AMD Ryzen 9 7950X

[tfdsaf]

As I kept saying and people like Ratirl kept laughing, the 7950x is less efficient than the 5950x. Probably the same will apply to the whole zen 4 range. Yikes

How can it be less efficient if its running at 25% higher performance? Your take is nonsensical!

I think this is a great showing by AMD, certainly blows everything right now out of the water, though I'm hoping for a price war with Intel once their 13 series processors are released. I think this and all new AMD processors can be 20% cheaper!

 

[Crackong]

AMD has an Eco mode on these, and I wouldn’t mind seeing extensive testing of these chips under modest coolers and 65W and 100W modes. I bet you’d still get really good everyday performance out of it. This is a 16C CPU mind you, we do have to contend with physics eventually. If you want crazy levels of sustained multithreaded performance, something has to give.
Put the 7950 against the 5950 at the same TDP and see what you get—I bet you’ll still get the rated boost clocks across some cores, they just won’t sustain near as long.

Even with Adler lake, those E cores can only buy you so much efficiency. When the workload hits, total energy consumed is going to be measured by the sustained load X time to complete. Pure max power draw is only half the story.

It is pretty obvious that the eco mode setting should be the original 'normal' setting
And the current setting should be the counter measure of Intel MCE behaviour, which unlocks all potential of the CPU, in the cost of heat and power consumption

 

[Aranarth]

How can it be less efficient if its running at 25% higher performance? Your take is nonsensical!

I think this is a great showing by AMD, certainly blows everything right now out of the water, though I'm hoping for a price war with Intel once their 13 series processors are released. I think this and all new AMD processors can be 20% cheaper!

Tomshardware has a nice graph to show efficiency improvments..

If you have been doing deep dives into cpu architecture for 20 years like I have then this stuff pretty simple. (I got into the more geeky side of this when amd released the k6-iii)
If you just started looking at this it is kind of heady and confusing to get your head around some of it.

 

[Solid State Brain]

Computerbase (de) made extensive tests with the CPU at various power levels:

AMD Ryzen 7000 im Test: So schnell sind 7950X und 7700X: Leistungsaufnahme und Effizienz

Ryzen 9 7950X & Ryzen 7 7700X im Test: Leistungsaufnahme und Effizienz / TDP und PPT steigen deutlich

www.computerbase.de

Google Translation:

 

[Adonisds]

https://twitter.com/x/status/1574398461642174464

Thank you very much for testing emulator performance. Did you enable AVX512 mode before testing RPCS3?

 

[zlobby]

While I don't disagree with you, what he stated is not without merit. From where I sit Zen 4 just looks like Zen 3 with DDR5 and a really nice overclock. I wager if AMD released a DDR4 capable Zen 4 there would be little thay sets them apart aside from the clockspeed adv. That's probably the reason why Zen 4 is DDR5 only. Sell us that and a lot less people would be unhappy.

Now that's something I'm willing to work with.

Apart from being 'Zen 3 unleashed' it's an entirely new uArch with lots of tweaks and reworks under the hood. Take the memory bandwidth from DDR5 for example. Or the fact it can boost automatically to insane clocks.
I haven't checked Zen 4 in great detail yet but I'm sure other things are also improved. Whether it's worth the ask price - that's for anyone to decide for themselves.

BTW, anyone got nitty-gritty numbers of memory latencies and in-depth memory testing?

 

[Easo]

At the end of the day, you should look at multiple reviews. As I said in my conclusion, gaming performance is fantastic onall these CPUs and you'll never notice a difference without FPS counter

That's the thing most people here forget - average person will see bajillion frames anyway on pretty much all modern CPU's and that is assuming they even know what FPS counter is or how to enable it. Enthusiasts are few but think everyone views world through their eyes where 1-2% difference is horror/amazing/fail/win, when it really really does not matter since game already pushes 150+ FPS (I am not counting CS:GO fanatics and the like).
GPU will make or break it for most people, not CPU.

 

[onemanhitsquad]

JaysTwoCents has the scoop...

 

[Garrus]

Well, the trash power draw is no surprise after AMD said these will be 170W TDP. Wonder what the people who keep shitting on Intel over this will say now, lol.

Looks like I'll be hunting for a 5950X for an upgrade as Zen4 is is just not good enough in a lot of ways, especially for the entry price, or maybe I get a 13xxx as those will at least let me keep my RAM.

Oh dear people are so confused about this stuff. The new silicon has better performance, and uses a more efficient 5nm node. YOU decide the settings. In no way is it less efficient. AMD showed at a fixed 65W the 7950X is 74 percent more efficient than the 5950X. Seems like you are about to make a huge mistake.

 

[erek]

JaysTwoCents has the scoop...

opinion?

basically 0% gain from 5950X to 7950X @ 4K imho

 

[Blueberries]

As someone has already mentioned, that is not how it works. It makes absolutely no difference what the temperature of the CPU is. Whether it is 95* or 50*, the heat output is the same, it is all about how that heat is transferred.

The dies are so small and the IHS is so thick, that the heat cannot be transferred fast enough to the cooler.

If the heat can be transferred quickly, the CPU will be cooler, but other components like the radiator or the water block will be hotter. But it is the same amount of heat.

What happens when you spin up your GPU fans to maximum? The GPU temperature will drop by 10 or 20 degrees, but all that heat will still be inside the case and you need to get rid of it using exhaust fans. And once you exhaust it, all that heat will be in your room.

Maybe I'm just stupid but this makes no sense to me.

Using an identical cooler:
CPU A has a load temperature of 90C
CPU B has a load temperature of 60C

You're telling me the air surrounding the heatsink will be the same temperature? How can this be?

 

[Zubasa]

Maybe I'm just stupid but this makes no sense to me.

Using an identical cooler:
CPU A has a load temperature of 90C
CPU B has a load temperature of 60C

You're telling me the air surrounding the heatsink will be the same temperature? How can this be?

Quite simple, one cpu has higher thermal resistance than the other.
Heat is getting out of CPU B faster than CPU A until it reaches equilibrium.
The temperature of the room it is in will be the same if the power thus energy dissipated is the same.
Also you are assuming the heatsink fins will be the same temp as the cpu, it is not.

For example when I had my 1950X OCed pulling 380W in Prime95 it reached 80C,
but my water temp was hotter than my 12700K unlimited @280W under Prime95 despite the CPU reaching 108C before I got my contact frame.
After I switch to the contact frame the CPU temp went down but the water temp is still the same as before.

 

[phanbuey]

Maybe I'm just stupid but this makes no sense to me.

Using an identical cooler:
CPU A has a load temperature of 90C
CPU B has a load temperature of 60C

You're telling me the air surrounding the heatsink will be the same temperature? How can this be?

If I put a cheap 120MMAIO on my GPU, it will only get to about 65-70C, which is quite low for a GPU but the radiator will burn you to touch it and the tubes will be melting.

Heat output in wattage is what matters not the diode temperature.

 

[oxrufiioxo]

If I put a cheap 120MMAIO on my GPU, it will only get to about 65-70C, which is quite low for a GPU but the radiator will burn you to touch it and the tubes will be melting.

Heat output in wattage is what matters not the diode temperature.

I don't understand why people don't get this.... My 5950X runs around 68C while fully stressed but puts out very little heat, My 3080ti on the other hand running around 65C is dumping a crap ton of heat into my case/room because it's consuming more than double the wattage even though it is running relatively cool.

 

[nguyen]

Man the long boot time is certainly a pain to deal with on a high-end build, AMD will be fixing this issue but I wonder how long will it take them doing so.

Now if Intel 13th + DDR4 can beat Ryzen 7000 + DDR5, AMD will have a hard time selling Ryzen 7000...

 

[MarsM4N]

7950X = best productivity CPU. Basically "raping" the Intels. Will be hard for them to catch up.
Thermals are kinda toasty, but the smaller process the hotter it get's. Guess we need to get used to it.

P.S.: for the kids out there, this is not a "Gaming" CPU. Save your cash and go for the Ryzen 5 (or 7 if you need something between). For gaming I would wait for the X3D variants.

JaysTwoCents has the scoop...

"The Ryzen 7000 system felt like a faster Intel system than a Intel system." Jayz2Cents Needs confirmation @W1zzard

Boot delay, because of RAM initialization, now that's something that could drive folks nuts who prefer to unplug their systems from the grid (f.e. with a smart socket).
Hope there is a way to set the "found optimum" RAM speeds fixed in the BIOS to skip this nonsense.

"95c by design, NO thermal throttling." That's something that should be thoroughly tested. Does performance suffer from bad coolers?

 

[A Computer Guy]

That is not true. It will reduce the clock speeds to stay at or below 95°C. In my definition that's "throttling", even if it's very mild (when you have decent cooling)

What happens to AIO's over extended period of time when the CPU is pumping out 95°C? Can they even keep up? Will they get overwhelmed and leak or explode?

Okay I need to know, is it even safe to run 95C for extended durations? I understand that AMD has rated their chips this high but everything else in your chassis will also be cooking unless you have some really exotic airflow.

Those are BBQ temperatures. You can cook ribs in there.

Sounds like a bad day for fans on air coolers.

 

[Ruru]

I'm actually surprised that the IGP's 128 shaders are even somewhat capable. Though still GT 1030 goes way ahead. But usable for older games.

 

[pantherx12]

Not sure if it's been said but high temperatures are because the IHS is stupidly thick to maintain cooler compatibility.

Bad move on AMD in my opinion. Could of included a shim for people that want to use old coolers, or different size spacers or something along those lines.

This will be a CPU that will benefit from.lapping and delidding.

 

[GeorgeMan]

I've seen this being repeated around, but on my 12700K I don't observe frequency dropping below spec when the CPU reaches the thermal throttling point. The CPU gradually decreases frequency (and voltages) and then oscillates quickly by a few hundred MHz to mantain temperatures at the configured level, like AMD is supposed to be doing here.

The Intel datasheets suggest that in some cases the processor duty cycle can be decreased to 1/4 of the regular rate (25% on, 75% off), if voltage and frequency changes do not manage to decrease temperatures, but this does not normally happen.

You're right, but the frequency/time plot in some reviews didn't reveal any fluctuation, just the normal 50-150MHz frequency drop (due to thermal saturation and power limits) in all core loads. Yet even the 7600X reached >90°C in many reviews... It's not that hard to cool ~100-150W CPUs with a 360 hydro cooler after all... I know the smaller manufacture nodes don't help with heat but it can't be so much worse than the previous iteration (5000 series).

THANK YOU for adding emulators. they always get overlooked in benchmarks. hopefully they become common with more games and emulators tested across multiple resolutions. like the normal gaming benchmarks you guys do here. an endless scrolling down of just emulation benchmarks.

Completely agree with this comment! And please know that RPCS3 doesn't have AVX-512 enabled by default, if it gets enabled it should blow away the older models.

 

[Ruru]

Not sure if it's been said but high temperatures are because the IHS is stupidly thick to maintain cooler compatibility.

Bad move on AMD in my opinion. Could of included a shim for people that want to use old coolers, or different size spacers or something along those lines.

This will be a CPU that will benefit from.lapping and delidding.

Are these soldered? Delidding a soldered CPU is a thing which isn't for anyone (trust me, I know personally )

 

[big_glasses]

Boot delay, because of RAM initialization, now that's something that could drive folks nuts who prefer to unplug their systems from the grid (f.e. with a smart socket).
Hope there is a way to set the "found optimum" RAM speeds fixed in the BIOS to skip this nonsense.

Shouldn't the CMOS battery make sure the settings/initialization/etc is kept even when disconnected?

I'm actually surprised that the IGP's 128 shaders are even somewhat capable. Though still GT 1030 goes way ahead. But usable for older games.

Yea, same. I expected the IGP to be only capable of desktop/cli useage

 

[jsven008]

94C at stock. Ooooof.

LOL! AMD said 95C (203F) isn't hot...

 

[W1zzard]

What happens to AIO's over extended period of time when the CPU is pumping out 95°C? Can they even keep up? Will they get overwhelmed and leak or explode?

The temperature at the AIO "side" is much lower. Technically bad things can happen if water inside an AIO starts to boil, because it expands, but I never even heard of such a case, nor experienced it. The thermal gradient CPU silicon -> IHS glue -> IHS -> coldplate -> water inside block would probably be too big for that to happen in the first place

Shouldn't the CMOS battery make sure the settings/initialization/etc is kept even when disconnected?

Assuming whatever takes so long can be saved and it gets saved to the CMOS

Showing RDR2 in the PS3 graph. I'm assuming that's RDR1 and that's just a typo.

Whoops, typo indeed. It is RDR1, fixing

https://twitter.com/x/status/1574398461642174464

Thank you very much for testing emulator performance. Did you enable AVX512 mode before testing RPCS3?

I used it at default settings. I did a quick test of the new AVX512 build yesterday, and gained around 10% in performance

 

[MarsM4N]

Shouldn't the CMOS battery make sure the settings/initialization/etc is kept even when disconnected?

Apparently, NO. Jay says in his video it's happening "everytime your system looses power long enough for the caps to drain, or you clear the CMOS, or update the BIOS, or make memory changes."
So how long does it take to drain the caps? Guess just a few hours. Needs to be tested, deal breaker for quite some.