Ryzen 4c/8t benchmark

[r9]

http://wccftech.com/amd-ryzen-4-core-benchmarks-intel-core-i7-7700k/

I was asking for this quite a while.
For one it give us a glimpse in the Ryzen quad cores that will start at $135 and second it takes poor optimization of windows scheduler out of the equation (moving threads between CCX and slow L3 cache inter connect).

Things to be taken into account the i7 7700 is clocked @4ghz and none of the games have any optimization for Ryzen.

The results look promising especially in GTA 5 where Ryzen was suffering quite a lot.
It would have been great if they used even faster memory and did the same test w/o smt.



You can build very capable gaming computer for not a lot of money.
Ryzen 3 $135
B series motherboard bundle with CPU at $69 at microcenter.
RX480 for $144 after rebate from newegg.
The total is less than $350.
That's like almost the same as buying just the i7 7700k,

 

[kruk]

But what if the core config for the quad cores will be 2-2 and not 4-0?

 

[Aenra]

Well.. got nothing to complain about this time around. Same RAMs, same clocks, i'm willing to presume same /GPUsHDDs. Amen.

As to the results, i can only imagine what will happen if we get optimizations plus a proper 4core. We have an interesting future ahead of us ^^
(and one slightly disappointed Aenra for picking this January to buy a Broadwell, but hey, never too late to build a new rig, lol)

But what if the core config for the quad cores will be 2-2

Why would they ever do that? Especially when considering that the design for a 4-0 is already there? :S

 

[r9]

But what if the core config for the quad cores will be 2-2 and not 4-0?

I think that would be way more complicated for AMD.
Even if it happens by then I hope Microsoft will have scheduler in place to properly manage threads.
It still bugs me how come they didn't handle this with Microsoft beforehand, knowing the basics of the architecture of Ryzen is not too difficult to anticipate the issues with thread handling

 

[newtekie1]

Why would they ever do that? Especially when considering that the design for a 4-0 is already there? :S

To improve yields. I can almost guarantee that every Ryzen chip being produced is an 8-core. I'm also pretty certain that this early in the production run there are quite a few of those 8-core chips that have bad cores in them. So if the chip has one bad core(or one bad core per CCX) then a core in each CCX is disabled, and the chip becomes a Ryzen 5 6-core. If the chip has 2 bad cores in a CCX(or up to two bad cores per CCX) then 2 cores per CCX are disabled and the chip gets sold as Ryzen 3 Quad-core. That way you aren't just throwing out the bad silicon.

 

[OneCool]

Interesting.

 

[r9]

To improve yields. I can almost guarantee that every Ryzen chip being produced is an 8-core. I'm also pretty certain that this early in the production run there are quite a few of those 8-core chips that have bad cores in them. So if the chip has one bad core(or one bad core per CCX) then a core in each CCX is disabled, and the chip becomes a Ryzen 5 6-core. If the chip has 2 bad cores in a CCX(or up to two bad cores per CCX) then 2 cores per CCX are disabled and the chip gets sold as Ryzen 3 Quad-core. That way you aren't just throwing out the bad silicon.

Hopefully quads and hex will become so popular that will make AMD disable functional cores to fill the demand and give us unlock possibility like with Athlon II.
I remember a good times buying athlon ii x3 for like $50 and unlocking it to quad core with L3 cache which made it Phenom II that went for $150.

 

[Aenra]

To improve yields

Dumb probably, but i honestly thought they'd have a new line for the quad cores, like Intel did. Thanks for grounding me, i occasionally need it

 

[OneCool]

To improve yields. I can almost guarantee that every Ryzen chip being produced is an 8-core. I'm also pretty certain that this early in the production run there are quite a few of those 8-core chips that have bad cores in them. So if the chip has one bad core(or one bad core per CCX) then a core in each CCX is disabled, and the chip becomes a Ryzen 5 6-core. If the chip has 2 bad cores in a CCX(or up to two bad cores per CCX) then 2 cores per CCX are disabled and the chip gets sold as Ryzen 3 Quad-core. That way you aren't just throwing out the bad silicon.

Yep. This is why we had 3 cores a few years ago...Wierd little AMD Frankenstein lol

 

[NinkobEi]

comparing either to a 4 year old i5 or i7 and I think we'll see pretty similar results. everyone knows IPC has been stuck for a long time

 

[newtekie1]

Dumb probably, but i honestly thought they'd have a new line for the quad cores, like Intel did. Thanks for grounding me, i occasionally need it

They probably will eventually. But they'll probably give them a new model number, like 1350X instead of 1300X. Outwardly it will be the same 4c/8t configuration, with the same cache size, and either the same clock speeds or a very minor bump. AMD will claim a small performance boost, but the reality is the performance improvement will really just be from no longer needing to communicate between CCXs anymore. That's just my guess though.

Yep. This is why we had 3 cores a few years ago...Wierd little AMD Frankenstein lol

This system is why we have a lot of the processors we do. And its been used for decades.

I mean, a Core i5 is just a Core i7 with a section of bad Cache. So Intel disabled that bad section and HT, and sell the chip as an i5.

 

[r9]

Dumb probably, but i honestly thought they'd have a new line for the quad cores, like Intel did. Thanks for grounding me, i occasionally need it

Eventually they might, but the first batch will be Ryzen 7 with disabled cores.
That what happened with Athlon II at first they used Phenoms II with disabled cores and L3 then after a while they had true athlon II design.
Having a true quad core Ryzen 3 down the line after they improve the batch provable will be able to hit much higher frequencies then their octacore counter parts.

 

[notb]

Dumb probably, but i honestly thought they'd have a new line for the quad cores, like Intel did. Thanks for grounding me, i occasionally need it

Hmm?
Intel only has quad cores in consumer segment (currently LGA1151), which is probably something like 99% of their sales. So OF COURSE they have a 4C design.
It's the main issue with HEDT platform, for which they have to maintain a separate design and manufacturing line, but with very little sales volume in mind - hence the relatively high pricing.

At this point it's almost sure that all Ryzen chips will be built on the same design (8C, XFR possibility and so on) with certain limits applied.
This is great for AMD, because they can offer the 8C/16T with a price tag we haven't been used to.

And think about the elasticity. The market will be almost saturated with high-end 8C before they release the 4C models, but at this point they don't have to dump the manufacturing line or change anything significant. Just think about the savings here.

Basically, at some point one of the companies will get brave enough (which makes AMD a more probable side), to sell you a CPU that is defined remotely. IMO that would be excellent. You would buy a basic model, run it as e.g. Ryzen 5 for a while, and if you decide that you'd prefer a 7, you could just pay the difference - not replace the chip.
This could be done today by both AMD and Intel. IMO the only thing stopping that are the customers.

 

[Vario]

Wow I am impressed. Those results are pretty good, nearly same as Intel at same clock speeds. Only problem is, can the Ryzen overclock to 5.0Ghz?

Given the 4C/8T Ryzen will be so inexpensive, its a real bargain.

 

[crazyeyesreaper]

What will be interesting is how the 4c 8t Ryzen chips are actually setup if its 2 cores 4 threads spread across 2 CCX complexes to reuse silicon that would actually result in possibly more L3 cache for the 2 cores to access. Due to how Ryzen is setup the L3 is 8mb for 4 cores. However if AMD offers a quadcore SMT chip based on 8 core Ryzen then 2x 2 core CCX and 16mb of L3 itcould prove to be an interesting CPU.

 

[mroofie]

So besides amd friendly tombraider... Its more or less the same as the big fellows

I think we should wait for q2

 

[kruk]

What will be interesting is how the 4c 8t Ryzen chips are actually setup if its 2 cores 4 threads spread across 2 CCX complexes to reuse silicon that would actually result in possibly more L3 cache for the 2 cores to access. Due to how Ryzen is setup the L3 is 8mb for 4 cores. However if AMD offers a quadcore SMT chip based on 8 core Ryzen then 2x 2 core CCX and 16mb of L3 itcould prove to be an interesting CPU.

Hardware.fr did a 4+0 vs 2+2 test: http://www.hardware.fr/articles/956-24/retour-sous-systeme-memoire-suite.html (graphs won't work in google translate)
The additional L3 cache in a 2+2 configuration currently doesn't seem to help. Interestingly, it also looks that the 4+0 config can't access the whole L3 cache, and a portion of the cache is reserved? We will see if scheduler patches bring some improvement for low threaded games (games that work great on an i3), but in well threaded games (i7 works much better) the 2+2 configuration will definitely suffer.

 

[RejZoR]

To improve yields. I can almost guarantee that every Ryzen chip being produced is an 8-core. I'm also pretty certain that this early in the production run there are quite a few of those 8-core chips that have bad cores in them. So if the chip has one bad core(or one bad core per CCX) then a core in each CCX is disabled, and the chip becomes a Ryzen 5 6-core. If the chip has 2 bad cores in a CCX(or up to two bad cores per CCX) then 2 cores per CCX are disabled and the chip gets sold as Ryzen 3 Quad-core. That way you aren't just throwing out the bad silicon.

It's how all CPU's and GPU's are made basically. Only the top models have to be in perfect condition to qualify for sales.

 

[Frag_Maniac]

http://wccftech.com/amd-ryzen-4-core-benchmarks-intel-core-i7-7700k/

The worrisome thing here, is they're trying to simulate a 1500X by merely disabling 4 cores. We don't even know if the 1500X will be stable at 4GHz yet. It's base clock is only 3.5, and their 8 core chips don't OC well.

 

[heky]

The worrisome thing here, is they're trying to simulate a 1500X by merely disabling 4 cores. We don't even know if the 1500X will be stable at 4GHz yet. It's base clock is only 3.5, and their 8 core chips don't OC well.

Sure, but there is no same arhitecture chip that i know of where the unlocked 4 core chip would clock less than an unlocked 8 core chip. And since their 8 core chips clock at around 4.0 to 4.2 ghz, the 4 core chips are highly unlikely to clock lower. I think it is safe to say they will clock higher.

 

[RejZoR]

The worrisome thing here, is they're trying to simulate a 1500X by merely disabling 4 cores. We don't even know if the 1500X will be stable at 4GHz yet. It's base clock is only 3.5, and their 8 core chips don't OC well.

"8 core chips don't OC well." Well, it's 8 core, 16 threads CPU. It's not suppose to OC well because it has that many cores... Intel has the exact same problem. Sure, they OC tiny bit better, but with so many cores, you'll be pumping out insane heat and voltages...

 

[silentbogo]

I'm pretty sure that if we see R3 and R5 CPUs rated at near 65W TDP, then most likely they are R7 with disabled cores. Though, given the current state of "problems" I'm not worried either way.
Decent batch of single-CCX quadcore chips will most likely be packaged as high-performance 4c/8t laptop CPUs with a maximum 35W TDP. I'd love to see those in action.

 

[kruk]

Sure, but there is no same arhitecture chip that i know of where the unlocked 4 core chip would clock less than an unlocked 8 core chip. And since their 8 core chips clock at around 4.0 to 4.2 ghz, the 4 core chips are highly unlikely to clock lower. I think it is safe to say they will clock higher.

We already know this:

So at release, we might not get higher clocked low core CPUs, but this doesn't mean that later there wont be a 1520(X), 1550(X) and so on with a base of 4.0 GHz+:

 

[r9]

We already know this:

View attachment 85142

So at release, we might not get higher clocked low core CPUs, but this doesn't mean that later there wont be a 1520(X), 1550(X) and so on with a base of 4.0 GHz+:

+1
First batch of quads will be the ones that didn't make the cut for octa and hexa core, lower binded.
And if AMD gave us 4.2+ quad it will kill the r7 in gaming and will cost only $135-200. I don't think is what they want to do. Profit margin is much higher for Ryzen 7.
Don't get me wrong I think dedicated silicon for r3 and 95w tdp would produce 4.5ghz chip. Which would be awesome.

 

[r9]

https://www.google.com/amp/amp.hoth...-status-dismisses-windows-10-thread-scheduler

Looks like windows scheduler doesn't need patch. So we not gonna see quick fix.
Setting the Power mode from Balanced to Performance should help.