Poll: Ryzen 7950X3D, 7900X3D, 7800X3D

[humbug]

Not sure what you’re getting at there?

During multi-threaded load (which is really why people like me buy such a CPU) the 5950x performance per watt is better than the 7950x, which is a step backwards for a company whose product would improve fundamentally rather Intels playbook of yoloing the power. (Which they were fairly and consistently critiqued for)

The 5950X completes Steve Burke's Blender run in 8.1 Minutes, this at 120 watts.

The 7950X completes the same run in 6 minutes flat at 263 watts.

The 5950X does it in 74% of the time for 47% of the power, so yes you're right.

Now lock the 7950X to 105 watts, Steve measured that at 158 watts and 6.4 minutes, recalculating that gives us 79% of the performance for 76% of the power, yes the 5950X is still more efficient but only just.

At 65 watts, 88 watts measured the 7950X completes it in 8 minutes dead, so now the 5950X is 98% the performance at 132% of the power.

If you set it to 105 watts, you will get as much as makes no difference the same performance per watts but also 27% higher performance.

 

[Spannerhead]

Sorry to invade this great thread where there are no arguments at all but does anyone know when these chips are gong to be avalible to pre order?

my medievil ryzen 1200 is approaching death . really need new hardware... annoyed this thing did not come out on 14th feb like the rumours said ... oh well

 

[FredFlint]

I'm not even sure waiting 12 months will help, as the IMC's won't be changing, only the BIOS. And I'm not sure how much BIOS changes can help in that regard. Although I'd love a shot at trying to drive all 128GB ram here at 6000Mhz if the BIOS updates could allow this, given I've tested and know that the kits I've got can hit that number in dual. But having seen the old x79 platform, I'm not putting too much hope over it. But would love to be surprised about it.

Think it will probably get to a point that 4*32 will work at 4800 without issues. AM4 improved RAM support a lot over its life, AM5 will do the same. That said, 4 stick has always been hit and miss for higher frequencies, more so with higher densities.

 

[Meddling-Monk]

Think it will probably get to a point that 4*32 will work at 4800 without issues. AM4 improved RAM support a lot over its life, AM5 will do the same. That said, 4 stick has always been hit and miss for higher frequencies, more so with higher densities.

That'll be nice also. Although I'm happy with 4000Mhz for the 1:1:1 deal, being able to push more from the RAM if low latency is not what you're after is always useful. Options are good.

 

[humbug]

Sorry to invade this great thread where there are no arguments at all but does anyone know when these chips are gong to be avalible to pre order?

my medievil ryzen 1200 is approaching death . really need new hardware... annoyed this thing did not come out on 14th feb like the rumours said ... oh well

Review on the 27'th, on sale, or pre order 28'th.

 

[Spannerhead]

Review on the 27'th, on sale, or pre order 28'th.

sweet tank you

 

[humbug]

Review on the 27'th, on sale, or pre order 28'th.

Sorry no that's wrong, it is the 14'th.

 

[Dave2150]

Sure, yet its faster at 65 watts than your CPU is at 140.

Indeed - how people aren't able to understand this is beyond me. They literally argue and think Zen4 is power hungry and hot, when the complete opposite is true.

The 7600, 7700 and 7900 at 65W are fantastic CPU's, efficient and powerful.

 

[humbug]

Indeed - how people aren't able to understand this is beyond me. They literally argue and think Zen4 is power hungry and hot, when the complete opposite is true.

The 7600, 7700 and 7900 at 65W are fantastic CPU's, efficient and powerful.

I find myself agreeing with you, amazing...

 

[hobib_ali]

Sorry no that's wrong, it is the 14'th.

changed

was the 14th feb but now its end of the month

will give people more time to think if they really need a 7950x3d

 

[humbug]

changed

was the 14th feb but now its end of the month

will give people more time to think if they really need a 7950x3d

I wish they would just make thier minds up, clearly reviewers already have them, JayZ2Cents is using his for his personal rig.

 

[Iron Man]

Only 1 week to go and looking forward to this. Only thing I’m unsure about is upgrading the bios when I build my new system but I’m told that Asus bios flashback works well so we will see. Still have nightmares when I tied to upgrade my bios back on the x570 for Zen3 as it didn’t go well!

 

[sprite-]

The 5950X completes Steve Burke's Blender run in 8.1 Minutes, this at 120 watts.

The 7950X completes the same run in 6 minutes flat at 263 watts.

The 5950X does it in 74% of the time for 47% of the power, so yes you're right.

Now lock the 7950X to 105 watts, Steve measured that at 158 watts and 6.4 minutes, recalculating that gives us 79% of the performance for 76% of the power, yes the 5950X is still more efficient but only just.

At 65 watts, 88 watts measured the 7950X completes it in 8 minutes dead, so now the 5950X is 98% the performance at 132% of the power.

If you set it to 105 watts, you will get as much as makes no difference the same performance per watts but also 27% higher performance.

I think you’re missing the point, beyond trying to use GamerNexus numbers to prove one (where in his conclusion he states blender isn’t reflective of production apps). It’s simply an analysis. Someone purchasing the product would never run eco mode, even he doesn’t recommend it.

You’d have to constantly flip between eco and non-eco to cover all the use cases you may face to optimise. Your only justifiable reason to use this is if you’re constrained by a small form factor so thermals considerations are important, or you’re running it in say a dedicated server or station with high up-time where multi thread workloads is it’s predominate (if not only) job.

Let’s go back to my original statement…… The 7950x is a step backwards for AMD, so why is that the case?

- The 3950x was more power efficient, cooler and had greater performance vs Threadripper (2950x)

- The 5950x was more power efficient, cooler and had greater performance vs the 3950x.

A pattern emerges…. Of course they’ll always hit a wall before a large architectural change but it’s been the big “boon” for AMD Products.

Then we get a 7950x, a product which uses significantly more power, is significantly hotter and is less power efficient.

How that can be anything less than a set back is beyond me, especially when Intel has been chastised (and rightly so) for pushing power to achieve generational performance increases in the same manner.

Is it a bad product? No, is it a disappointment from a company who was showing Intel up by delivering power and thermally efficient CPUs with a sizeable performance gain. Absolutely

 

[Deleted User 456458]

How do you conclude that AM5 is step backwards? That literally makes no sense.

I've bought and tested two Ryzen 7000 systems over the last few weeks, R5 7600 and a R7 7700. Both running EXPO, fully stable, no issues. Just fantastic performance (check reviews) and rock solid stability.

There are no DDR5 "shenanigans" - just people trying to mix different memory kits, or buying kits not on QVL list, the usual noob errors.

Calling people noobs while panicking if a mem kit isn’t on a qvl list. lol

 

[Greebo]

2 Sticks can get you 64GB and 2 sticks can generally reach a higher speed than 4 sticks.

There is always a trade off when going for max capacity and more sticks as its harder on the IMC and was the same on DDR4.

And yet never seemed go be an issue. My various 4ddr4 sticks always ran at their xmp speeds without issue and would overclock. Yes i wasnt expecting to break any records as for that you needed 2 sticks or even 1 but XMP worked. it seems with ddr5 xmp and expo doesnt work if using 4 sticks and you arent guaranteed to hit the quoted memory speed doing it manually either.

So yeah, ddr5 needs a year to mature.

 

[Greebo]

How do you conclude that AM5 is step backwards? That literally makes no sense.

I've bought and tested two Ryzen 7000 systems over the last few weeks, R5 7600 and a R7 7700. Both running EXPO, fully stable, no issues. Just fantastic performance (check reviews) and rock solid stability.

There are no DDR5 "shenanigans" - just people trying to mix different memory kits, or buying kits not on QVL list, the usual noob errors.

qvl lists was never an issue with ddr4 though. And with some baords the qvl lists are shockening short and limits you for some speeds to one brand/model when you might have a choice of 10 to buy.

 

[Greebo]

Think it will probably get to a point that 4*32 will work at 4800 without issues. AM4 improved RAM support a lot over its life, AM5 will do the same. That said, 4 stick has always been hit and miss for higher frequencies, more so with higher densities.

i want 4*32 at 6000 though. And there is no guarantee that if you buy a board now, that it will get any better. By next year there will be ddr5 10,000 ram around but that is no good still if when you buy 4 * 32 you end up running a lowly 4000 only still...........................

 

[FredFlint]

i want 4*32 at 6000 though. And there is no guarantee that if you buy a board now, that it will get any better. By next year there will be ddr5 10,000 ram around but that is no good still if when you buy 4 * 32 you end up running a lowly 4000 only still...........................

AMD list the spec on its site, I have not looked in a bit but think it says 4 * 32GB will run at 3600 so anything above that is overclocking which is not guaranteed. If you need 128GB of 6000+ RAM, don’t know if any platform supports that?

 

[junpit]

I think you’re missing the point, beyond trying to use GamerNexus numbers to prove one (where in his conclusion he states blender isn’t reflective of production apps). It’s simply an analysis. Someone purchasing the product would never run eco mode, even he doesn’t recommend it.

You’d have to constantly flip between eco and non-eco to cover all the use cases you may face to optimise. Your only justifiable reason to use this is if you’re constrained by a small form factor so thermals considerations are important, or you’re running it in say a dedicated server or station with high up-time where multi thread workloads is it’s predominate (if not only) job.

Let’s go back to my original statement…… The 7950x is a step backwards for AMD, so why is that the case?

- The 3950x was more power efficient, cooler and had greater performance vs Threadripper (2950x)

- The 5950x was more power efficient, cooler and had greater performance vs the 3950x.

A pattern emerges…. Of course they’ll always hit a wall before a large architectural change but it’s been the big “boon” for AMD Products.

Then we get a 7950x, a product which uses significantly more power, is significantly hotter and is less power efficient.

How that can be anything less than a set back is beyond me, especially when Intel has been chastised (and rightly so) for pushing power to achieve generational performance increases in the same manner.

Is it a bad product? No, is it a disappointment from a company who was showing Intel up by delivering power and thermally efficient CPUs with a sizeable performance gain. Absolutely

I don’t think it’s as simple as you suggest. The design which occurs many years before assumes a power profile as part of the design. Lots of variables entering into this like yield at the power profile and performance. It’s never about power in = performance out. I agree that the 5000 series was more power efficient at reduced power. It was designed to run as the last chip in AM4 so they were forced to accept the AM4 design constraints. Just because it can’t perform at lower power doesn’t make it a worse product unless your competing in some kind of performance per watt. The race is for physical performance at a sensible amount of power.

They’ve decided the TDP they needed to facilitate this.

It’s worth realising that the performance per watt will get better under AM5. Any sensible comparison we can make now needs to compare the processors in the same platform.

 

[sprite-]

I don’t think it’s as simple as you suggest. The design which occurs many years before assumes a power profile as part of the design. Lots of variables entering into this like yield at the power profile and performance. It’s never about power in = performance out. I agree that the 5000 series was more power efficient at reduced power. It was designed to run as the last chip in AM4 so they were forced to accept the AM4 design constraints. Just because it can’t perform at lower power doesn’t make it a worse product unless your competing in some kind of performance per watt. The race is for physical performance at a sensible amount of power.

They’ve decided the TDP they needed to facilitate this.

It’s worth realising that the performance per watt will get better under AM5. Any sensible comparison we can make now needs to compare the processors in the same platform.

All chip manufacturers focus on one main thing, how much performance can they achieve with the smallest amount of energy used.

It allows them to make a better product and reduce cost, it would have never of been AMDs power target at design stage. Remember, higher power consumption for the retail product will reduce their potential yields.

They were simply faced with a resurgent Intel due to their P core design and Intels willingness to push power limits to compete. They could no longer hold the advantage without also resorting to such behaviour.

Don’t get me wrong, I’m hoping AMD can respond and get back on track. Where true architectural improvements drive innovation, rather than “boot up the watts baby” approach. Otherwise it’ll just become the norm.