Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
AMD 7000 Series - A disappointment!
- Thread starter exoticSpice
- Start date
- Sort by reaction score
The performance/watt metric is amazing. IPC is acceptable. ST increase acceptable.
It should. But what are you going to use to power everything other than the RTX4090?
How should we consider the M2 if AMD's new CPU is considered a disappointment because its theoretical improvement in ST is only 15% (vs. 12% for the M2)?
However you want really. The thing that really plagued Zen 4 reactions were all the dumb rumours claiming they'd have like 20% IPC gain. The M2 was mostly expected to be an A15 part (of which there weren't lofty rumours re: performance) so the comparatively small IPC improvement is more manageable.
Also, M2 has large GPU, E-core, and NPU improvements so that helps compared to just the P-cores.
7950X yes 16 Core however I’ve decided to go With an RTX 3080 Ti should tide me over. Got a crazy deal at $600 bucks. Just my 8700K could bottleneck me but not so much at 1440p.
I have to say I was quite impressed by the new chips in particular the 7950X. Most surprising was not just the performance improvement over Alder Lake i9-12900K but the power efficiency from it as well (over 40%). That and the fact they also added AVX-512 functionality onto the chip makes it a sweet deal.
I am looking forward to seeing this being benched and used as a hackintosh. It should potentially be a nice speedy system.
I am looking forward to seeing this being benched and used as a hackintosh. It should potentially be a nice speedy system.
One thing that tech observers probably haven't noticed is that the new chips have much better power consumption.
This is where and how the increase in IPC and 42% system performance vs Zen 3/62% performance per watt (7950X) has largely come from (aside from the overall shift down to 5nm).
In the previous gen Ryzen 5000, the power drawn by the chip on various I/O subsystems at idle/load was a lot higher compared to Intel offerings. In the new chip, AMD has significantly worked on the I/O die (now at 6nm) bringing power consumption down to less than 20% of what it was before.
To confirm this, Hattedsquirrel provides an in-depth look on the power usage of the Ryzen 5000 series which shows in its tests that the 5000 series chip has higher than normal power usage compared to Intel, and that due to technical issues in how the chip works, these figures are not reported by software to the system vs hardware measurements taken from the chip, meaning the results of power draw on Ryzen 5000 chips are potentially much higher.
This is where and how the increase in IPC and 42% system performance vs Zen 3/62% performance per watt (7950X) has largely come from (aside from the overall shift down to 5nm).
In the previous gen Ryzen 5000, the power drawn by the chip on various I/O subsystems at idle/load was a lot higher compared to Intel offerings. In the new chip, AMD has significantly worked on the I/O die (now at 6nm) bringing power consumption down to less than 20% of what it was before.
To confirm this, Hattedsquirrel provides an in-depth look on the power usage of the Ryzen 5000 series which shows in its tests that the 5000 series chip has higher than normal power usage compared to Intel, and that due to technical issues in how the chip works, these figures are not reported by software to the system vs hardware measurements taken from the chip, meaning the results of power draw on Ryzen 5000 chips are potentially much higher.
Power consumption of Ryzen 5000 series CPUs
The power consumption of a Ryzen 5000 processor is significantly higher than what the software reports. Here is why.
hattedsquirrel.net
If you want to look at the effect of 12-nm I/O die, compare 5600G and 5600X, 5700G (not the iGPU one) and 5800X in your first graph. I assume HotHardware had done sufficient due diligence for the numbers to be comparable. We could see at idle the difference is about 6-8W. That's pretty much the extra power 'hog' contributed by 12-nm I/O die vs a possible hypothetical 7-nm I/O die in Zen2 & Zen3.
The I/O die is a power hog is an urban myth in Zen2 & Zen3, at least for consumer Ryzen systems.
Do note that these tests from HotHardware were most likely done at stock settings. Once you overclock the memory and/or CPU, the I/O die will consume more power than rated for. In part that was deliberate act by AMD's firmware and/or users in an attempt to address stability in an overclocked state of the systems.
EDIT:
Forgot to say the obvious! I do agree 6nm I/O die will be much better. I would speculate the package power of Zen4 Ryzen at idle would be as good as (if not better than) Zen2/Zen3 APUs which is <10W.
The new chiplet-based AM5 Ryzen also come with 2-CU RDNA2 iGPU. That's great news to use these processors for building home labs, VMs, SOHO servers & etc without waiting for the APUs. More than enough PCIe slots/sockets and bandwidth for expansion in such use cases. Will definitely retain some of the users from the necessity of moving to EPYC or Threadripper Pro.
Last edited:
Aren't those all 1 CCD systems? Seems like the IO Die would draw more power on multi-CCD chips, or am I wrong about that? My 5900X, last time I checked seemed to report a roughly 20W difference between Core and Core+SOC when under load.
I prefer to look at "CPU package power" or PPT. Single-CCD and two-CCD Ryzen are about 11W and 15W at idle with stock settings.
I always suspect Psoc (from SVI2) doesn't cover the full picture of uncore. AMD officially in their Ryzen Master likes to present you with Pcore, Psoc individually (together with % of PPT though). Not so sure about the rationale behind it..
Sales are looking very bad for AMD. Everyone was complaining about the high cost of adoption, i.e., DDR5, new motherboard - that's on top of the pricey CPUs. Now it seems consumers/hobbyists are not adopting it as much as AMD was expecting it seems
AMD Ryzen 7000 Sales Plummet by >70% in the Second Week
AMD Ryzen 7000 Sales Plummet by >70% in the Second Week
Yup. AMD made a mistake not launching the B-Series boards at the same time for the more budget conscious folk. I am not sure how useful it would have been for AMD to make a LGA AM4 board, and it feels like it would have muddied the waters for AM5 (where the number historically has indicated what version of DDR the board supported).
Intel does have the advantage here as they typically only support a socket for 2 years so for 14th Gen you are going to have to get a new board anyways.
AMD aimed their CPUs prices too high given all the new tech people must adopt. Hopefully they don't fall victims to the same trappings Intel did back after the 3rd Gen Core ix.
The issue is the price of entry, DDR5, high end motherboards, and yes high priced CPUs, new coolers are needed as well. At least with Intel, the we can use our existing DDR4 and there's cheaper motherboards.
I think we have to factor in the fact that people have less disposable income as well, with inflation running rampant around the world, many people are making hard choices on what they want to spend their money on, and new PCs are just a luxury.
Intel 13th gen CPUs max out at 332watts and 85C on Air Cooler. Good temps for that power and get 40K in Cinebench.
250watts for 38k in Cinebench and around 60C. Better than Zen4 where it's always 95C under load in Cinebench and get around 38K.
Zen 4 is a not great and also very expensive.
250watts for 38k in Cinebench and around 60C. Better than Zen4 where it's always 95C under load in Cinebench and get around 38K.
Zen 4 is a not great and also very expensive.
Should we be comparing temps between the two now? AMD and Intel are now running two different boost algorithms.
3d cache is coming to the 7000 series but I don't think for any AM4 socketed CPUs. I agree legacy users would benefit the most but I see AMD pushing their user base to AM5 rather giving them more years on AM4 - just my $.02
In the past AMD offered transistion CPU's (AKA a CPU that supports both old and new memory type, like Intel has done). This seems to be the first time they haven't done that (well in recent history). I think it would be harder to do it due to the socket type change. They cannot make a board/chip that supports both LGA and PGA.
AMD Ryzen 7000X3D 3D V-Cache CPUs Rumored To Be Unveiled At CES In 16, 12 & 8 Core Flavors
While many people were expecting the 3d cache variant, I think the speed to which they'll be releasing them (sooner then later) took some by surprise. Some people think these chips could be announced by late January
While many people were expecting the 3d cache variant, I think the speed to which they'll be releasing them (sooner then later) took some by surprise. Some people think these chips could be announced by late January
Note this is only the reference cards having this issue. And even then it is unknown how many are affected. The custom cards (and it isn't clear if the plain XT is also affected) are fine.
No question, but in Debauer's testing there is a custom card and its running hotter (though its because of the increase power draw).