The Kaveri Pre-Launch Thread (A10-7800 and A10-6800k @3,5 Ghz)

[AtenRa]

1.65V is not stock. To be stock it would have to be a JEDEC standard and actual 1.65V chips produced at the fabs. Its not any.

Im sorry but you are talking BS,

Stock = as it was made by the manufacturer.
For example Kingston sells 2800MHz 1.65V 8GB DDR-3 kit with lifetime Warranty. That is stock, you dont get 1600MHz and you OC to 2800MHz, you buy the ram at 2800MHz completely evaluated and tested to work at that frequency and voltage by the manufacturer.

Now, could you give it a rest about memory ???

 

[monstercameron]

this has devolved into a debate of semantics[like quiet and uber mode], intel guys like to brag about what is on the market now and ddr3-2400 can be had for ~$70 stock or not, and the price premium is negligible atleast in the us...lowest on newegg ddr3-1600 is $64, i payed ~$70 for ddr3-2400

ddr3-1600 2x4gb kits
http://www.newegg.com/Product/Produc...2%20x%204GB%29

 

[ShintaiDK]

Im sorry but you are talking BS,

Stock = as it was made by the manufacturer.
For example Kingston sells 2800MHz 1.65V 8GB DDR-3 kit with lifetime Warranty. That is stock, you dont get 1600MHz and you OC to 2800MHz, you buy the ram at 2800MHz completely evaluated and tested to work at that frequency and voltage by the manufacturer.

Now, could you give it a rest about memory ???

Stock applies a standard. 1.65V is not a standard. Neither is anything above 2133Mhz. The chips used on the DIMM is still 1.35V or 1.5V.

 

[Vesku]

Uh, it is stock for the vendor you are buying from. It may violate the JEDEC standard but companies do offer out of the box 1.65V DDR3 2133 some with 'limited' lifetime warranties. I've never seen someone try to imply that the term 'stock' when referring to computer components means industry standard rather than the usual definition of unaltered out of the box settings and usage. The term comes from the "retail inventory" definition of stock as in you are running it how it would run for anyone who bought it unaltered.

A 'stock' A10-7850K would be running at AMD's stated settings, for example. Really taking this thread off topic with such an odd personal definition of a commonly used term.

monstercameron is right, btw, can get 'stock' 2x4GB DDR3 2400 1.65V in the ~$80 range. Considering the Samsung 1.35V green modules seem to be unavailable at MSRP, I'd be tempted to run 1.65V 2400 over 1.5V 2133 for ~$30 more. IF there is some user confirmation that it works in the specific motherboard being used.

 

[ShintaiDK]

Limited lifetime is a fun definition.

The lifetime in question is rarely the lifetime of the purchaser, but rather the length of time the manufacturer continues to make the product. Once a product is discontinued, the length of time the manufacturer remains liable under the warranty is specified in the warranty documentation.

Essentially means warranty is gone when its EOLed.

 

[pcsavvy]

Wow, I thought this was about a Kaveri pre-launch and the speculations that abounded until AMD released the actual product. Users can actually run them on their home computers to see if expectations/speculations matched/exceeded/lessened by reality. Not nitpicking about memory and warranties. Warranties are only as good as the company that stands behind them.

 

[monstercameron]

Wow, I thought this was about a Kaveri pre-launch and the speculations that abounded until AMD released the actual product. Users can actually run them on their home computers to see if expectations/speculations matched/exceeded/lessened by reality. Not nitpicking about memory and warranties. Warranties are only as good as the company that stands behind them.

hmmm the kaveri prelaunch thread

 

[Vesku]

Limited lifetime is a fun definition.




Essentially means warranty is gone when its EOLed.

Yes, which will be the case whether your 'limited' lifetime RAM is JEDEC spec or not.

Is there any reason to expect Kaveri will be more restrictive in what memory it will tolerate than Trinity/Richland? Can we expect Kaveri to be a bit more 'finicky' based on its GF 20nm node and resulting lower max CPU clocks or is that not relevant at all when it comes to the on die memory controller?

 

[Ajay]

Stock applies a standard. 1.65V is not a standard. Neither is anything above 2133Mhz. The chips used on the DIMM is still 1.35V or 1.5V.

Stock != Standards based. Never has. Seriously, why are you harping on this, it's pointless.

 

[Abwx]

hmmm the kaveri prelaunch thread

People s memory is definitly short...

 

[Markfw]

1.65V is not stock. To be stock it would have to be a JEDEC standard and actual 1.65V chips produced at the fabs. Its not any.

1.65 v is stock if its the manufacturers spec. I have a lot of 1.65v stock memory.

But this has nothing to do with the Kaveri Launch, stay on topic, or the next divergence will be an infraction.

 

[pcsavvy]

:thumbsup:

 

[PPB]

Ok, so back on topic. What secret sauce is behind having a 2x larger IMC than Richland?

Is it that big because of the failed attempt at propelling GDDR5M? Is it actually a DDR3/DDR4 IMC, saving the latter for a future platform update (and thus screwing whoever bought FM2+ thinking it would last)?


The GDDR5m sounded fancy for embedded/mobile solutions, while DDR4 sounds reasonable for desktop, where you can actually populate four slots of RAM. Either way AMD made a decision and somehow that IMC is bigger that it should (for DDR3 only, that is).

 

[Vesku]

My guesses regarding the memory interface section of the die map: either intentionally fudged or quad channel DDR3 support. Perhaps will get some quality analysis along the lines of that mysterious AMD die shot thread which we now know would have to be Excavator or beyond (was some speculation it was Steamroller at the time).

 

[Shivansps]

Quad channel makes sence, but will be not getting it on desktop, maybe some OEM will go for a itx with quad ddr3 with a mobile kaveri.

 

[Rickyyy369]

This was posted over on OCN & SA forums:

http://foro.noticias3d.com/vbulletin/showthread.php?t=421821

7850k vs 5800k, overclocked & stock. Guy said it took 1.5v to get to 4.55ghz on the 7850k. The gaming benchmarks at the bottom were done with a 7950 since apparently he cant find any drivers for the IGP.

 

[NostaSeronx]

Quad channel makes sence, but will be not getting it on desktop, maybe some OEM will go for a itx with quad ddr3 with a mobile kaveri.

It's for dense servers which are BGA and it supports DDR4 as well.

AMD currently can only access:
DDR4-1600, 4GB - 16GB, 11-11-11 per DIMM
DDR4-1866, 4GB - 16GB, 13-13-13 per DIMM
DDR4-2133, 4GB - 16GB, 15-15-15 per DIMM.

DDR4 without a switch:
4 DIMMs * 16GB = 64GB
DDR4 with a switch:
16 DIMMs * 16GB = 256GB

It expected by 2015 though that the speeds will hit 2866 and higher.

 

[Abwx]

This was posted over on OCN & SA forums:

http://foro.noticias3d.com/vbulletin/showthread.php?t=421821

7850k vs 5800k, overclocked & stock. Guy said it took 1.5v to get to 4.55ghz on the 7850k. The gaming benchmarks at the bottom were done with a 7950 since apparently he cant find any drivers for the IGP.

Nice find, if thoses numbers are genuine IPC is improved by 18.9% on FritzChess, an integer bench, and 14.9% on CB R11.5 wich is mainly FP intensive.

 

[SPBHM]

some of the gains are quite nice (like the stock 7850K pretty close to the 5800K at 4.45GHz in many things), and if aida cache test is accurate, that's a pretty big cache performance improvement

 

[DaZeeMan]

Another link to some Kaveri related benches. Focusing on the new Plextor SSD though, not the Kaveri specifically.
http://forums.hardwarezone.com.sg/storage-clinic-119/%5Bplextor%5D-fast-enuff-ya-its-plextor-afterall-4536563.html

 

[inf64]

In aida64 there will be "benchmark" (pure x86) that will put stock 7850K around the same level as 4 module 8150 :whiste:. And it will not be about AES.

From that website link:

IPC in x86 (@ OCed speeds):
C11.5 SR is 13.5% faster
C15 SR is 7.6% faster
Fritz chess SR is 17% faster

IPC in games (720p for making the CPU matter more; @ OCed speeds; tester used OCed 7950 @ 1100/1500Mhz ! ) :
Grid2 SR is 8.6% faster
RE5 SR is 39% faster!
Metro LL SR is 8% faster
Hitman Absolution SR is 33% faster! (even in 1080p the difference is 31%)
Batman AC SR is 12% faster
Skyrim no 720p data but in 1080p SR is around 6% faster (stock 7850K is faster than OCed @ 4.5Ghz PD).

In other benchmarks the tester didn't post number yet but he posted this:

De serie el luxmark 553 puntos. Mi 5800K da 455 con 1151Mhz y 4450Mhz de cpu y 2544Mhz de memoria. Pffff

P.D el i3 + 7750 ddr5 da 528 puntos.

P.D2 en raytrace y sph fluid simulation casi el triple que trinity

Yup! But i have no drivers for this igp.

I used an hd 7950 @ 1100/1500mhz to see bottleneck evolution.

Stock 3700mhz cpu. Overclock 4560mhz with 1,5v. Really really good temps.

Imc works right up to 2544mhz like trinity.

 

[itsmydamnation]

some of the gains are quite nice (like the stock 7850K pretty close to the 5800K at 4.45GHz in many things), and if aida cache test is accurate, that's a pretty big cache performance improvement

likely from the fixed WCC, remember bulldozer/etc is a mostly write through for the L1D which means L1D write becomes mostly L2 write limited.

 

[mikk]

Nice find, if thoses numbers are genuine IPC is improved by 18.9% on FritzChess, an integer bench, and 14.9% on CB R11.5 wich is mainly FP intensive.

Not that special for the better core scaling. The 1 core instruction performance probably is still behind K10.5.

 

[itsmydamnation]

Not that special for the better core scaling. The 1 core instruction performance probably is still behind K10.5.

purely workload dependant clock for click piledrive wins some K10 wins some.


edit:lol wrong cpu.....
http://www.anandtech.com/bench/product/203?vs=699

 

[monstercameron]

not bad kaveri, not bad, there does seem to be a good amount of OC overhead.
@rickyyy369 from your link with the same ram being used on both why do you think kaveri has more latency for being an improved imc?