Discussion AMD Ryzen MegaThread! FAQ and Resources

8350rocks · Feb 16, 2017

jdwii :

juanrga :

Isn't that what Intel already does what makes that technology so great?

XFR is an all core turbo. It has been since the Discovery tablet nearly 5 years ago.

TMTOWTSAC · Feb 16, 2017

To be fair, it's not like WCCFTech is the source of every crazy rumor. They're more like that friend who believes every single thing they hear and forwards it to you. And they have no trouble publishing conflicting stories.

https://www.youtube.com/watch?v=KCSA7kKNu2Y

Anyway, back on topic. It sort of sounds like AMD's response to the latest Turbo Boost that tries to identify the fastest core and shift single core workloads to it. As usual though, it's going to be highly dependent on software support.

nikoli707 · Feb 16, 2017

im excited for amd. im still happy with my 2600k but ive been thinking about an upgrade lately. everything i keep learning about ryzen keeps getting better and better so come the first week of march i will be neck deep into reviews and benchmarks. pretty cool.

jwcrellin · Feb 16, 2017

I've been wanting to go 8 core but I don't want to spend $1000 on a i6 6900K. Ryzen might be the solution for me.

jdwii · Feb 16, 2017

nikoli707 :

Man you got the Kraken X62 it came out like 2 weeks after i got my X61 ha ha. I really can't wait to see what we can get Ryzen to when OC i got lucky with my chip but i see kaby-lake users getting 5.0Ghz like crazy like the new norm. Ryzen i personally don't think will hit that.

2600K(sorry gamer ha ha) is on its way out its a decent amount slower in IPC compared to skylake-kabylake and if the rumored benchmarks are right about Ryzen on the single core i think IPC will be better for Ryzen then the 2600K.

Sandy.........................Ryzen..Haswell(IPC), on average is my best guess now.

nikoli707 · Feb 17, 2017

jdwii :

yeah the kraken x62, nzxt s340 elite, and evga g2 850w were my christmas present to myself.

i never really intended on going so flashy and high dollar but i had $350 of basically free money from recycling lead acid batteries, like 1200lbs of them, so i treated myself. plus im redoing my office and im one of the sheep jumping on the rgb theme bandwagon

but i hoping for the best for amd and even though its looking like haswell ipc that still very good but the proof is in the pudding. im going to let this all play out and see where pricing goes. im hoping skylake-e will have much better thermals and overclocking headroom, and hopefully priced much lower than usual due to ryzens competition. as much as i like the 7700k i wont be doing 4c8t in my upgrade. i need more cores for adobe premier.

8350rocks · Feb 17, 2017

nikoli707 :

jdwii :

yeah the kraken x62, nzxt s340 elite, and evga g2 850w were my christmas present to myself.

i never really intended on going so flashy and high dollar but i had $350 of basically free money from recycling lead acid batteries, like 1200lbs of them, so i treated myself. plus im redoing my office and im one of the sheep jumping on the rgb theme bandwagon

but i hoping for the best for amd and even though its looking like haswell ipc that still very good but the proof is in the pudding. im going to let this all play out and see where pricing goes. im hoping skylake-e will have much better thermals and overclocking headroom, and hopefully priced much lower than usual due to ryzens competition. as much as i like the 7700k i wont be doing 4c8t in my upgrade. i need more cores for adobe premier.

Adobe premier might do well on Ryzen...the FX series always showed strongly in adobe premier.

jdwii · Feb 17, 2017

nikoli707 :

jdwii :

yeah the kraken x62, nzxt s340 elite, and evga g2 850w were my christmas present to myself.

i never really intended on going so flashy and high dollar but i had $350 of basically free money from recycling lead acid batteries, like 1200lbs of them, so i treated myself. plus im redoing my office and im one of the sheep jumping on the rgb theme bandwagon

but i hoping for the best for amd and even though its looking like haswell ipc that still very good but the proof is in the pudding. im going to let this all play out and see where pricing goes. im hoping skylake-e will have much better thermals and overclocking headroom, and hopefully priced much lower than usual due to ryzens competition. as much as i like the 7700k i wont be doing 4c8t in my upgrade. i need more cores for adobe premier.

As others know i think Haswell level of IPC is really good almost remarkable given the fact that i had zero hope for Amd's branch prediction as well as their odd decision of a 4ALU+2AGU design. Possibly adding twice the L2 cache over comes this?

As 8350rocks said a 8 core part will definitively increase your productivity in adobe premier and i can understand wanting twice the cores if the cores are actually worth it.

jwcrellin · Feb 17, 2017

That AsRock board is looking kinda pretty.
http://www.tweaktown.com/news/56321/asrocks-ryzen-motherboards-led-x370-killer-sli-ac/index.html

juanrga · Feb 17, 2017

The information that I have is that XFR is not an all-core turbo.

There are eight frequencies defined in Zen. Four of them are for engineer's interest only like F_boot (the boot frequency) or F_min, the minimal working frequency (actually 550 MHz on desktop chips). The other four frequencies are of consumer interest: F_P0, F_TMT, F_TST, and F_MAX.

F_P0 is the base frequency. It was 3.15GHz for the engineering sample used by Canard in their preview. It is 3.6GHz for the 1800X model.

F_TMT is the all-core turbo. It was 3.3GHz for the engineering sample. I don't know its value for the 1800X. I know it is somewhat between 3.65GHz and 3.8GHz.

F_TST is the 1-core turbo. It was 3.5GHz for the engineering sample. It is 4.0GHz for the 1800X.

F_MAX is a second 1-core turbo. This frequency is not predefined by AMD. It is a variable frequency that depends on measurements of several on-the-fly parameters like the temperature. Different 1800X chips will have a different F_MAX depending on silicon lottery and the quality of the cooling you have installed. This is an auto-overclocking turbo. This is XFR.

salgado18 · Feb 17, 2017

juanrga :

At the very least, if Turbo frequency is with one core, logic dictates that XFR, which is above it, is with one core. It should help with single-threaded benches, which is Ryzen's weakness versus Kaby Lake and friends. Although it would be great if there were a full-core XFR, but I think that would extrapolate the rated TDP, and that can't happen from the manufacturer's side, right?

juanrga · Feb 17, 2017

Another leaked score for Ryzen

http://www.passmark.com/baselines/V9/display.php?id=77190436599

Compared to other chips:

i7-6900K @ 3.20GHz = 17714

i7-5960X @3.00GHz = 15979

Ryzen @3.4GHz = 15085

i7-6800K @3.40GHz = 13583

Using the 8C results, we obtain Zen IPC is ~20% behind Haswell and ~25% behind Broadwell in this bench.

jwcrellin · Feb 17, 2017

I'll take 20% slower for 1/2 the cost.

salgado18 · Feb 17, 2017

juanrga :

The eternal question: what about turbos for each one? I don't trust benchmark software to get the correct frequency.

juanrga · Feb 17, 2017

salgado18 :

Turbos are 3.5GHz for Haswell, 3.7GHz for the 8C Broadwell, and 3.6GHz for the 6C. Ryzen turbo is 3.8GHz.

The Ryzen ID is ZD3406BAM88F4_38/34_Y this corresponds to a chip whose base frequency is 3393 MHz and software reports 3.39GHz. It all looks correct to me.

juanrga · Feb 17, 2017

I am not sure what think about this

http://wccftech.com/amd-ryzen-5-1600x-cpu-benchmark-leak/

salgado18 · Feb 17, 2017

juanrga :

They compared the results of the R5 1600X, with 3.7 GHz turbo, agains an i5 7600k, with 4.22 GHz turbo. They got 1888 and 2130 points respectively in CPU-z's single threaded bench. Calculating i5's score to 3.7GHz, we get 1867 points, which puts Ryzen at the same IPC as Kaby Lake. Something wrong is not right here

I say wait for oficial benchmarks and reviews.

Edit: OORRR XFR is at play, and the actual clocks were much higher than 3.7. However, to reach i5 performance, it needed to boost up to at least 4.4 GHz, if we consider a very optimistic 20% IPC deficit, and 4.6 GHz for a 25% difference. Could it be possible?

Crumpet 1 · Feb 17, 2017

salgado18 :

juanrga :

They compared the results of the R5 1600X, with 3.7 GHz turbo, agains an i5 7600k, with 4.22 GHz turbo. They got 1888 and 2130 points respectively in CPU-z's single threaded bench. Calculating i5's score to 3.7GHz, we get 1867 points, which puts Ryzen at the same IPC as Kaby Lake. Something wrong is not right here

I say wait for oficial benchmarks and reviews.

Edit: OORRR XFR is at play, and the actual clocks were much higher than 3.7. However, to reach i5 performance, it needed to boost up to at least 4.4 GHz, if we consider a very optimistic 20% IPC deficit, and 4.6 GHz for a 25% difference. Could it be possible?

Frankly i'm happy with either of those situations.

cdrkf · Feb 17, 2017

juanrga :

I don't know much about the CPU-Z benchmark- not one I've seen used much. It's quite possible it's *integer focused*, in which case Kaby and Ryzen have very similar capabilities. Even if it's FPU based, so long as it doesn't use higher than 128 bit instructions in theory Kaby and Ryzen are the same in terms of max throughput. The issue of IPC between Ryzen and Kaby isn't really one of who has the most execution resources but comes down to other things like the branch prediction capabilities and cache, if this is a simple enough test that might not come into play at all.

That said it's nice to see a test where Ryzen is roughly equal to Kaby clock for clock- given what we've seen of the design I think it's quite possible there will be more of these- Ryzen doesn't have the inherent deficiencies that the construction cores had (with the execution of 256 bit+ FP instructions, which are are very specific use case).

gamerk316 · Feb 17, 2017

My favorite time of the day: MATH TIME:

The performance of the chip was tested in CPU-z with the in-built bench utility. This utility helps evaluate the single and multi-threaded performance of the processor running on the PC. The Ryzen 5 1600X had a score of 1888 in single-threaded performance bench and 12544 points in multi-threaded performance bench. In the task manager, we can see that the chip has boost clock enabled since it is clocking beyond it’s base frequency (e.g. 3.56 GHz).

Core scaling is meh; You'd expect a theoretical 1888 * 12 = 22656, but get a 12544 (55%). If we ignore SMT for a moment, we get an expected 1888 * 6 = 11328, which implies gain of about 11% assuming perfect scaling. That's...not great.

For comparison purposes, we used a Core i5-7600K running in our test rig and loaded the same CPU-z version (v1.78.1 x64). The quad core (non hyper-threaded) chip achieved a score of 2130 in single-threaded and 8206 in multi-threaded.

For comparison, core scaling is expected 2130 * 4 = 8520, compared to a 8206 (96%), so CPUz is scaling properly.

Now for IPC: We already calculated SMT gains of about 11%, so I'm going to disregard the SMT cores for Ryzen except for subtracting 11% from the initial performance result, and calculate for just the six physical cores. That's as good a comparison I can do here; I have to do this to factor out SMT as much as I can:

Performance = IPC * Clock * Num_Cores

Ryzen:

12544 - (12544 - (12544 * .11)) = IPC * 3.56 * 6
12544 - 1378 = IPC * 21.36
11166 = IPC * 21.36
IPC = 522.75

Kaby:

8206 = IPC * 3.6 * 4
8206 = IPC * 14.4
IPC = 569.86

Kaby comes in about 8.64% faster on a per-core basis, disregarding SMT effects. Which puts Ryzen right around Skylake level performance, give or take.

gamerk316 · Feb 17, 2017

juanrga :

How do you know it's the 8 core variant?

Anyway, I'm going to first compare the two Skylake chips, to figure out how well the benchmark scales:

6900k:
17714 = IPC * 3.2 * 8
17714 = IPC * 25.6
IPC = 691.95

6800k:
13583 = IPC * 3.4 * 6
13583 = IPC * 20.4
IPC = 665.83

So I come out with scaling of just under 4%, close to what we've seen in other benchmarks. So I can safely disregard scaling as a factor here [it's there, but minimal].

Now for Ryzen, which I will consider a 8 core chip for this benchmark, as I didn't consider HTT effects above:

Ryzen:
15085 = IPC * 3.4 * 8
15085 = IPC * 27.2
IPC = 554.6

Or about 18.22% slower then a 6800k, or right back to our Ivy-Haswell level performance.

So two benchmarks, two results. In fact, a LOT of benchmarks show the same behavior right now: Some show Ivy-Haswell performance, while others show competitive with Skylake. I'm suspecting both are true: There's a bottleneck for certain workloads which causes Ryzen to lag behind in certain metrics.

But based on what I'm seeing, "best case" per-core performance seems to be around Skylake, and "worst case" around Ivy-Haswell. Throw in the fact Ryzen has more cores to run off of, then I'd be willing to predict that in best case applications that scale across all CPU cores, Ryzen can match or beat Kaby Lake. Again, "best-best case", but the potential is there.

cdrkf · Feb 17, 2017

gamerk316 :

The poor SMT scaling on the CPUz benchmark backs up my hypothesis that that particular benchmarks pushes the execution resources more than anything else- meaning there is little left to gain from SMT (I would expect the same is true for Intel if you were to look at an i7 for example).

I think what is worth noting here is *if the pricing we're seeing is accurate* (and we're close enough to say it's definitely possible)- AMD are offing *a lot* of CPU for the money.

It actually makes a lot of sense when you take into account Ryzen is CPU only- in terms of die size (and thus overall manufacturing cost) an 8 core Ryzen die isn't really any larger than a 4 core Kaby Lake + IGP. Yes it under cuts Intel's enthusiast igp less 'high end' cpu's, however I think the take away from this is that in reality it's not Ryzen being cheap (size for size Intel sell similar physically sized parts for the money) but rather Intel were milking the enthusiast market for all it was worth

The key thing is that AMD appears to be *close enough* to make 'moar cores' a worthwhile choice- take a slight hit in single thread but gain big in multi threaded software vs price equivalent Intel.

simon12 · Feb 17, 2017

Which chip do people think would go in a budget conscious gaming PC with an RX 480, will an R3 1100 be enough?

cdrkf · Feb 17, 2017

juanrga :

I've had a thought about this (pure speculation here)- however do we know much about how the different frequency / turbo values are defined? I wonder if F_TMT is actually a derivative of F_MAX- whilst F_TST simply defines the default value for F_MAX unless the system determines a new (higher) value based on operating conditions? If F_TMT is defined as a proportion of F_MAX (e.g. F_TMT = 0.9 * FMAX) then as XFR increases the maximum single core turbo the max all core turbo would increase as well. Given the chip monitors power and thermals at a high rate the F_MAX would vary on loading anyway so I don't see any logical reason why it couldn't increase the all core turbo slightly if cooling allows for it (though the all core turbo would always be less than the highest single core frequency)...?

cdrkf · Feb 17, 2017

simon12 :

Personally I've seen a number of modern games that respond well to more than 4 threads- given the rumoured prices I'd suggest looking at the 4 core / 8 thread models if possible.

4 threads are enough *now*, but 6 - 8 threads do show benefits in many of the most demanding games so given the fairly close pricing going for the 4 core / 8 thread R5 might be a better long term investment (although going cheap and upgrading later is always an option).

Discussion AMD Ryzen MegaThread! FAQ and Resources

Distinguished

Distinguished

Splendid

Reputable

Splendid

Splendid

Distinguished

Splendid

Reputable

Distinguished

Distinguished

Distinguished

Reputable

Distinguished

Distinguished

Distinguished

Distinguished

Commendable

Judicious

Glorious

Glorious

Judicious

Splendid

Judicious

Judicious

Share this page