guess" without total redesign to add e cores instead of a much larger chip like amd uses plus when amd dropped 5k x3d intel went into panic mode to get more performance because they have been just throwing more power to keep up with amd's performance.
7k x3d took the intel gaming crown away and intel is holding their balls only thing intel has now is people buying what ever they put out.

All amd cores are p cores and clock/ cool equally
I just bought my first amd system and I doubt I'll go back to intel anytime soon.

Lol, Intel did P+E for performance, where it beats the 7950X in both applications and games, with the X3D being faster in games. E cores have nothing to do with cost cutting and everything to do with more cores per chip, next gen is 8+32. Besides, AMD is doing same cores per die optimization thing with their Zen C, just late, as usual. The flagship chips are both ~250mm dies if you count the IO die on a cheap process.

14900K

7950X

 

[Princess Garnet]

Besides, AMD is doing same cores per die optimization thing with their Zen C, just late, as usual.

I agree with the rest of your message here; that the e-cores have a misunderstood place, but this part seems like a selective take for two reasons.

The first is that AMD and Intel both do different things, and often in response to what the other does, so which is "first" or "second" is often a bit less clear. It seems like a pointless remark to make, but even if you want to, I'm not sure I see all these AMD tardy scenarios in the consumer x86 space anyway? The first to 1 GHz? The first 64-bit? The first multi-core? The first to move the IMC to the CPU? The first non-monolithic? The first stacked cache? The first mixed process node sizes (yes, Intel's going to do that too so the jab at AMD being "cheap" for this is a case of "well, about that..."). Besides the final few, I seem to recall AMD doing all those first and then Intel doing them too. And for the final few, AMD did non-monolithic and stacked cache first and Intel is finally about to do one, and then the other. It's ultimately a silly discussion to have, but even if you want to have it, AMD isn't the constant tardy party anyway.

The second it's a selective comment is because in this particular situation, AMD was "second" insofar as they had no need for lesser cores like Intel does to begin with. Intel was fine selling the same moderately improved quad core for over a dozen generations. They put out a six core, Hyper-threaded product as early as LGA 1366 while most consumers were on LGA 775 on single or dual cores only, yet the consumer landscape only got quad cores all the way through the 7th generation. Ryzen came, and then when it got decent per core performance, Intel was suddenly facing down more scenarios where they were losing (first in multi-threaded, but then the gap was closing in single-threaded too once Zen 2 launched and then Zen3 surpassed them entirely). Intel then had to respond, and this explains the timely increase of the 8th, then 9th, to finally 10th generations in core/thread counts. All in response to Ryzen. Then when Zen 2 doubled core counts again, Intel couldn't match it with its current cores at all because power and thermal demands would be way too high. We even saw the Core i9 lose cores going from 10th to 11th generation. Their answer was e-cores, and yes it's a fine answer, but pretending like this is some "usual Intel first" when it came in response to AMD being ahead in core count and Intel being totally unable to match them is a very peculiar way to look at it.

AMD's ZenC cores isn't quite the same as what Intel is doing anyway. Intel's p-cores and e-cores cores are actually different entirely. AMD's ZenC cores are actually just regular cores, just the opposite direction of what their v-cache approach is. That is, they are still the same as whatever their usual core is architecturally but have less (lower level) cache instead of having more of it. The end goal is the same as what Intel's hybrid approach is after though, yes. The intent with a lesser core that comes with space savings is that for things that don't lose performance from what the core skips out on (in the case of ZenC cores, this is less cache), you "get more performance out of the same potential die space". Because that right there is what it ultimately boils down to now. They're both doing it different ways, and much of what each of them do is in response to what the other is doing, sometimes in non-traditional ways. Seeing that as "AMD is late as usual" isn't even historically accurate.

 

[dgianstefani]

I agree with the rest of your message here; that the e-cores have a misunderstood place, but this part seems like a selective take for two reasons.

The first is that AMD and Intel both do different things, and often in response to what the other does, so which is "first" or "second" is often a bit less clear. It seems like a pointless remark to make, but even if you want to, I'm not sure I see all these AMD tardy scenarios in the consumer x86 space anyway? The first to 1 GHz? The first 64-bit? The first multi-core? The first to move the IMC to the CPU? The first non-monolithic? The first stacked cache? The first mixed process node sizes (yes, Intel's going to do that too so the jab at AMD being "cheap" for this is a case of "well, about that..."). Besides the final few, I seem to recall AMD doing all those first and then Intel doing them too. And for the final few, AMD did non-monolithic and stacked cache first and Intel is finally about to do one, and then the other. It's ultimately a silly discussion to have, but even if you want to have it, AMD isn't the constant tardy party anyway.

The second it's a selective comment is because in this particular situation, AMD was "second" insofar as they had no need for lesser cores like Intel does to begin with. Intel was fine selling the same moderately improved quad core for over a dozen generations. They put out a six core, Hyper-threaded product as early as LGA 1366 while most consumers were on LGA 775 on single or dual cores only, yet the consumer landscape only got quad cores all the way through the 7th generation. Ryzen came, and then when it got decent per core performance, Intel was suddenly facing down more scenarios where they were losing (first in multi-threaded, but then the gap was closing in single-threaded too once Zen 2 launched and then Zen3 surpassed them entirely). Intel then had to respond, and this explains the timely increase of the 8th, then 9th, to finally 10th generations in core/thread counts. All in response to Ryzen. Then when Zen 2 doubled core counts again, Intel couldn't match it with its current cores at all because power and thermal demands would be way too high. We even saw the Core i9 lose cores going from 10th to 11th generation. Their answer was e-cores, and yes it's a fine answer, but pretending like this is some "usual Intel first" when it came in response to AMD being ahead in core count and Intel being totally unable to match them is a very peculiar way to look at it.

AMD's ZenC cores isn't quite the same as what Intel is doing anyway. Intel's p-cores and e-cores cores are actually different entirely. AMD's ZenC cores are actually just regular cores, just the opposite direction of what their v-cache approach is. That is, they are still the same as whatever their usual core is architecturally but have less (lower level) cache instead of having more of it. The end goal is the same as what Intel's hybrid approach is after though, yes. The intent with a lesser core that comes with space savings is that for things that don't lose performance from what the core skips out on (in the case of ZenC cores, this is less cache), you "get more performance out of the same potential die space". Because that right there is what it ultimately boils down to now. They're both doing it different ways, and much of what each of them do is in response to what the other is doing, sometimes in non-traditional ways. Seeing that as "AMD is late as usual" isn't even historically accurate.

5775C was the first large cache CPU.

Non-monolithic is a cost savings for AMD, as is mixed nodes, not a perk for the consumer, latency is worse for both CPUs and GPUs. 14900K is the swansong of monolithic. At least 15th gen Arrow lake is using very advanced tile packaging with all sorts of backside power magic etc., so we can hope the "chiplet" equivalents have better latency.

All those early "eight core" AMD Phenom CPUs etc performed worse than the so called terrible quad core Intels.

Then if you want to you can look at AMD in the GPU section, starting to take RT seriously when NV is on it's third generation. Same with the AI stuff.

What's interesting to me is the pro AMD mindshare of the enthusiast community right now, E/C cores are innovations, not to be complained about, if AMD had released the hybrid architecture though it would be celebrated. Same with RT.

I'm also fairly certain Intel was forced into the adding cores shenanigans for marketing reasons, despite consumer software not really scaling that well beyond 6/8 cores, due to their failures with nodes. If Intel 10/7 was ready earlier we would have seen those higher core count sooner, perhaps before Zen. Remember these architectures and designs are made years before production.

 

[DAPUNISHER]

I am going to be bold and make a suggestion. Let's focus on helping the TS with their build/upgrade instead of engaging in the tired VS trope. Let's also hope they report back with a happy ending. There's nothing worse than when they ghost you, and you are left wondering how things turned out.

 

[Princess Garnet]

5775C was the first large cache CPU.

A perfect example of what I meant by how things can be traced to either being the "first" depending on definition.

My mistake though, I meant the first 3D vertically stacked cache. I figured that would have been clear.

Even if we mean "large cache", the 5775C wasn't the first. In its time, the Pentium Pro had a lot. Or even the Pentium 4 being the first (I think?) with L3.

Non-monolithic is a cost savings for AMD, as is mixed nodes, not a perk for the consumer, latency is worse for both CPUs and GPUs. 14900K is the swansong of monolithic. At least 15th gen Arrow lake is using very advanced tile packaging with all sorts of backside power magic etc., so we can hope the "chiplet" equivalents have better latency.

Yes, there's pros and cons here. I won't dispute that, but I thought the discussion was about who does something first that the other then follows, and moving away from monolithic is in that category, no?

All those early "eight core" AMD Phenom CPUs etc performed worse than the so called terrible quad core Intels.

Agreed (I choose a 2500K back then and AMD almost wasn't even on my radar), but I didn't say AMD had the first octo core. My comment with Ryzen was how they finally got their act together and it freed us from Intel pushing the same iterative quad core as our best choice (outside HEDT) for so long.

Then if you want to you can look at AMD in the GPU section, starting to take RT seriously when NV is on it's third generation. Same with the AI stuff.

Again, agreed. nVidia is perhaps lightyears ahead on ray tracing, but since the discussion was regarding Intel and CPUs, I wasn't talking about nVidia or GPU here. The Radeon division is definitely different than the Ryzen one.

What's interesting to me is the pro AMD mindshare of the enthusiast community right now, E/C cores are innovations, not to be complained about, if AMD had released the hybrid architecture though it would be celebrated. Same with RT.

Once again, agreed, and this was the initial thing I agreed with you on.

The e-cores are definitely misunderstood, and from a "performance per die space" stance, they are better (at least than Intel's own p-cores) for linearly scaling workloads. They justify themselves there.

But they do seem to have scheduling concerns (I haven't used a 12th generation+ processor for clarity, so I admit I'm going on what I've read of with things), and you seemed to have been critical of latency and scheduling concerns yourself.

 

[LFaWolf]

This thread is a bit old, but I don't know if OP has made a purchase. I am a developer, and we use the same apps as you do. I recently got a 7900x for a developer because it is so cheap. It is like the middle child that has been ignored. We also have a 14900K that I plan to use for myself. Here is what I found -

1. VS is multithreaded. You can specify the specific number of concurrent builds. I don't know if this makes a difference for your project since I don't know what kind of projects you run, but it allows our asp.net core projects to utilize all of my threads (i9-10940x). Most people leave it alone and it still uses all available cores.

2. Webpack is multithreaded. If you use it, you can specify the number of parallelism, but I think the default is 100. It will use all your available cores.
3. There is no noticeable penalty for AMD chiplet or CCX design when compiling code. Code compiling using VS and Webpack scales well to the number of cores from 7900x to 14900k.
4. Therefore, if you have a lot of test cases in your projects, the more cores you have, the faster they will build.
5. I run 3x 2TB for my system, C: for all the programs, D: for projects and scripts, E: for data.

Let me know if you have any questions.