Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[unseenmorbidity]

ccr vöcvvvvv

That cooler has a delta-T of ~50C at 150W coming from an i7 920 D0 @ 4.2GHz. That die is 263mm^2, so has an easier time transferring heat than Ryzen will at its measely ~192mm^2.

So, given that 8-core Ryzen uses ~65W at 3.4Ghz (CPU-only) and ~150W at 4.4Ghz, has a likely safe temperature limit around 80C, and 37% less surface area for transferring heat, I suspect you will find yourself running about ~75C at 4.1GHz on 8-cores with that heatsink. Accounting for the added power draw for the higher temperature, you will likely "creep" upward from about ~72C a few degrees, which is how I come to ~75C.

A Noctua NH-D15, from my math, based on theory and very little real world data (so a few salt mines are needed to satiate the requirements here - as with the above), will allow about 4.25Ghz at the same temperature. Be mindful: these are equilibrium temperatures with assumed wattage outputs... lots of assumptions, in fact, such as 22C ambient temperatures (my house temps).

Standard water-cooling, however, should be able to keep the case(IHS) temperature low enough to keep the delta at the die-IHS interface high, which increases thermal efficiency, which reduces power needs, and allows for better results. The actual amount of heat is, by no means, a problem. It's getting the heat out of the die through its small interface with the IHS that is problematic.

A heatsink with a larger thermal mass will take longer to reach equilibrium temps, so you could probably run 4.4Ghz with gaming loads, if the CPU can handle it. Overclocking should improve with BIOS updates, so keep up to date

calculate 1600x with a noctua d15

go!

 

[looncraz]

We have to remember this is 8c cpu the 6900 at 4.3 uses 210w

Yikes. I'm hoping Zen follows the idealized curve more closely than that.(i.e., not needing a lot of voltage to jump through a barrier).

I had a feeling AMD was going to top Intel in efficiency. Anyone paying attention could have seen what a huge progress they have made on the dozer cores with Carrizo.

Absolutely. I'm sure AMD kept the clock mesh as well, which will have been tuned for about 3.4Ghz (about the max it can handle, IIRC).

Honestly I can't make sense of AMD's TDP numbers. The 65W sku's ship with 95W coolers; the 95W sku's ship with 140W coolers.

The 4C4T 1100 is rated 65W; the 8C16T 1700 is rated at 65W

Then the 6C12T 1600X is rated 95W

And the only leak we have re: power consumption (from Canard PC December leak) shows the 3.3 Ghz ES drawing 93W under load.

It's simple: AMD needs to keep the IHS as cool as possible to transfer the heat at acceptable temperatures. These CPUs are going to run hot - but aren't actually putting out a lot of heat, they just have a high concentration of heat in a small area and difficulty transferring that heat away.

The low-power 8-core is running in the high-efficiency region of the clock mesh technology, assuming it is still being used, so it can be much more efficient than a six core running at higher clocks.

I have a D14 myself and does a pretty good job on my 4.8Ghz 2500k. What kind of power draw does a 4.8Ghz has and what kind of heat dissipation, I don't exactly know, but I could measure the delta from stock and get a ballpark figure.

Now these things (the D14 and D15), can dissipate up to 220W. That seems like a lot of headroom from the stock 95W.

I have a D15. My 2600K outputs 150W into it at 4.5Ghz when at 60C. It climbs to 180W at 75C. I'm using PWM, so the temperatures could be kept lower with higher fan speeds, but I intentionally equalize at 77C (~186W) using "Very High" Intel Burn Test, as this keeps my system quiet during my normal loads (it barely gets any louder when encoding with HandBrake for hours, for example).

The die size here is 216mm^2, ~12.5% larger than Ryzen's die. Ryzen 8-core should have about as much heat output as Sandy Bridge 4-core at the same frequencies (which is absolutely amazing when you stop to think about it), but worse thermal transfer, which is something that should mean that you should expect higher temperatures at the same frequency.

 

[looncraz]

ccr vöcvvvvv

calculate 1600x with a noctua d15

go!

LOL! There a few more factors involved with this, but I'll assume a 3+3 configuration.

~93.35W at 4Ghz on all cores @ 60C. Given you will not maintain 60C die temp with the D15 (65C transient), you will see a gentle increase towards 107.95W over 15~30 seconds with temperatures reaching towards 70C and likely exceeding it in spots (all of my temperatures are average die temps...).

If you run max fan speeds, though, you will be able to maintain those temperatures (almost exactly) at 4.2Ghz. 4.4Ghz will push you right to 80C and 135W power usage.

 

[.vodka]

I have a D15. My 2600K outputs 150W into it at 4.5Ghz when at 60C. It climbs to 180W at 75C. I'm using PWM, so the temperatures could be kept lower with higher fan speeds, but I intentionally equalize at 77C (~186W) using "Very High" Intel Burn Test, as this keeps my system quiet during my normal loads (it barely gets any louder when encoding with HandBrake for hours, for example).

The die size here is 216mm^2, ~12.5% larger than Ryzen's die. Ryzen 8-core should have about as much heat output as Sandy Bridge 4-core at the same frequencies (which is absolutely amazing when you stop to think about it), but worse thermal transfer, which is something that should mean that you should expect higher temperatures at the same frequency.

I happen to have similar temperatures on my 4.5GHz/1.3v 2500k with my heatsink which is similar to a Noctua NH-D14.

At least Ryzen is supposed to be soldered, I remember reading The Stilt hinting it would be in some thread a while ago.

amazing data!

I'm crossing my fingers it turns out like this and I don't have to move over to watercooling. Mind you if it's absolutely needed I will. I'm also looking forward to what refined BIOSes do to overclocking capability... after all this is a fresh out of the oven platform that will get lots of tweaking going forward this year.

This first batch of chips is going to be rough around the edges, second or third batch near the end of the year should be a different story.

 

[Jan Olšan]

Honestly I can't make sense of AMD's TDP numbers. The 65W sku's ship with 95W coolers; the 95W sku's ship with 140W coolers.

This is not new or "AMD's". Core i5-6600K and i7-6700K already are marketed as "91W TDP", but at the same time, Intel states that the cooling needs to be rated for 130 W. Pretty sure Kaby Lake is the same deal.
Basically it just means that you are supposed to cool the chips better than what was required for CPUs with equal heat output in the past.
Imagine that both chips are 95W, but one made with 14nm is more finicky and needs better cooling to be comfortable with that power consumption.

 

[Saylick]

This first batch of chips is going to be rough around the edges, second or third batch near the end of the year should be a different story.

Oh man, imagine a "Q6600 G0"-style stepping improvement on Ryzen within one year of launch.

 

[ultima_trev]

Something seems aloof.

AnandTech's own test of the i7 6900K achieved 1547 multi thread (but only 153 single thread, versus 162 in AMD's slides).

Still seems promising. I wonder how aggressively XFR was working for 1800X here.

 

[looncraz]

Something seems aloof.

AnandTech's own test of the i7 6900K achieved 1547 multi thread (but only 153 single thread, versus 162 in AMD's slides).

Still seems promising. I wonder how aggressively XFR was working for 1800X here.

https://semiaccurate.com/forums/showpost.php?p=283125&postcount=7108

 

[.vodka]

Oh man, imagine a "Q6600 G0"-style stepping improvement on Ryzen within one year of launch.

Sure. Expect that to happen at some point this year or around this time next year. It's always happened before as processes and designs get tweaked

E6xxx B0 -> G0
Q6xxx B0 -> G0
E8xxx C0 -> E0
i7 9xx C0 -> D0
i7 4770k -> i7 4790k
i7 6700k -> i7 7700k


Initial 2GHz 90nm Athlon 64 that ended up at 3.2GHz
Phenom II x4 940 -> x4 955 and later revisions, like the 3.7GHz 980BE
FX8150 (3.6GHz base) -> FX8350 (4GHz base) -> FX8370/e (similar clocks, lower power after years of refining)



Yeah, it'll happen. Polaris 10/11 were used as pipe cleaners for Glofo's 14nm process last year (look at the initial RX480s and the XFX GTR RX480 in power, overclocking capability and thermals), and it'll continue to improve. If Samsung is also manufacturing Ryzen, well, that's another thing to look for if there's enough differences when pushing those chips at their extremes.

We've already seen progress, engineering samples last year around December were running at low 3GHz, now we have the R7 1800x that does 4GHz out of the box in some form.

I mean, they've reserved the R9 name for an eventual even higher end product... I wonder what that could be. Zen+?

 

[ultima_trev]

https://semiaccurate.com/forums/showpost.php?p=283125&postcount=7108

Good find. Illustrates how much AIO cooling can benefit versus crappy fan/HSF concoctions.

 

[OrangeKhrush]

I've lubbed my wallet like a gentleman.

 

[badb0y]

I am ready to jump into that octa-core. WITNESS ME!

 

[piesquared]

Zen is looking great for AMD. The high IPC and efficient design allows Ryzen to go toe to toe and win some [edit](but dominating the competition's products in $/perf) at the high end and combined with the 14nm LPP process should work out really well for mobile where the clocks will be in the sweetspot. Especially with a Vega class GPU on board. In a short period of time AMD could really raise the ceiling for their minumum graphics performance with Raven Ridge and have Ryzen with Vega at the high end.

 

[piesquared]

Server also looks quite promising with the option to pair Naples with Vega. Looking forward to hear more about the Infinity Fabric.

 

[PPB]

It's just scary the SMT yield AMD achieved on their second try globally, and their first try for all pipelines. I think it's really a benefit of having INT and FP separated like AMD does from several generations ago. The Port scheme of Intel IMO seems more or less a nightmare to have a good yielding SMT going on.

 

[Grooveriding]

I am ready to jump into that octa-core. WITNESS ME!

Nice.

Still not going to get hyped about leaks. Need to see real benches, but I want to see them pan out. It's not even expecting to see new levels of performance over what Intel has. It's wanting to finally see legitimate competition for performance levels that are currently only available from Intel.

I like that it looks like the competition will be aimed for Intel's very best and most expensive consumer CPUs; their hex and octa cores. Intel is going to town right now on their HEDT pricing and it would be nice to see a shake up.

If the Ryzen octocores are legitimate competition to the 6900k, then I don't see much value in Intel's 4c/8t chips going forward. I expect there will be a ST advantage for a 7700k, but there are solid gains today in more than 4 cores. A lot of games are taking advantage of 6+ cores, then you have encoding. It comes down to how large the ST deficit is between KL and Ryzen, and for the enthusiast, how they compare to one another with overclocking. If when all is said and done Intel only has about a 10% lead in ST, then I personally would recommend the 8 core Ryzen over the 4 core Intel chip.

Hopefully reviews are coming this month.

 

[Magic Hate Ball]

Noctua Nh D15 is just 1-2 degrees higher than the best of best AIO's under the same loads. So I think if Ryzen 7 chips can hit 4,5 Ghz with realistic voltages, then Noctua Nh D15 won't have a problem to keep up with that clock.

I believe this. I just got my NH-D15S in the mail tonight and holy cats this thing is yuuuuge.

 

[GroundZero7]

Yaup..That's why I'm gonna get the Corsair H115i AIO...don wanna crack my mobo sticking a D15 in a tower.

 

[bjt2]

How did amd outdo intel in the power efficiency game? The 1700 @ 3ghz has a 65w tdp while a 3.2ghz 6900k is 140w... mind boggling i tell ya.
Currently the 6900k is $1050 on newegg with a $50 discount, lets see how fast this will drop.

Zen has higher pipeline length or lower FO4: for the same clock it requires less Vcore.
Actually at the ISSCC conference was shown a slide in which the Vcore for Zen was 0.9V. I suppose at stock clock. 0.9V at 3.4-3.6GHz on a worse process is remarkable. But only possible with lower FO4.

 

[bjt2]

I had a feeling AMD was going to top Intel in efficiency. Anyone paying attention could have seen what a huge progress they have made on the dozer cores with Carrizo.

See here: http://www.bitsandchips.it/52-english-news/8058-xfr-and-its-anchestors

 

[bjt2]

Honestly I can't make sense of AMD's TDP numbers. The 65W sku's ship with 95W coolers; the 95W sku's ship with 140W coolers.

The 4C4T 1100 is rated 65W; the 8C16T 1700 is rated at 65W

Then the 6C12T 1600X is rated 95W

And the only leak we have re: power consumption (from Canard PC December leak) shows the 3.3 Ghz ES drawing 93W under load.

It's an ES. And 93W is on the 12V rail. Multiply for .86-.88, that is the VRMs maximum efficiency and you'll get the actual CPU consumption (80-82W)...

 

[EliteRetard]

Here's a curious question, if this XFR does turn out to be kinda like self overclocking. How does the chip know when it's clock speed has gotten to high and started causing errors? Just because it's below temp and power limits, doesn't mean the clock speed is stable/error free.

 

[bjt2]

Here's a curious question, if this XFR does turn out to be kinda like self overclocking. How does the chip know when it's clock speed has gotten to high and started causing errors? Just because it's below temp and power limits, doesn't mean the clock speed is stable/error free.

Since Carrizo there are small replicas of critical circuits spread around the chip, that are periodically exercised and tested for minimum Vcore at given frequency... Have a look at the link I posted some posts ago in this thread.

 

[Dante80]

Been lurking for some time, decided to make a post on this.

What I understand so far.

1. Almost all leaks come from the ES samples that AMD gave out to the motherboard manufacturers for board development/fitting etc. Those samples may have binning issues or disabled XFR/boost/other capabilities (as far as I remember, only haswell got a better ES OC than a product OC).

2. In the performance front, this time it seems that AMD is hitting their IPC goal numbers (40% above BD). In absolute numbers, that is around 10%-15% below the latest and greatest from the blues. It is fine, especially if you take into account stock TDP, core and price numbers.

3. The price/performance ratio when you compare the chip to Broadwell-E is very competitive.

4. AMD may be aggressively binning their parts for first launch. Meaning that getting the pricier X models may not only give you better stock clocks and XFR, but also more OC headroom.

5. Jim Keller is pretty good at his job. Lisa Su is also good at her job. AMD has been too terrible, too long, and that changing is a breath of fresh air for the market.

6. Regarding gaming performance vs the 1151 parts. I think that Ryzen does not have the clocks and single core performance to match the 6700/7700 products. And that is fine. By the time that a game comes out that pushes current CPUs to the point of effectively throttling your gaming experience (for example, sub-60 FPS minimums at your playing resolution), the games existing would already be in need of more cores than what your standard i5 or/and i7 can handle.

Moreover, more cores/threads may mean better simultaneous streaming and encoding performance for enthusiasts that carry the market (streamers, youtubers et). This is a plus.

In other words. If you already have a decent OCing i5/i7 for gaming right now, there is no need to replace it for a ryzen system. If you are making a new PC though and are a gamer, it may make sense to buy one right now, if only for more future-proofing (since the real day to day performance delta in gaming is non-existent, almost all games are GPU limited).

Lastly, I'm wondering whether we are going to see any VR gaming benchmarks too. This is relevant to my interests, and from what I understand the Vive and other VR headsets really do want as many cores and as much GPU as you can throw at them, though GPU certainly helps much more. Having more CPU cores does help in getting more consistent (minimum) framerates when multitasking, which is vital for the 3D effect to not become unpleasant.

 

[Timmah!]

"How did amd outdo intel in the power efficiency game? The 1700 @ 3ghz has a 65w tdp while a 3.2ghz 6900k is 140w... mind boggling i tell ya.
Currently the 6900k is $1050 on newegg with a $50 discount, lets see how fast this will drop."

Since quoting is broken for whatever reason for me....

this is what interests me personally as someone who owns X99 platform only for about half year.... if all these leaks about performance and prices are true and AMD is truly gonna sell their 8-cores for as little as 1/3 of the price of Intel´s ones, while performing pretty much the same, do you think we are in for massive price-drops of entire Intel´s HEDT line-up? Current BW-E i mean... No way i am going to sell my entire rig only to replace it with Ryzen, too much hassle with that for me, but i could think about at least upgrading my CPU (6850k) for something faster, if the price becomes right as the result of Ryzen´s competition. Ideally 6950x... what do you think are the chances its gonna drop to say 1000?