Simulated AMD Ryzen 5 Series Chips as Fast as Ryzen 7 at Gaming

[btarunr]

It's not rocket science to simulate smaller upcoming Ryzen series chips when you have a Ryzen 7 1800X. By disabling two out of its eight cores and adjusting its clock speeds, TechSpot simulated a Ryzen 5 1600X processor. While the Ryzen 5 1600X was a near-perfect simulation by TechSpot, the 1500X isn't entirely accurate. AMD is carving out the 1500X by disabling an entire CCX (quad-core complex), leaving the chip with just 8 MB of L3 cache, disabling four cores on the 1800X still leaves the full 16 MB L3 cache untouched. The Ryzen Master software lets you disable 2, 4, or 6 cores, but not specific cores, so it's entirely possible that disabling 4 cores using Ryzen Master turns off two cores per CCX. Nevertheless, the gaming performance results are highly encouraging.

According to the gaming performance figures for the simulated 1600X six-core and 1500X quad-core Ryzen chips put out by TechSpot, the 1600X barely loses any performance to the 1800X. Today's AAA PC games have little utility with 8 cores and 16 threads, and you'll hardly miss the two disabled cores when gaming on a 1600X powered machine. The simulated 1500X loses a bit more performance, but nothing of the kind between the quad-core Intel Core i7-7700K and the dual-core i3-7350K. When paired with a GeForce GTX 1080 Ti in "Mafia III," for example, you lose 12.8% performance as you move from the $499 1800X to the $189 1500X (simulated); but you lose 35% performance as you move from the $329 i7-7700K to the $189 i3-7350K. Find more interesting results in the source link below.



View at TechPowerUp Main Site

 

[phanbuey]

in other words... they are meh

sorry had to... love my 1800x

 

[btarunr]

in other words... they are meh

In yet other words, i5-7600K and i3-7350K are screwed.

 

[phanbuey]

In yet other words, i5-7600K and i3-7350K are screwed.

bout time... competition! but lets see if it's CCX vs cross-ccx binning.. and if the cache makes a difference.

enabling two cores per ccx means full cache no?

 

[btarunr]

enabling two cores per ccx means full cache no?

Yup, but since specs sheet says 1500X has just 8 MB L3, they're disabling an entire CCX.

 

[DeathtoGnomes]

sounds like a goofy way to manipulate L3 cache. Simulation like is means its only speculation and hype, a bit too close to fanboism.

 

[InVasMani]

Why not simply setup "Process Lasso" to only use every other Ryzen cores during restraint and further more setup all the non gaming processes to operate on the L3 cache CCX cores? That way all the less critical backround cores run on the slower cache by default and during restraint games run exclusively off of the faster L1 + L2 cache and cores though while not under restraint still benefit from the other cores.

 

[Manu_PT]

In yet other words, i5-7600K and i3-7350K are screwed.

I laugh so much when only 3 weeks after the failure (gaming wise), people still say things like this. Ahah.

Meanwhile someone will post and say these chips are ideal to gaming instead of the R7, while before, R7 was "DA FUTURE BRO, optimization!!"

Absolutely love it xD

 

[theGryphon]

Yup, but since specs sheet says 1500X has just 8 MB L3, they're disabling an entire CCX.

So the whole 2+2 thing was BS afterall, and it must be 4+0? I suppose that's good news, unless they found a way to hide half of the cache from the CCX's, and it can still be 2+2... I guess we'll know for sure only after they're out.

 

[sweet]

So the whole 2+2 thing was BS afterall, and it must be 4+0? I suppose that's good news, unless they found a way to hide half of the cache from the CCX's, and it can still be 2+2... I guess we'll know for sure only after they're out.

I guess the author of this article don't pay enough attention to the R5 line up.

1500x has 16 MB L3, so it's 2+2 with 2 CCX.

1400 has only 8 MB L3, indicating that a full CCX is disabled and 4+0.

 

[InVasMani]

I laugh so much when only 3 weeks after the failure (gaming wise), people still say things like this. Ahah.

Meanwhile someone will post and say these chips are ideal to gaming instead of the R7, while before, R7 was "DA FUTURE BRO, optimization!!"

Absolutely love it xD

The R7 is still better than the R5 overall, but the R5 is better dollar value and we all know tech sites generally overstate and understate performance metrics a touch the raw numbers are what they are, but the conclusions are rather intentionally opinionated sometimes overly so. The R7 is still really new and far from a lemon though it's not a gaming juggernaut per say either however it was never expected to dominate in every metric against Intel as it is AMD came out and overachieved really what people were anticipating and expecting and that's a good thing finally a bit of fight in them competition is good. The R5 1600X looks like really solid value.

I'm still not sure how the L3 on Ryzen works with the CCX and cores I know that it's slower and lags comparatively speaking, but I don't know if that lags the entire CPU or just the processes traversing the L3 cache IE can you manually setup core affinity and priority for backrounds tasks to run on the L3 cache while the L1 and L2 are reserved for games?? Process Lasso allows for lots of customization on that level, but I don't what happens once the L3 cache is touched is the entire CPU stuck in limbo or can it do other things with L1/L2 caches on different cores and only what's in L3 cache is being slowed down latency wise?

 

[phanbuey]

I guess the author of this article don't pay enough attention to the R5 line up.

1500x has 16 MB L3, so it's 2+2 with 2 CCX.

1400 has only 8 MB L3, indicating that a full CCX is disabled and 4+0.

Nice!!

 

[Kanan]

AMD said that it will be 2+2 / 3+3 cores activated for Ryzen 1600/1500. Just read through the article, much more worth than "simulated" stuffs on TechSpot. Also some people here seem to never learn, what's so hard to understand about it, that AMD said it? And it's also making some fucking sense, because with 2+2 and 3+3 they can simply use more defective CPUs as with 4+0 and 4+2, it's simple maths. Deactivated L3 (8 MB instead of 16 MB) could simply mean *some* of the CPUs are 4+0, but not all of them, OR the L3 cache could be simply disabled by half. Fruitless speculation by TechSpot.

Anyway, we will see soon.

 

[ShurikN]

The 1500X makes the equally priced 7350k look like shit. Plus you get a remarkable stock cooler.

 

[bistrocrat]

every time I open something about Ryzen R5 leaks or R7 reviews - only thing I can read about (99% of the time) is:
"*in annoying parrot voice* how is 1080, gaming, how is 1080 gaming, how is 1080 gaming"

 

[nemesis.ie]

The UEFI in my Asrock board allows setting e.g. 2+2 or 4+0. So no need to use Ryzen Master (RM is shorter going forward!) and guess what it is disabling.

 

[Dethroy]

Obligatory link to my spreadsheet

 

[noname00]

Am I the only one that would prefer a quad core CPU with only one CCX instead of one with two CCXs that have only two cores each activated?

From what I know, the connection path between the two CCXs works at memory speed, and you get better performance with higher memory speed. Wouldn't this path be removed if only one CCX is used, and and there will be no need for high memory speed? or at least the impact won't be that high.
Also - only one CCX will have better thermal performance, and that might translate in slightly higher overclocking potential.

 

[Dethroy]

Am I the only one that would prefer a quad core CPU with only one CCX instead of one with two CCXs that have only two cores each activated?

From what I know, the connection path between the two CCXs works at memory speed, and you get better performance with higher memory speed. Wouldn't this path be removed if only one CCX is used, and and there will be no need for high memory speed? or at least the impact won't be that high.
Also - only one CCX will have better thermal performance, and that might translate in slightly higher overclocking potential.

You are right. But it doesn't make sense from a business point of few. Yields ultimately dictate these decisions. It's not like they have a choice...

 

[snakefist]

I would like to see 1 CCX 4-thread CPU. It may overclock differently...

On the other hand, perhaps CPUs with disabled core could have them enabled later, and people would get more cores for less money. I'm not saying it is possible, but it happened before.

My next CPU is likely to be 4/8 Ryzen (and will probably buy cheapest 4/4 for some secondary purposes), but not until benchmarks and comparisons are out. Those *won't* be "ultimate gaming platforms" by any means (and I'm not the "ultimate gamer"), but getting 80% performance of Intel for, say, 60% price is a pretty good deal.

I would also like that 2-cores CPUs become obsolete, the technology is here since early 2000's, it's hardly relevant now and the time for retirement is here. That means I3 and low-cost dual-cores from both manufacturer, they likely can be replaced with 4-cores and only then the *real* core optimisation will come.

 

[bug]

Am I the only one that would prefer a quad core CPU with only one CCX instead of one with two CCXs that have only two cores each activated?

From what I know, the connection path between the two CCXs works at memory speed, and you get better performance with higher memory speed. Wouldn't this path be removed if only one CCX is used, and and there will be no need for high memory speed? or at least the impact won't be that high.
Also - only one CCX will have better thermal performance, and that might translate in slightly higher overclocking potential.

That's what yields maximum performance, yes.
But since this is about defective chips, you'd only get to reuse chips with a fully working CCX.

 

[Donny Stanley]

the 1500X isn't entirely accurate. AMD is carving out the 1500X by disabling an entire CCX (quad-core complex), leaving the chip with just 8 MB of L3 cache, disabling four cores on the 1800X still leaves the full 16 MB L3 cache untouched.

No, the 1400 will be a 4+0 configuration with just 8MB cache, meaning yes, one CCX will be disabled. However, the 1500X features a 2+2 configuration with the full 16MB of cache. This is confirmed by Anandtech here. Furthermore, you can choose to disable an entire CCX in the BIOS in some motherboards (I believe the ASUS Maximus Hero is one).

 

[Kanan]

No, the 1400 will be a 4+0 configuration with just 8MB cache, meaning yes, one CCX will be disabled. However, the 1500X features a 2+2 configuration with the full 16MB of cache. This is confirmed by Anandtech here. Furthermore, you can choose to disable an entire CCX in the BIOS in some motherboards (I believe the ASUS Maximus Hero is one).

Can you actually read?

We have confirmation from AMD that there are no silly games going to be played with Ryzen 5. The six-core parts will be a strict 3+3 combination, while the four-core parts will use 2+2. This will be true across all CPUs, ensuring a consistent performance throughout.

Again and again...

 

[ShurikN]

I'm guessing the 4+0 config will be on the Ryzen 3.

 

[john_]

Waiting for the "New feature/bios/program/ritual unlocks cores in Ryzen 5 and Ryzen 3 cpus" article in a few months.