AMD Bristol/Stoney Ridge Thread

[SPBHM]

I'm curious to see how memory and the IGP overclocks, for $99 it's somewhat interesting but, once you want better gaming performance I think you will need a new CPU+GPU, while if you buy a G4600 later on you kind of just need a new VGA, and the HD 630 might not be that terrible compared to it I don't think...

 

[amd6502]

If you can overclock the memory controller (NB) to run 2667 DDR4 memory then my guess is the iGPU can get another 10+ percent performance gain with a corresponding GPU overclock.

Sure, a Ryzen 3 1300X plus RX 460 would be a much better performer and could handle higher rez, but it's also well over a whole price notch up. Adding a discrete GPU that doesn't suck is roughly about the cost of the A12 ($99)! So it is a very compelling ultra budget proposition.

 

[NostaSeronx]

AM4 supports both Bristol and Raven. So... if cheap on budget, but will get money later on. Then, this option is good and if Bristol works, then Raven works. Even then Bristol Ridge supports AVX2, where Kaveri/GV does not. That is an upgrade right there, esp. if AVX2 and Intel C++ is used. (AVX2-Intel doesn't have special paths yet)

There is also the Bristol -> 8th Mainstream -> 9th Mainstream path which will be the successor to the cancelled AM1: Beema(Bhavani) -> Carrizo-L -> BristolR-L path.

 

[Abwx]

http://www.anandtech.com/show/11669/amd-releases-bristol-ridge-to-retail-am4-gets-apus

Spoiler: IMG

 

[dark zero]

A budget chip that cost decently, but being beaten hard by Intel offerings...

Except in GPU which they retook the crown again. Intel iGPU is BAD, and needs hard a new brand revamp. If they needs, they should buy Power VR GPU patents....

 

[Abwx]

The more the threads the more BR will show some advantage over Celerons, if we are to believe Computerbase.de multitasking tests with just two apps those latter CPUs are hopeless, and that s without accounting for their weak IGP and restricted ISA.

https://www.computerbase.de/2017-01...m-multitasking-test-the-witcher-3-plus-winrar

https://geizhals.de/?cat=cpuamdam4&xf=12099_Desktop~596_Bristol+Ridge~820_AM4

Prices are decent overall, this will make a PC that has some gaming usability at compressed prices...

 

[dark zero]

So... BR is ending to be a decent deal against the Celerons and the non HT Pentiums?
Oh boy... seems that the Dual Core Era is about to end.

 

[VirtualLarry]

Yeah, dual-core is pretty-much nearly dead at this point, going forward.

 

[jpiniero]

Yeah, dual-core is pretty-much nearly dead at this point, going forward.

For general purpose usage, dual core is all you need really. That's sort of Intel's problem in a nutshell; we are way past overkill for GP usage.

 

[Abwx]

That s not a matter of cores but of threads, a 4 threads CPU is a minimal requirement as well as the necessary cache sizing, and i m not pointing this aspect because BR support four threads, as Computerbase.de tests show that AMD s APUs suffer greatly with two simultaneous tasks, albeit much less than Intel s 2C/2T and competitively with the i3s.

It s worth noticing that the FX is on top in this matter thanks to its L3 cache, and i suspect that a FX4300 is better than Kaveri FI due to its 4MB L3, not talking of the FX4350 wich has the full 8MB L3.

 

[Insert_Nickname]

That s not a matter of cores but of threads, a 4 threads CPU is a minimal requirement as well as the necessary cache sizing, and i m not pointing this aspect because BR support four threads, as Computerbase.de tests show that AMD s APUs suffer greatly with two simultaneous tasks, albeit much less than Intel s 2C/2T and competitively with the i3s.

It s worth noticing that the FX is on top in this matter thanks to its L3 cache, and i suspect that a FX4300 is better than Kaveri FI due to its 4MB L3, not talking of the FX4350 wich has the full 8MB L3.

This is otherwise known as the module penalty. Bulldozer and Piledriver both loose about 20% performance when you load up the second thread in a two thread module. Steamroller fixed this somewhat, and Excavator no longer suffers from it.

With regards to cache, the loss of 2MB L2 and having no L3 penalises Excavator somewhat. But the 2MB L2 that's there is quite high performance. Coupled with fast, low latency RAM Excavator performance is quite good for what it is.

 

[Abwx]

This is otherwise known as the module penalty. Bulldozer and Piledriver both loose about 20% performance when you load up the second thread in a two thread module. Steamroller fixed this somewhat, and Excavator no longer suffers from it.

With regards to cache, the loss of 2MB L2 and having no L3 penalises Excavator somewhat. But the 2MB L2 that's there is quite high performance. Coupled with fast, low latency RAM Excavator performance is quite good for what it is.

Module penalty comes on top of what is actually threads contention, if we look at the posted tests.

FI the i7 2600K, wich is a potent multitasking CPU better than any i5 generation, see its CB R15 ST score collapse from 119 to a meager 69 once an integer load use all cores.

If CB is using all cores the resulting MT score is only 35% lower despite the integer load using all cores simultaneously, in comparison the i5 6600K CB MT score is down by 68% while its ST score is 77 down from 165.

https://www.computerbase.de/2017-01...m-multitasking-test-the-witcher-3-plus-winrar

We can see that the integer load execution efficency is good on all CPUs when an FP load is fired at the same time but this latter is efficently executed (with said integer load also running) only on i7s and FXs.

 

[Insert_Nickname]

Module penalty comes on top of what is actually threads contention, if we look at the posted tests.

FI the i7 2600K, wich is a potent multitasking CPU better than any i5 generation, see its CB R15 ST score collapse from 119 to a meager 69 once an integer load use all cores.

If CB is using all cores the resulting MT score is only 35% lower despite the integer load using all cores simultaneously, in comparison the i5 6600K CB MT score is down by 68% while its ST score is 77 down from 165.

https://www.computerbase.de/2017-01...m-multitasking-test-the-witcher-3-plus-winrar

We can see that the integer load execution efficency is good on all CPUs when an FP load is fired at the same time but this latter is efficently executed (with said integer load also running) only on i7s and FXs.

I see your point. I wasn't aware of those test, so thanks. I'm only sorry they didn't include EX.

Anyway, EX has most merit as a low budget contender, and not much else. IGP is decent for the price, but EX isn't suited as a gaming CPU, so its a trade-of. Mainstream will still have to wait for Raven Ridge.

 

[amd6502]

fellow dozer fans, I may no longer be able to hold off too much longer from the idea of a zen upgrade. I may try one before trying the a12 which will hit the shelves within less than two week.

From looking at the line-up it seems the 1600x or possibly the 1600 is by far the best bang for the $.

what is known about the popularity of 1600's versus 1700-1800x? Do the hexacores outsell the octacores, or vice versa?

The yields are supposedly very very good, but I'm thinking their binning process is more frequency oriented than true defect oriented.

If they throw out the bottom 10% performers for frequency more than half of the dies would cut it for hexacores while less than half, 43% would cut it for octacores (1700's and higher):

0.9^6=53% and 0.9^8=43%

And 2/3 ~ 0.9^4 would cut it for the creme de la creme of quadcores (1500x).

Now the numbers for tossing bottom 8% would be slightly different (51%,61% and 72%).

If dropping the bottom 14% only 30% make it as octacores and 40% as hexacores.

The boost frequency among 1500 and higher Ryzens ranges from 3.7 to 4, so they are using multiple cutoff points.

If I had to make a guess I would say hexacore production and sales outnumber octacore. Whether wrong or right, it definitely seems close to me. What's your guess?

 

[wahdangun]

fellow dozer fans, I may no longer be able to hold off too much longer from the idea of a zen upgrade. I may try one before trying the a12 which will hit the shelves within less than two week.

From looking at the line-up it seems the 1600x or possibly the 1600 is by far the best bang for the $.

what is known about the popularity of 1600's versus 1700-1800x? Do the hexacores outsell the octacores, or vice versa?

The yields are supposedly very very good, but I'm thinking their binning process is more frequency oriented than true defect oriented.

If they throw out the bottom 10% performers for frequency more than half of the dies would cut it for hexacores while less than half, 43% would cut it for octacores (1700's and higher):

0.9^6=53% and 0.9^8=43%

And 2/3 ~ 0.9^4 would cut it for the creme de la creme of quadcores (1500x).

Now the numbers for tossing bottom 8% would be slightly different (51%,61% and 72%).

If dropping the bottom 14% only 30% make it as octacores and 40% as hexacores.

The boost frequency among 1500 and higher Ryzens ranges from 3.7 to 4, so they are using multiple cutoff points.

If I had to make a guess I would say hexacore production and sales outnumber octacore. Whether wrong or right, it definitely seems close to me. What's your guess?

x core cpu usually better binned and can reach 4 Ghz, than non-x.

"
The Silicon Lottery released their binned Ryzen CPUs today and included the following statistics in their product pages. This gives us more of an idea on the differences among the lineup in terms of overclocking potential and should help us set our expectations. AMD has clearly squeezed as many MHz out of their CPUs as the process allows.

Ryzen 7 1700
93% reach 3.8GHz @ 1.376V
70% reach 3.9GHz @ 1.408V
20% reach 4.0GHz @ 1.440V

Ryzen 7 1700X
100% reach 3.8GHz @ 1.360V
77% reach 3.9GHz @ 1.392V
33% reach 4.0GHz @ 1.424V

Ryzen 7 1800X
100% reach 3.8GHz (assumed)
97% reach 3.9GHz @ 1.376V
67% reach 4.0GHz @ 1.408V
20% reach 4.1GHz @ 1.440V

Note:
Their test setup used the Realbench stress test for 1 hour on an Asus Crosshair VI, cooled by a Corsair H105 with 2 X 8GB of 2400MHz CL15 RAM."

 

[Jan Olšan]

what is known about the popularity of 1600's versus 1700-1800x? Do the hexacores outsell the octacores, or vice versa?

https://www.mindfactory.de/Hardware/Prozessoren+(CPU)/AMD+Desktop/Sockel+AM4.html

Look here for sales breakdown of the individual AM4 chips (or other CPUs too). It's just one store but it seems to match Amazon best sellers list.

 

[amd6502]

Thanks, those numbers are very helpful. It's too close to call (Octas: ~9.6k; Hexa 10.5k), but given that Ryzen 7 had a running start it seems Hexas are outselling octacores.

1.8k 1800x
3.4k 1700x
4.4k 1700
3.7k 1600x
6.8k 1600 (has the lowest boost of the bunch, 3.6)
1.6k 1500x

The boxed edition of BR A-series have just arrived with Ryzen 3 and have just barely sold by the handful.

The numbers are showing Octacores would be much more common to produce (than hexas that don't make the cut as octas), so it seems they must have a way to make a dispropotionate number of Octacores go to server use (Zepplin) by gathering statistics on the wafers or somehow else.

I'm thinking 1600x with its nice boost frequency is pretty rare compared to 1600; it also is the best value for gamers who like more than 4 cores. I think a 1600 on sale will do for me.

 

[VirtualLarry]

https://www.newegg.com/Product/Product.aspx?Item=9SIA2F86570539

BR is here!

 

[AtenRa]

I dont see the A12-9800E 35W TDP or any other BR in Newegg.

 

[amd6502]

Yes availability looks to be close to zero at the moment, but at least it's a sign that they're probably on the way and arriving before back to school season.

 

[formulav8]

AMD is not sampling to reviewers I guess? Would like to still see a full review of the 9800 model.

 

[AtenRa]

AMD is not sampling to reviewers I guess? Would like to still see a full review of the 9800 model.

yes me to, i would really like to see a full review for both 65W and 35W models.

 

[Asterox]

This is otherwise known as the module penalty. Bulldozer and Piledriver both loose about 20% performance when you load up the second thread in a two thread module. Steamroller fixed this somewhat, and Excavator no longer suffers from it.

With regards to cache, the loss of 2MB L2 and having no L3 penalises Excavator somewhat. But the 2MB L2 that's there is quite high performance. Coupled with fast, low latency RAM Excavator performance is quite good for what it is.

If we compare Cinebench R15 scores(i added a couple of 4 thread processors for a better comparison), first Bulldozer CPU-s where garbage for a big severe headache.FX-4100 is barely faster than X4 Athlon, and obviously Piledriver is very similar story or a serious headache.

AMD is logicaly gone all in for Zen/Ryzen. But at least one 8 Threads Excavator CPU(on very cheep 28nm) for old AM3+ socket, well that would by a nice last greetings or goodbye.

 

[formulav8]

But at least one 8 Threads Excavator CPU(on very cheep 28nm) for old AM3+ socket, well that would by a nice last greetings or goodbye.

Will definitely not happen. But would be interesting to see, especially with a 'good' L3 cache.

 

[Abwx]

A 14nm shrink would be more than competitive against Ryzen, moreover with a L3 cache as pointed by Formulav8, and also with a correctly sized L2, but on 28nm it would be hopless perf/watt wise.