The Official Kaveri Review Thread (A10-7850K, etc)

[Gloomy]

Stick a fork in it, it's done.

 

[monstercameron]

how criptic...http://www.hardwareheaven.com/revie...g1sniper-a88x-power-use-and-overclocking.html

In our very limited time with the A10-7850K before publishing we ran a quick overclocking test and were easily able to hit 5.0GHz on the A88X using air cooling. More results to follow in future articles...

 

[PG]

I'm shocked that the CPU is still losing against a 3-year old Sandy Bridge processor.

I was thinking the same thing. Previous Anandtech reviews of Bulldozer and Piledriver said that the poor single threaded performance was a huge concern. The Phenom II, Bulldozer, and Piledriver have all had around 1 to 1.1 or so at best for single threaded performance in Cinebench 11.5.
Now this new cpu comes out and it's the same thing! That score needed to be around 1.5 to be on par with Sandy, and that's at the very least. I was even hoping for more. This is a major fail.

 

[meloz]

I was excited about Kaveri, since I have a dual-core Sandy Bridge and I begin to feel that upgrade itch. But after reading Kaveri reviews just one word comes to mind: meh.

The promised CPU improvements are underwhelming. Bulldozer was just rubbish and polishing it year-after-year is not working. It is hardly a compelling upgrade for someone using SNB desktop, to say nothing of Haswell, sans AVX2 even.

For users of SNB or later generation CPU aiming to upgrade their CPU power, it makes no sense to buy into Kaveri. And if they need addition GPU horsepower it makes more sense to slot an AMD GPU card in your old Intel system rather than build an all new Kaveri desktop. Much cheaper, bigger bang for buck.

Kaveri's iGPU is strong (much stronger than Intel's HD lolphics, anyway), but I see two reasons why it will not swing most consumers going for Haswell:

1. While kaveri's iGPU is stronger than Intel's solution, it is not strong enough to satisfy the gamers. They will still need a GPU card to get good framerates at 1080p to play their choice titles. AMD are targeting gamers who need more performance than what Intel iGPU provides, but do not have the budget to build a desktop with GPU card and Haswell processor. How big of a segment is this? Don't these kind of people buy consoles, anyway?

2. This one is personal: with no official open-source linux drivers, Kaveri is no go for me. With binary blobs every upgrade is a gamble, you never know when you run into an 'oops' that cannot be investigated let alone solved because there is no open-source driver in the kernel. AMD prepare this slide:

but "forget" to put any official drivers in the linux kernel. Apparently service to open-source only means programs, forget the underlying OS....it is like one part of company does not know what the other is aiming to do.

Sigh.

Just once, I would like AMD to have a non-retard CEO. Their own leadership is their biggest enemy. About the only positive thing I can say about Kaveri is that it will continue to flame the fire under Intel's bottom to keep improving HD Graphics.

 

[JDG1980]

Not having HSA available on day 1 was a major fail. According to Hardware Canucks, "the software necessary for support isn't ready on AMD's part and compatible applications are non-existent". I understand that AMD only has limited influence over third-party developers (though you'd think they would at least have been lobbying Adobe pretty hard for a while), but not having their own drivers available at launch is utterly absurd. AMD really needs to hire more competent software developers.

 

[unon]

It seems they had put too much on GDDR5m which apparently got lost due to Elpida bankruptcy. Otherwise kaveri would've been a Conroe, at least as far as the igp is concerned. The igp frequencies would've been higher too.

The worst thing is kaveri is too big and the prices are too high(even with bf4(?)). If AMD had just focused on HSA, only a 2channel controller I'm sure kaveri could fit in <180mm^2(or 200mm^2 with 8CU's) and could be sold for <100$ which at least would've been good value at least. I wanted to buy one to work with HSA but the prices seem way too much. A8 7600 doesn't look that bad if it goes under 100$.

 

[9enesis]

(facepalm) IMO , what amd should have done with this arc is: made one 65W APU-k for enthusiasts(aka AMD-fans)with GDDR5 or Qch DDR3 and kept the consumer baseline products with dual chanel memory and TDP from 45W down, because their 65W and 95W parts make no sense whatsoever.
that is about it.

TBH , all of the reviews i've seen so far scream for FAIL AMD.

 

[SPBHM]

HSA is right now what x64 was back in 2003 unfortunately, so you need to look for classic CPU and GPU performance,

I think the 45W version achieved a nice balance, the A8 7600 is he most interesting Kaveri right now,

CPU performance looks competitive enough to the i3 (but still the same old huge single thread performance gap), but not really adequate for the 7850K price, the GPU is nice, but outside of the OpenCL stuff is hard to really see the gain when comparing to the 6800K which is disappointing,

still, it's funny to see the old 5750 (re branded as 6750) still easily beating even the "great" Iris Pro.

unfortunately I don't see this product changing much for AMD at the moment.

lack of DG or talks about memory going higher than dual channel DDR3 was disappointing

still looks bad compared to Pentium + 7730
http://pclab.pl/art56011-7.html

 

[Gloomy]

What are the chances of HSA taking off? Do you guys think Intel will support something similar?

Kaveri looks like garbage to me, but HSA seems interesting. Hopefully it has a future.

 

[monstercameron]

I was excited about Kaveri...
snip
2. This one is personal: with no official open-source linux drivers, Kaveri is no go for me. With binary blobs every upgrade is a gamble, you never know when you run into an 'oops' that cannot be investigated let alone solved because there is no open-source driver in the kernel. AMD prepare this slide:

snip

just WTH are you on about? there are linux drivers available, and the FOSS driver is there, it's just a matter of enablement, It's the same RadeonSI driver for all gcn cores. This is bs, it is the same thing with intel it took a little while for even phoronix to bring up linux on the baytrail t100...

 

[monstercameron]

Not having HSA available on day 1 was a major fail. According to Hardware Canucks, "the software necessary for support isn't ready on AMD's part and compatible applications are non-existent". I understand that AMD only has limited influence over third-party developers (though you'd think they would at least have been lobbying Adobe pretty hard for a while), but not having their own drivers available at launch is utterly absurd. AMD really needs to hire more competent software developers.

the new driver is set to release soon, and AMD does employ man competent software devs, quit ya trollin and go somewhere else.

 

[Fjodor2001]

From the Anandtech review conclusion:

The reality is quite clear by now: AMD isn't going to solve its CPU performance issues with anything from the Bulldozer family. What we need is a replacement architecture, one that I suspect we'll get after Excavator concludes the line in 2015.

Some questions regarding that:

1. Does it mean we cannot expect much gains from Excavator either?

2. Is anything known about what AMD's plans are beyond Excavator? Will there be a completely new CPU uarch line (non-Bulldozer based)?

3. If Excavator is the "end of the current CPU uarch line" for AMD, then what is the corresponding "end of current CPU uarch line" for Intel? Will they e.g. make a complete uarch redesign after Broadwell or Skylake, and start a new "line"?

 

[unon]

From the Anandtech review conclusion:

Some questions regarding that:

1. Does it mean we cannot expect much gains from Excavator either?

2. Is anything known about what AMD's plans are beyond Excavator? Will there be a completely new CPU uarch line (non-Bulldozer based)?

3. If Excavator is the "end of the current CPU uarch line" for AMD, then what is the corresponding "end of current CPU uarch line" for Intel? Will they e.g. make a complete uarch redesign after Broadwell or Skylake, and start a new "line"?

I think AMD can come near Intel st and mt wise if they add an alu/agu, fatten the fpu, use lower latency l2 along with l3 and put a trace cache. Intel process advantage cannot be undermined and actually with equal process BD processors should have higher frequency than Core. Making a new CPU might be a waste of resources and they should focus on using HBM(not for whole main memory) ASAP that will be much more important.

edit:Actually with FD-SOI AMD might be able to do better.

 

[NTMBK]

From the Anandtech review conclusion:

Some questions regarding that:

1. Does it mean we cannot expect much gains from Excavator either?

2. Is anything known about what AMD's plans are beyond Excavator? Will there be a completely new CPU uarch line (non-Bulldozer based)?

3. If Excavator is the "end of the current CPU uarch line" for AMD, then what is the corresponding "end of current CPU uarch line" for Intel? Will they e.g. make a complete uarch redesign after Broadwell or Skylake, and start a new "line"?

Leaked roadmaps show Excavator only going up to 65W- even if they improve performance/W again, I don't expect total performance to be better than the 95W Kaveri.

 

[monstercameron]

I think AMD can come near Intel st and mt wise if they add an alu/agu, fatten the fpu, use lower latency l2 along with l3 and put a trace cache. Intel process advantage cannot be undermined and actually with equal process BD processors should have higher frequency than Core. Making a new CPU might be a waste of resources and they should focus on using HBM(not for whole main memory) ASAP that will be much more important.

edit:Actually with FD-SOI AMD might be able to do better.

it is obvious that amd subscribes the the "once bitten twice shy" way of things, they don't want the compete with intel head-on but rather subvert the whole system. Mantle and HSA is proof that AMD would rather try something new than play the long game with intel [surely a losing fight]. The emphasis going forward IMHO will be HSA style compute [all 3 flavors ]

 

[sniffin]

They were stupid to launch with a 95W desktop flagship, when the biggest gains are clearly in the 45W and below range. Nice going AMD, way to make your product look bad

 

[monstercameron]

They were stupid to launch with a 95W desktop flagship, when the biggest gains are clearly in the 45W and below range. Nice going AMD, way to make your product look bad

the only thing that looks bad is pricing, everything else has been improved, especially graphics.

 

[Asterox]

HSA is right now what x64 was back in 2003 unfortunately, so you need to look for classic CPU and GPU performance,

I think the 45W version achieved a nice balance, the A8 7600 is he most interesting Kaveri right now,

CPU performance looks competitive enough to the i3 (but still the same old huge single thread performance gap), but not really adequate for the 7850K price, the GPU is nice, but outside of the OpenCL stuff is hard to really see the gain when comparing to the 6800K which is disappointing,

still, it's funny to see the old 5750 (re branded as 6750) still easily beating even the "great" Iris Pro.

unfortunately I don't see this product changing much for AMD at the moment.

lack of DG or talks about memory going higher than dual channel DDR3 was disappointing

still looks bad compared to Pentium + 7730
http://pclab.pl/art56011-7.html

With regard to the 45W A8-7600(120 dolars), it is Kaveri APU with the bigest increase in CPU Multithread performance no doubt or 600mhz higher defoult CPU frequency+up to 20% beter IPC.

http://forums.anandtech.com/showthread.php?p=35917274&highlight=#post35917274

Hardware-canucks

"When I first opened the box, I couldn&#8217;t believe my eyes. &#8220;Who in their right mind samples a low-end APU processor to highlight their new lineup?&#8221; I thought. After years of bleeding edge Intel and AMD samples being sent in an effort to showcase the best of the unaffordable best, getting an A8-7600 just seemed odd.

I quietly progressed through the benchmarks and now, after more than 20 hours of testing, I get it. This particular APU isn&#8217;t about record setting overclocking, extreme gaming or pushing the boundaries of x86 performance. It&#8217;s a stark reminder of reality. While part of me misses the race towards the top, that&#8217;s not what folks are looking for these days. AMD&#8217;s goal this time around was to showcase how ridiculously capable their more affordable APUs are.

The A8-7600 is the real standout of AMD&#8217;s new lineup. It epitomizes everything we could want from Kaveri by incorporating the graphics and multimedia performance of a 95W A10-6800K into a part with a TDP of just 65W. More importantly, it can be easily configured down to 45W without taking a massive cut in the areas that count the most to its intended market: GPU accelerated decoding, transcoding and gaming. I&#8217;m not saying forget about the K-series parts but the A8 7600 makes one hell of a convincing case for itself in HTPC systems and simple set-top boxes. Intel doesn&#8217;t have a thing that can touch it within their current Haswell product stack.

By combining graphics horsepower that&#8217;s focused in key areas and &#8220;good enough&#8221; (for its segment at least) CPU performance into an efficient, affordable package, AMD has created the perfect poster child for their design mantra and a great companion for SFF systems. It may not have the overclocking chops of its more expensive siblings but the A8 7600 is infinitely more capable of displaying the best Kaveri has to offer."


http://www.hardwarecanucks.com/foru...1-amd-kaveri-a10-7850k-a8-7600-review-28.html

 

[Informant X]

I am so confused! Building a HTPC/lite gaming PC that will game at 720P in most likely.

Don't know if should go Kaveri, or the A10-6800K, or cheapest i3 with a R7 260X???

If I went Kaveri I would get 8 gigs PC 2133 ram, or if I went i3, i could get only 4 gigs of a cheaper ram, so the price difference would basically negate each other.

 

[Spawne32]

Wow, talk about a blow to the gut. Not only is it slow, its painfully slow compared to what they touted. Looks like im still on track to buy a 4670k as the prices come down.

 

[sniffin]

the only thing that looks bad is pricing, everything else has been improved, especially graphics.

Steamroller looks hopeless however. The entire package would look 2x more enticing in 35W mobile form, and yet they launch with this. I've ranted about this to many times so I won't go there, but this is clearly not a desktop chip. And yet thats what they give us day 1. What even goes through there heads, jesus

 

[monstercameron]

I am so confused! Building a HTPC/lite gaming PC that will game at 720P in most likely.

Don't know if should go Kaveri, or the A10-6800K, or cheapest i3 with a R7 260X???

If I went Kaveri I would get 8 gigs PC 2133 ram, or if I went i3, i could get only 4 gigs of a cheaper ram, so the price difference would basically negate each other.

with current pricing you are better of with cpu+dgpu for gaming.

 

[JDG1980]

From the Anandtech review conclusion:

Some questions regarding that:

1. Does it mean we cannot expect much gains from Excavator either?

2. Is anything known about what AMD's plans are beyond Excavator? Will there be a completely new CPU uarch line (non-Bulldozer based)?

3. If Excavator is the "end of the current CPU uarch line" for AMD, then what is the corresponding "end of current CPU uarch line" for Intel? Will they e.g. make a complete uarch redesign after Broadwell or Skylake, and start a new "line"?

If the leaked die shot is accurate, Excavator will have a beefed-up FPU, which should help some.

If AMD moves away from the "construction equipment" cores, I suspect the backup plan would be to beef up the "cat" cores (Bobcat->Jaguar, etc.) to the point that they are competitive at higher clock/power levels. This would basically be a replay of what Intel did back in 2006: Netburst wasn't working, it was too inefficient and IPC was terrible, so they took their mobile Pentium-M and beefed it up to become the Core 2 Duo. I can easily see AMD trying something similar. It would also save them the trouble of maintaining 2 different lines of x86 chips that have little in common with each other. (They'd probably still need to keep a slim-line cat core for tablets, alongside the beefed-up version, but at least this would minimize duplication of effort.)

I don't see Intel dropping the P6 architecture because there is no reason to. A new architecture would be a big financial risk, might not work out, and would be unnecessary since Intel currently has a big lead. They won't do a new architecture until Moore's Law peters out and AMD finally catches up to them.

 

[sefsefsefsef]

Steamroller looks hopeless however. The entire package would look 2x more enticing in 35W mobile form, and yet they launch with this. I've ranted about this to many times so I won't go there, but this is clearly not a desktop chip. And yet thats what they give us day 1. What even goes through there heads, jesus

The reason to go desktop first is yields. Reject mobile chips get repurposed as high TDP desktop parts.

 

[mrmt]

Send out 45W - 65W SKU models. Review sites compare them to a10-6800k.

They are extending the TDP fraud from the FX line to the rest of the consumer line up. I can't expect for Beema and Mulins. Kaveri 45W is consuming more power than Intel 54W SKUs. I wonder if they can say to server customers with a straight face that 3.1Ghz SKUs are 45W. Let's watch the bins of Opteron SKUs.

Not having HSA available on day 1 was a major fail.

That's not unheard of when talking about AMD. They didn't have working OpenCL runtime on their catalyst package when they launched their DX11 card, and their OpenCL SDK is still in bad shape today. Developer relations and software support aren't AMD strong points.