Official AMD Ryzen 5000 Series & 7000 Series Thread

[Android1]

AMD Ryzen 9 7950X Geekbenched, Crushes i9-12900K, in the League of the i9-13900K

If these numbers prove to be correct when the retail chips are benchmarked, AMD has a very capable product on their hands.

An AMD Ryzen 9 7950X "Zen 4" 16-core/32-thread processor was put through the Geekbench 5.4.5 benchmark, and it's becoming all too clear that AMD has a highly competitive product on its hands. The 7950X yielded a single-threaded score of 2217 points, and 24396 points in the multi-threaded tests. With these scores, the 7950X is about 14% faster than the "Golden Cove" P-cores of the i9-12900K "Alder Lake" processor in the single-threaded tests, and comes out as being 41% faster than it in the multi-threaded test. Against the leaked i9-13900K "Raptor Lake," the 7950X is shown being about 4% slower in the single-threaded test (against the "Raptor Cove" P-cores); and about 7.8% slower in the multi-threaded test.

Sources: techPowerUp!, Benchleaks (Twitter), VideoCardz.com

[acroig]

I can't wait for the benchies.

[Android1]

I can't wait for the benchies.

Me too!

[Nunz]

If you plan on buying Zen4, go buy these sticks *RIGHT NOW* before they are all out of stock and/or prices go up. They also make them with RGB if you prefer that for $5-10 more. I just grabbed a set and they absolutely destroy the TeamGroup 6200 kit I had.

https://www.kingston.com/en/memory/g...density=16gbit

The heatsinks are good quality as well. Better temps at higher voltage than the TeamGroup set. They're extremely cheap for the best PMIC you can have (Richtek) and the best RAM chips (Hynix).

[0091/2]

I haven't been following Intel arch designs. Can someone explain to me why is 24c only 32t? Are their P core single thread only and E core for the MT?

[the_sextein]

The 8 P cores are hyperthreaded and the 16 E-cores are not.

[0091/2]

The 8 P cores are hyperthreaded and the 16 E-cores are not.

[Riptide]

If you plan on buying Zen4, go buy these sticks *RIGHT NOW* before they are all out of stock and/or prices go up.

I'm within the return period for the 6000 MT/sec CL40 stuff I bought so went ahead and got this upgrade for another $55.

[Nunz]

I'm within the return period for the 6000 MT/sec CL40 stuff I bought so went ahead and got this upgrade for another $55.

I would have just manually OC'd it to match but hey, get it!

I'm running it at stock voltages at 6600 without touching anything which is pretty nice. Once I get back in a few days, we'll shoot for 7000.

[Nagorak]

^^ lol

Hard to say when the 3d cache versions would go up for sale if they debut at CES. Maybe Q2 if not earlier.

Makes it real tough to justify spending money on an initial launch CPU if you know in your of that you'd end up wanting one of the 3d versions anyway.

OTOH, launch motherboards and memory should work fine with it. So if you don't mind a hold-over CPU and getting perhaps 50% back when you go to sell it then no big deal.

Decisions decisions..

As someone who bought Ryzen 1000 at launch, my advice is to wait 3-4 months before buying. Being an early adopter on AM4 was really rough. Hopefully this time goes better, but it's still safer just to wait and see how the dust settles.

Waiting also has the advantage of allowing you to see what Intel brings, as well as the X3D chips, which I think are the better buy if you are primarily gaming.

[Riptide]

I had issues with my 1800x on launch too and like most people it was memory related. I was able to achieve stability quickly by setting the memory back to 2133mhz. Clearly not optimal but the system operated and I was OK until firmware updates came out.

While I'm not super concerned about having a completely inoperable client at or near launch, I will eventually want a 3D cache chip. So some money would be wasted in a hold-over scenario. The primary consideration would then be which launch option will hold the best value when I go to re-sell in six months.

I'm tempted to wait for Intel however one of my admittedly specialized use cases is CPU mining. Intel options really can't come close to competing with the efficiency of Ryzen in that department.

[acroig]

What's the chance AM5 won't have any of those issues?

[the_sextein]

There are rumors going around that the 7950 his really hot. That's been my fear the whole time. If the 7000 series is power efficient, that's great but heat is my biggest issue. I'm just going to wait for the final word on thermals from 3rd party testers. If these end up running at higher temps than the 13900K then I'm out.

[Nunz]

What's the chance AM5 won't have any of those issues?

Zero

[Jay20016]

What's the chance AM5 won't have any of those issues?

Crossing my fingers the growing pains will not be as abundant but as with anything new there are going to be quirks. Hopefully the success of Ryzen has allowed AMD to invest into this more than they were able to with OG Ryzen.

But, for now I'll enjoy my newly installed 5800x3D in the same board that has seen a 1700x, 3700x, and now this 5800x3D and hope when I do jump up to the next generation they're gone.

[acroig]

Zero

Crossing my fingers the growing pains will not be as abundant but as with anything new there are going to be quirks. Hopefully the success of Ryzen has allowed AMD to invest into this more than they were able to with OG Ryzen.

But, for now I'll enjoy my newly installed 5800x3D in the same board that has seen a 1700x, 3700x, and now this 5800x3D and hope when I do jump up to the next generation they're gone.

I really wanted to go AMD for that same reason Jay but I'm starting to lose faith. Intel's 13700K will be out way before the 7800x3D will.

[pax]

Some good results with manual voltage settings:

https://twitter.com/harukaze5719/sta...85953056681984

Looks like at least with golden samples you can expect almost half drop in watts and temps

Aida64

122w down to 68w

93c down to 56c

[Riptide]

Zero

So what you’re saying is.. there’s a chance!

[Nunz]

AIDA64 is not a stability test worth running or trusting, so there's no guarantee that the system is stable at those voltages. Those are also really low wattage numbers for the temps shown.. leads me to believe there's a mounting issue. The higher the wattage, the more it would exacerbate the problem.

[acroig]

AIDA64 is not a stability test worth running or trusting, so there's no guarantee that the system is stable at those voltages. Those are also really low wattage numbers for the temps shown.. leads me to believe there's a mounting issue. The higher the wattage, the more it would exacerbate the problem.

Excellent point, let's see what GN and other sites make of this on the 27th.

[pax]

Wouldnt be surprising to see the need for a bracket on AM5 as they got for intel. But even if the results here are extreme it points to a possible decent improvement with UV.

I just hope they let undervolting on X3D parts as well. Even a mild undervolt can make a big diff.

[acroig]

https://wccftech.com/amd-ryzen-9-795...3-debut-rumor/

[Nunz]

Wouldnt be surprising to see the need for a bracket on AM5 as they got for intel. But even if the results here are extreme it points to a possible decent improvement with UV.

I just hope they let undervolting on X3D parts as well. Even a mild undervolt can make a big diff.

Decent improvement with UV if it's stable. Cutting power consumption in half would require a massive undervolt; I'd venture at least a 600mv drop.. which again, no way

[pax]

Wish we knew how much of an undervolt he did here but UV has been a regular thing on ryzen since the 3000's at least. Im not that surprised by the results other than he had a golden sample but you dont need a golden sample to get a good undervolt on ryzen.

But ya further tests to determine stability is warranted.

[SSXeon]

Some good results with manual voltage settings:

https://twitter.com/harukaze5719/sta...85953056681984

Looks like at least with golden samples you can expect almost half drop in watts and temps

Aida64

122w down to 68w

93c down to 56c

I saw the 95*C rumors, and if Intel can keep their CPUs relatively cool for Alderlake with 200+W, I can't see AMD failing on that. I could be wrong, but I don't see that happening.

Lets wait for real benchies before we jump the gun with ES chips. Remember the 5950X was running like ~120W vs my 5800X @ ~140W.

https://wccftech.com/amd-ryzen-9-795...3-debut-rumor/

Yeah and rumored @ 200Mhz lower clocks but like 20-30% faster!

[Nunz]

AMD is using a significantly denser node than Intel. Heat dissipation and thermal density are more of an issue than what Intel is going to experience. I can easily see 200watt having issues considering that was already difficult to cool on Zen3, and now AMD has stuffed 50% more transistors into Zen4.

[Riptide]

I have a EK 360mm AIO with a six fan setup on there so it better be able to keep up with this thing or it's all going back for a refund.

[acroig]

NDA up on 9/20?

[pax]

Ya a one week NDA lift is kind of unusual. Gives people plenty of time to see the independent reviews.

[acroig]

Ya a one week NDA lift is kind of unusual. Gives people plenty of time to see the independent reviews.

Can't wait.

[Nunz]

I have a EK 360mm AIO with a six fan setup on there so it better be able to keep up with this thing or it's all going back for a refund.

Hah. That's the issue with heat caused by thermal density.. the limitation isn't going to be your cooler; hell, it might not even be the IHS. Transferring the heat from that tiny little die generating heat at the rate of the sun is going to be a tough issue.

I think Zen4 is going to show that the way we currently handle the cooling aspect of processors is ancient and needs to see innovation. Direct-to-Die from the factory will help, but that's still a band-aid solution.

I haven't heard much about the tests that were being done with water channels being cut into the substrate of the die. The closer we get to the heat source, the better cooling we'll get. If we can have an IHS on a processor that has water-channels running through the substrates down to the transistor level, and then slap a waterblock on top of that IHS.. that's the next generation of cooling if you ask me.

We don't have much further to go before we simply won't be able to cool these chips. I don't care what AMD says - hitting 90c at stock is not good for the life of a processor. Three things cause degradation .. heat, wattage, and current. If you only have one of the three, you'll likely be OK.. but as these noses get smaller, the *safe" wattage/current is dropping rapidly. There's a guy that degraded his 12900K running Y-Cruncher for a few hours with the chip at 90-95c with a ~350-375watt load. As we continue shrinking nodes, that allowable wattage is only getting smaller.

[Riptide]

Can’t they start using liquid metal between the die and the IHS straight from the factory to help the issue?

[Nunz]

It helps, but eventually that won't be enough either.

Solder is much safer for them to use for factory installs so I doubt we'll see that change until the entire concept of processor cooling moves forward.

This kind of product is what I think should be offered from factory. These guys make their stuff in Thailand; very popular with the XOC crowd. I'm looking at grabbing their DDR5 copper waterblock, but their CPU blocks have garnered a lot of recognition the past year. They're still expanding.. last I looked they didn't even have a website, you used to order through their Facebook and pay w/ PayPal lol.

http://www.supercoolcomputer.com/en/...direct-die-amd

[the_sextein]

I agree, heat has really become an issue to the point where I value low temps as much as high performance. Back when I bought the 9900K it wasn't a big deal but the 9900K in prime 95 made me realize that something needed to happen. I kept thinking they would come up with something, but each new generation of CPU just keeps getting worse and worse. I don't think performance users really care about power usage as long as it doesn't become a thermal issue. I knew AMD would have lower power usage but with the clock bumps I figured heat could be a problem still. I was hoping they had figured something out this time, but it doesn't look like it from what has been leaked. Sure you can under volt it but will it still hit the advertised speeds if you do? Probably not so you are paying for speeds that you can't use if you actually multi thread with it or it will kill the chip. Heck, even the 7600X is hitting 90C+ from the looks of it.

[Riptide]

Lump me in with those who don't care about power use until it becomes a legit thermal issue.

Kingston running a sale today guys. CL32 6400 memory for $235. 32GB.

I called and they even credited me the $30 difference vs. when I ordered two days ago. And I got someone on the phone in 2 minutes.

Not bad!

[Nunz]

Thank you Riptide! Just ordered 4-5 days ago so I should be eligible for that refund as well. $235 is INSANE

[pax]

It helps, but eventually that won't be enough either.

Solder is much safer for them to use for factory installs so I doubt we'll see that change until the entire concept of processor cooling moves forward.

This kind of product is what I think should be offered from factory. These guys make their stuff in Thailand; very popular with the XOC crowd. I'm looking at grabbing their DDR5 copper waterblock, but their CPU blocks have garnered a lot of recognition the past year. They're still expanding.. last I looked they didn't even have a website, you used to order through their Facebook and pay w/ PayPal lol.

http://www.supercoolcomputer.com/en/...direct-die-amd

Does that include the delidder? Derbauer is supposed to delid a 7000 gen ryzen soon... If its safe to use I can see it being well worth it for the X3D versions.

[acroig]

https://wccftech.com/kingston-adds-a...0-mbps-speeds/

[Nunz]

Does that include the delidder? Derbauer is supposed to delid a 7000 gen ryzen soon... If its safe to use I can see it being well worth it for the X3D versions.

With the capacitors being on the top-side of the chip, I really can't see how delidding will be possible without extreme risk. High risk, high reward though..

The supercool direct-die block does not come with a delidder, as far as I know. It's only ~$80 USD for the block and direct-die IHS, so I'd assume not. I don't see any pics of a delidder in there.

[pax]

That big black screw on top of a piece there looked a bit like a delidder for a moment there... thought it was to twist it off but guess its just there to press the cooling block back on tightly.