AMD Ryzen 7000 non-X Processor SKUs Confirmed with 65W TDP, Boxed Coolers

[tussinman]

While I agree that DDR5 usually shows benefits in cases where the frame rates are already high, the pricing disparity between DDR5 and DDR4 isn't that bad now. Techspot used DDR5 6000 CL36 in their comparison with DDR4 last year. In the USA, It's around $180 on Newegg and DDR4 3600 CL16 is about $153 there. I think that price difference isn't significant enough to choose the slower DDR4.

Scenario he's talking about (non K CPUs with mid level b boards) calls for ddr4 3200 which is half the price of the ddr5 you listed and 1/3rd of the price of the next tier higher ddr5 which is what a lot of reviewers used

3600 isn't guaranteed on non K CPUs (mine won't even boot 3600 on a tier higher H series board) and it's not guaranteed on B boards which is why the reviewers that did use 3600 almost always used a z board

 

[n-ster]

Sexy 7900 price, hopefully that can also translate to the X3D variant being well priced in their line-up as well.

 

[Space Lynx]

Sexy 7900 price, hopefully that can also translate to the X3D variant being well priced in their line-up as well.

the 7900x is already $440 on a regular basis... not sure why you would be excited about 10 dollars less for slightly less performance

 

[Mussels]

His post lacks a lot of context even more not knowing the full system specs.

I mean my 4th gen laptop boots instantly to windows and feels super duper snappy, but that's because its got no extra hardware to boot from and runs at 768p

You obviously have'nt seen a business from inside, did you? R&D is the first thing thats taxed upon sales, meaning R&D has to be paid back first before they even think of selling it for cheaper. Always has bin like that. 3 years of engineering usually goes into chips.

Or more, but yes - they always start with high prices and drop them once they pass the thresholds they need to profit from the endeavor

Often times, they release new products to shake up the prices instead of dropping prices of existing products again and again

 

[Minus Infinity]

the 7900x is already $440 on a regular basis... not sure why you would be excited about 10 dollars less for slightly less performance

Wow, just checked Aussie prices and yes it's dropped a lot, can now find a few stores selling at $699 down from $1K. As for x3d variant, needs to be $499-519 to compete against 13700K at least if your sole emphasis isn't gaming, which it shouldn't be with this class cpu IMO.

 

[chrcoluk]

You obviously have'nt seen a business from inside, did you? R&D is the first thing thats taxed upon sales, meaning R&D has to be paid back first before they even think of selling it for cheaper. Always has bin like that. 3 years of engineering usually goes into chips.

Thats seems to be shorter than the gaps between generations, so make each generation last longer then we get cheaper chips?

What about motherboards though?

 

[Space Lynx]

Wow, just checked Aussie prices and yes it's dropped a lot, can now find a few stores selling at $699 down from $1K. As for x3d variant, needs to be $499-519 to compete against 13700K at least if your sole emphasis isn't gaming, which it shouldn't be with this class cpu IMO.

my guess is x3d is going to pricey this round, AMD knows its the only Ace they have left.

 

[Minus Infinity]

my guess is x3d is going to pricey this round, AMD knows its the only Ace they have left.

If it boosts a lot of non-gaming software too, it could be worth it. If it's just for gaming, I don't even see the need for 79xx class variants to have v-cache. Zen 4 is already very good at gaming, but gets easily beaten in a large % of productivity applications. Milan X does show some big improvements in a few things I find relevant like CFD (eg OpenFOAM) and if it helped in things like COMSOL, Matlab, Anysys to name a few, I'd consider it. But RL holds large leads in a lot of that software over Zen 4, eg COMSOL sims were well over 50% faster in RL than Zen 4. If v-cache even gets parity, it's still a lot more money for basically better gaming performance than RL. I'd save the money for the MB and memory.

 

[ZoneDymo]

I only used crystaldiskmark, but I can also tell an overall "snappiness" feeling when using Intel raptor lake in day to day usage, even just using file explorer. It's def not placebo.

And I don't notice any difference going from a 2600k to a 12600k in overall "snappiness"

 

[Space Lynx]

And I don't notice any difference going from a 2600k to a 12600k in overall "snappiness"

you need to have a 176 layer nvme drive to know what I am talking about, and the processor powerful enough to feed it.

If it boosts a lot of non-gaming software too, it could be worth it. If it's just for gaming, I don't even see the need for 79xx class variants to have v-cache. Zen 4 is already very good at gaming, but gets easily beaten in a large % of productivity applications. Milan X does show some big improvements in a few things I find relevant like CFD (eg OpenFOAM) and if it helped in things like COMSOL, Matlab, Anysys to name a few, I'd consider it. But RL holds large leads in a lot of that software over Zen 4, eg COMSOL sims were well over 50% faster in RL than Zen 4. If v-cache even gets parity, it's still a lot more money for basically better gaming performance than RL. I'd save the money for the MB and memory.

true, even a ryzen 5600 with slower ddr4 ram is only like 20-30 fps slower at 1440p, and in some games not even 5-10 fps. cpu and ram really don't matter anymore. just need a good gpu, 6800 xt or higher at 1440p and mostly will be good to go

 

[Chrispy_]

you need to have a 176 layer nvme drive to know what I am talking about, and the processor powerful enough to feed it.

Haven't SSD's basically flatlined in terms of access latency and random IOPS?
The results here are basically a wash with high-end NVMe and entry-level TLC drives jostling for position. Clearly, it's not the latency or access times of modern SSDs that are holding back OS "snappiness"

IMO it's mostly down to your OS configuration, what system tray apps are running, how many services and realtime scanners you have for disk and network, and how much crap is going on. I can make VMs at work that are effectively god-tier storage and specification and they can run amazingly or run like crap depending on location and config.

Some of the worst offenders for Windows are your ISP's latency, DNS service speed, and hardware firewall (if you use one). All of those overheads have the capability to add half a second of wait all over the place in Windows for asinine reasons. Turning off most of the Microsoft telemetry/lookup/help/search features that head out to the web for requests have a massive impact on snappiness - far more effect than any hardware advances in the last decade can achieve! (alongside those god-tier VMs on six-digit enterprise flash arrays and servers, I daily an i7-2600(non-K) with 24GB of mismatched DDR3 and a 180GB Intel SATA drive. By simply locking down services/lookups and bypassing the hardware firewall during testing, I can make this PC feel better in Windows than anything new with a clean, unmodified version of Windows 10/11 connected to the internet)

If I was spitballing an estimate, I would say that hardware performance of any pseudo-modern PC is a negligible sub-20% of the total time taken from you interacting with the OS and getting a result on screen. The overwhelming 80% of the short delay is web-based or optional OS service features like smartscreen, Cortana, AV heuristics, etc.

 

[Space Lynx]

Maybe it is placebo, but I feel like on raptor lake with a 176 layer nvme drive, everything just felt a smidge faster, even just switching between file explorer/web browser, etc.

@W1zzard being the nvme reviewer, what has your real world experience been regarding this? is it just placebo, or have you noticed that 176 layer gen4 drives do indeed feel a bit "snappier" again this is a generic term, as its not about statistics or numbers, just real world "feeling" when using.

 

[Chrispy_]

You're hyperfixating on a meaningless stat; The number of layers of the NAND on the drive has NOTHING to do with its performance.

At best, once the cache on the drive is exhausted from huge sustained sequential writes, the raw write speed of 176 layer NAND might actually mean something. In reality, you'll never get to that point, unless you're intentionally testing for it

Realistically, the SSD controller and its firmware have the largest impact on how fast a drive is. If you end up dealing with the raw NAND speeds of individual NAND packages, then the drive is probably garbage and the controller/firmware have completely failed to do their job.

I'm going one step further than and pointing out that it doesn't even matter how fast your drive is. The variables that affect "snappiness" and "responsiveness" are mostly independent of storage speed these days. It's been that way for around 15 years once SATA drives matured enough to saturate their interface and shift the bottleneck in most OS workloads over to something else.

So:

• The number of layers on the NAND is irrelevant, because the number of NAND packages and channels matters far more.
• The number of NAND packages and channels is irrelevant, because the performance of the controller and firmware's cache capabilities matter far more.
• The performance of the controller/firmware are negligible, provided that the access latency isn't problematic and that numbers lie within the class-leading ranges I posted above

Provided the drive isn't a turd, and you're not continuously moving tens of GB of data into and out of RAM, then the drive performance is only a minor, almost negligible factor in how responsive and snappy an OS feels.

 

[Space Lynx]

You're hyperfixating on a meaningless stat; The number of layers of the NAND on the drive has NOTHING to do with its performance.

At best, once the cache on the drive is exhausted from huge sustained sequential writes, the raw write speed of 176 layer NAND might actually mean something. In reality, you'll never get to that point, unless you're intentionally testing for it

Realistically, the SSD controller and its firmware have the largest impact on how fast a drive is. If you end up dealing with the raw NAND speeds of individual NAND packages, then the drive is probably garbage and the controller/firmware have completely failed to do their job.

I'm going one step further than and pointing out that it doesn't even matter how fast your drive is. The variables that affect "snappiness" and "responsiveness" are mostly independent of storage speed these days. It's been that way for around 15 years once SATA drives matured enough to saturate their interface and shift the bottleneck in most OS workloads over to something else.

So:

• The number of layers on the NAND is irrelevant, because the number of NAND packages and channels matters far more.
• The number of NAND packages and channels is irrelevant, because the performance of the controller and firmware's cache capabilities matter far more.
• The performance of the controller/firmware are negligible, provided that the access latency isn't problematic and that numbers lie within the class-leading ranges I posted above

Provided the drive isn't a turd, and you're not continuously moving tens of GB of data into and out of RAM, then the drive performance is only a minor, almost negligible factor in how responsive and snappy an OS feels.

with 176 layers we saw faster 4k reads though? that is the most important stat I thought?

 

[AnotherReader]

with 176 layers we saw faster 4k reads though? that is the most important stat I thought?

Faster 4k random reads at a queue depth of 1 haven't happened in a long time because NAND flash is slow for random reads.

 

[Psychoholic]

These should do pretty well.
I run my 7950X at ECO/105W mode anyway since pretty much day 1.

 

[ARF]

These should do pretty well.
I run my 7950X at ECO/105W mode anyway since pretty much day 1.

A normal company would have launched them from the get-go on day 0.
But hey no, let's put our imaginary performance dominance over the ugly dark-shaded Intel and its bad practices... Meh!

 

[Chrispy_]

with 176 layers we saw faster 4k reads though? that is the most important stat I thought?

Coincidence, not correlation.
The newest controllers with the highest performance use the newest NAND with the most layers. Those new controllers would also do well with other NAND, provided it's good enough to keep up.

The number of layers in NAND is basically "what year was it made". The improved performance has nothing to do with layer count, but it's cost-effective to keep increasing layers with each new, faster, better NAND generation, so the number keeps going up.

NAND with more layers is typically better, but it's like looking at clockspeed of a CPU in isolation without taking into account how many cores there are, what architecture it's using, or what the IPC is. You wouldn't want to use a hypothetical 5GHz Intel 486 instead of a 4.8GHz i9-12900K, would you?

A normal company would have launched them from the get-go on day 0.
But hey no, let's put our imaginary performance dominance over the ugly dark-shaded Intel and its bad practices... Meh!

What normal company?
This behaviour of top-down launches has been the status quo for Intel/AMD/Nvidia/Apple/Samsung for the last decade or more.

 

[ARF]

What normal company?
This behaviour of top-down launches has been the status quo for Intel/AMD/Nvidia/Apple/Samsung for the last decade or more.

I don't think a 170W Ryzen 9 7950X and 7900X followed by 65W Ryzen 9 7900 is a top-down approach.
Actually, they are fixing the terrible mistake with the uber high and unsustainable by all normal measures TDP of the former launches.

 

[Minus Infinity]

Coincidence, not correlation.
The newest controllers with the highest performance use the newest NAND with the most layers. Those new controllers would also do well with other NAND, provided it's good enough to keep up.

The number of layers in NAND is basically "what year was it made". The improved performance has nothing to do with layer count, but it's cost-effective to keep increasing layers with each new, faster, better NAND generation, so the number keeps going up.

NAND with more layers is typically better, but it's like looking at clockspeed of a CPU in isolation without taking into account how many cores there are, what architecture it's using, or what the IPC is. You wouldn't want to use a hypothetical 5GHz Intel 486 instead of a 4.8GHz i9-12900K, would you?


What normal company?
This behaviour of top-down launches has been the status quo for Intel/AMD/Nvidia/Apple/Samsung for the last decade or more.

Didn't Samsung specifically mention next gen 230+ layers ssd's would have improved random IOPS as well as sequential. I hope it's toke a token 5% but something at least of order 20%. Only reason I would even bother with a gen 5 ssd. Sequential is marketing BS.

 

[tfdsaf]

You obviously have'nt seen a business from inside, did you? R&D is the first thing thats taxed upon sales, meaning R&D has to be paid back first before they even think of selling it for cheaper. Always has bin like that. 3 years of engineering usually goes into chips.

I've owned my own business, but what has that got to do with AMD's pricing? I don't care how much it costs them, Intel has a far superior product and if they want to compete, they need to lower prices, simple as that.

The 13600k destroys the 7600x and 7700x, all for $330. In games its reasonably similar in performance but destroys the 7600x and beats the 7700x a fair bit in multithreaded apps.

 

[pressing on]

I've owned my own business, but what has that got to do with AMD's pricing? I don't care how much it costs them, Intel has a far superior product and if they want to compete, they need to lower prices, simple as that.

The 13600k destroys the 7600x and 7700x, all for $330. In games its reasonably similar in performance but destroys the 7600x and beats the 7700x a fair bit in multithreaded apps.

But we are talking about CPUs like the AMD 7600 which should be priced less than $250. AMD will have a chance against the Intel i5 13400/F, 13500 and 13600 because they are essentially variants of the 12th gen Alder Lake 12600K (locked frequency, lower base clocks, lower maximum turbo except the 13600). They won't have the performance jump of the true 13th gen Raptor Lakes, although the 13500 and 13600 may have more L3 cache (and Efficient cores) than the 12600K which will help. It will probably all come down to pricing, and it's no surprise that some people are expecting a price war in this segment of the market to break out in January. There are rumours that Intel will price these i5s somewhat higher than the 12th gen chips, such as the 13400F at $210 and the 13400 at $240. This could price the 13500 at $260 and the 13600 at $280.

The far away and best seller of this range of Intel i5s to date has been the 12400F and even at the increased price the 13400F could be expected to follow suit. So there is an opportunity for AMD but competing with a chip like the 13400F is going to be tough.

 

[AnotherReader]

The far away and best seller of this range of Intel i5s to date has been the 12400F and even at the increased price the 13400F could be expected to follow suit. So there is an opportunity for AMD but competing with a chip like the 13400F is going to be tough.

Given that the 7600X is equivalent to the 12600k, I don't see how the 13400F would be a competitor. Right now, the 7600X can be found around the $240 mark. The real problem for AMD is the high price of AM5 motherboards.

 

[tfdsaf]

Given that the 7600X is equivalent to the 12600k, I don't see how the 13400F would be a competitor. Right now, the 7600X can be found around the $240 mark. The real problem for AMD is the high price of AM5 motherboards.

Only in games, but it loses BADLY in apps. Again, we are talking up to 40% difference in multithreaded apps, while still costing $300, $270 with some promotions and price cuts in certain markets!

The 7600x needs to be $230 at most with the 7600 $200 at most. Even the 7700x loses in apps to the 13600k, around 10% slower, but its $10-20 more expensive!

AMD are in a terrible position right now, with new mobo's being extremely expensive, even the cheapest garbage ones selling for $180+ and their cheapest CPU costing at the least $260 in specific markets and under specific promotional price cuts. Officially AMD has not lowered the MSRP of these CPU's, so they will be selling for $300 in many markets, including Eastern Asia, South America, parts of Europe, Australia and New Zeland, etc...