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

[cbn]

Compare a low end discrete, 7750 or 7770:
http://www.tomshardware.com/reviews/radeon-hd-7770-7750-benchmark,3135-8.html

Power use? >200W:
http://www.tomshardware.com/reviews/radeon-hd-7770-7750-benchmark,3135-14.html

That test was run with a Core i7 Sandy Bridge E overclocked to 4.2 Ghz-> http://www.tomshardware.com/reviews/radeon-hd-7770-7750-benchmark,3135-4.html

So most people (with lesser CPUs like Core i3) will have much lower total power usage with HD7750 during gaming.

APU are in another league that low end CPU/low end discrete cannot touch, much more power efficient and cheaper.

Maybe not:

That Core i3 is leaving a lot of room for a discrete Video card to be added in.

 

[Torn Mind]

The TDP the the 7770 is 80 watts with very low idles. The greater memory bandwidth and more stream processors is worth it. The 7750 outclasses its Kaveri counterpart with its GDDR5 memory by a large margin while only having a TDP of 55W.

 

[blackened23]

Compare a low end discrete, 7750 or 7770:
http://www.tomshardware.com/reviews/radeon-hd-7770-7750-benchmark,3135-8.html

Power use? >200W:
http://www.tomshardware.com/reviews/radeon-hd-7770-7750-benchmark,3135-14.html

APU are in another league that low end CPU/low end discrete cannot touch, much more power efficient and cheaper.

They're really not in another league. The i3 uses a lot less power and some low end dGPUs use very low power. You suggest the 7750 or 7770. Why. Also the cards themselves don't use 200W, you forgot to mention that is total system draw which is entirely dependent on the power draw of the other components. In fact, for some reason, the power draw tests you linked are using a system based on a 3960X. Really? You're going to link total system draw with a SB-E chip? That isn't going to be in the same ballpark as a Haswell i3 or celeron.

As far as dGPU, I was thinking more along the lines of a GT640 or a GTX650, both of which would use less power than the 7750 and 7770, respectively. The total system draw with a Haswell i3 should not exceed 100W, which certainly is in the same ballpark as an APU. Similar in terms of cost as well.

 

[Silverforce11]

Maybe not:

That Core i3 is leaving a lot of room for a discrete Video card to be added in.

~117W for the new A10 in Civ5 with heaps of units on screen per Hardwarecannucks.

As for the 7750/70 TDP, i think we all know by now not to go with TDP.. and see it in actual reviews.

I would be surprised if its system load is anywhere near that and not beyond 150W total, from the wall.

 

[Obsoleet]

I wasn't impressed when reading Anand's review of iGPU gaming, ridiculous settings gives a slideshow, who buys a cheap APU and game at those settings?

Reading here: http://www.hardwarecanucks.com/foru...1-amd-kaveri-a10-7850k-a8-7600-review-24.html

Now I am officially impressed. 1080p gaming at normal/medium settings, fluid gameplay in all newer titles. Whats even more impressive is BF4 at 1080p with low/medium is running so fast.

http://www.hardwarecanucks.com/foru...1-amd-kaveri-a10-7850k-a8-7600-review-26.html

And they ran their Ram at 1600mhz!

Very happy with this performance improvement. At these low range of FPS, a bump to around 37 fps makes a massive difference to playability.

Thanks for this. I long since removed AT from my list of sites I check for reviews, since I've noticed so much stupidity (or, calling it what it really is- obvious NV/Intel bias) like this. AT can't be trusted for reviews, and not sure they ever were trustworthy. Checked out AT's review only because they had the A10, but was again disappointed- years down the line.

Added HardwareCanucks to my Firefox RSS feeds. Thanks again.

 

[Homeles]

Thanks for this. I long since removed AT from my list of sites I check for reviews, since I've noticed so much stupidity (or, calling it what it really is- obvious NV/Intel bias) like this. AT can't be trusted for reviews, and not sure they ever were trustworthy. Checked out AT's review only because they had the A10, but was again disappointed- years down the line.

Added HardwareCanucks to my Firefox RSS feeds. Thanks again.

At this point, I'm thoroughly convinced that AnandTech's perceived bias is all in people's heads. People will believe whatever they want in order to feel persecuted.

Not that anyone's perfect, but I'm just not seeing it.

 

[AnandThenMan]

Not that anyone's perfect, but I'm just not seeing it.

It's really, really obvious.

 

[Silverforce11]

I dont think they are biased, just pretty slack for not taking time to test at playable settings. Nobody game on ultra settings on APU at 1080p for starters so what is the point of testing it.. it was a complete waste of my time to read that review I had to seek out other sites to see a more realistic test scenario.

 

[AnandThenMan]

If it's not bias, then it's either a complete lack of effort, or perhaps incompetence. And I question the last two as being possible given the excellent write-ups on the tech itself.

 

[Shivansps]

Anand did the exact same thing with the Iris Pro review tested low quality at low res, them scaling up quality and resolution at the same time.

I dont remember anyone complaining about "bias" back them.

 

[Homeles]

I dont think they are biased, just pretty slack for not taking time to test at playable settings. Nobody game on ultra settings on APU at 1080p for starters so what is the point of testing it.. it was a complete waste of my time to read that review I had to seek out other sites to see a more realistic test scenario.

IMO, the review industry in general could use work. AnandTech less so than others.

It's really, really obvious.

So are the voices in a Schizophrenic's head. Once again, it's all in your head. You saying it's obvious doesn't mean anything.

What can be asserted without evidence can be dismissed without evidence. Since it's "really, really obvious," I'm sure you'll have no problem digging that evidence up.

I'm not intending to be rude here; it's just that I've seen this stated over and over, by people with very obvious pro-AMD bias, and have yet to see damning evidence.

It's like you didn't even read my post you quoted.

 

[meloz]

The one thing I take for this is that desktop CPUs are now "completed". When looking at what's on the horizon they will simply not provide much more performance going forward.

Both Intel and AMD are bringing us such minuscule performance increases from one generation to the next that it is pointless to upgrade your desktop PC. We're heading towards 10 year upgrade cycles.

It only appears that way because most actions on modern PC (with SSD) happen within less than a second of passing command, so performance improvement is not perceptible to users in day-to-day use.

That does not mean Intel is not making big leaps. AVX2 is a very big deal, it increases performance 30% or more in many software when the said software is compiled with AVX2 support. The only reason no one is talking about it right now is because most users are Windows obsessed, and in the world of Windows it takes time for commerical software vendors to compile and sell users a new copy. And when do you think Microsoft will ship a copy of Windows which is compiled with AVX2 enabled, a decade from now?

If you want you can run an entire OS (Gentoo Linux, it is all compiled locally) and get full benefit of AVX2 today. For other linux distros, you could just build the programs you hammer heavily locally with gcc 4.8.x, and again, get great benefit from AVX2. AVX2 will become a big talking point when we get Haswell-EP and Haswell-EX, every VPS user will want to build their nginx, postregsql and php with AVX2 to take full advantage of this new capability.

 

[DominionSeraph]

At this point, I'm thoroughly convinced that AnandTech's perceived bias is all in people's heads. People will believe whatever they want in order to feel persecuted.

Not that anyone's perfect, but I'm just not seeing it.

This. You're impressed with a review site because they tested it on low settings against nothing but Intel integrated graphics? Let's face it, all you want to point at is playable numbers with AMD on top, with no concern for where it actually places in the performance hierarchy.

The A10-7850K will be destroyed by a 760K + 7790, which can be had for the same price. Hey, guess what those are? An AMD processor and an AMD GPU that go in an AMD motherboard! But I suppose I'm an Intel/Nvidia shill for pointing that out, because it's not what you want to hear.

 

[zir_blazer]

The A8-7600 isn't that bad for its MSRP price, actually, it got quite a bit of praise. The problem is the high end models, as Kaveri seems to scale really bad. The A10-7700K got 400 MHz faster base clock, 200 MHz higher Turbo, the unlocked Multiplier, yet it cost nearly 30% more. The A10-7850K got actual extra working silicon with its bigger GPU and the price difference with the A10-7700K isn't that much, but over the A8-7600 the overall extra performance is meh.

Overally, the A8-7600 seems quite good. Yes, CPU performance is better on Intel, but as an overall platform, it is competitive if you want a better integrated GPU. Don't think on the fact that you can upgrade the Intel platform with a discrete Video Card and get better gaming performance, because for people where Kaveri will be a good choice (People with limited budget that want a gaming machine and are unlikely to upgrade due to lack of funds), chances are that it won't get any meaningful upgrade.


BTW, its me the only one that notice that no one seems to really have an idea about what is going on with the new Chipsets? AnandTech mentioned that A75 doesn't work with FM2+ and that there is a new A78 Chipset to replace it, yet I see Motherboards like this. Guru3D also speaked about an A87X Chipset.
Also, AnandTech said that A88X brings PCIe 3.0 support, which is not true as far that I know. PCIe lanes from Chipset are 2.0, its the Kaveri PCIe controller the one that is 3.0 capable, and even Motherboards with A55 can support its PCIe 3.0.

 

[DominionSeraph]

Anand did the exact same thing with the Iris Pro review tested low quality at low res, them scaling up quality and resolution at the same time.

I dont remember anyone complaining about "bias" back them.

And note how they don't complain about the unplayable FPS with Intel graphics. No calls to scale it down to 640x480 so Intel can keep up.

 

[yottabit]

Anand did the exact same thing with the Iris Pro review tested low quality at low res, them scaling up quality and resolution at the same time.

I dont remember anyone complaining about "bias" back them.

Except in that review the framerates all stay at relatively reasonable rates. Even on the harder tests there are results in the 15-25 FPS range. It's completely ridiculous to draw conclusions, especially in terms of % improvement, on framerates like 2-5 FPS like they show in the Sleeping Dogs max quality test.

I'm very sad about this particular review from Anandtech... there seems to be a large discrepancy in quality sometimes between reviews written by Anand and other authors. I definitely don't think Anand would have published an article like this.

I can't believe in the F1 benchmark they leave out Iris Pro (with no good explanation other than "various circumstances") and then say none of the Intel GPUs could keep up...

From the AMD side they are missing overclocked performance, memory scaling, etc... and there's just lots of weird stuff in this article and not enough real explanation.

 

[DaZeeMan]

Dual Graphics really work?

Colour me surprised!

Even on richland?! Looks like I've missed something.

The Hardware France guys talked AMD into letting them use an as yet unreleased driver to showcase the dual graphics, without frame pacing. Their review noted 'flashes' and such due to the lack of frame pacing and such. We may see dual graphics with the new driver release this month, but who knows (I'm guessing that it'll be in the next release, not this month's release).

So, for now, AFAIK yeah no dual graphics for us regular peeps. The main thing is that with some titles, dual graphics can work, and add a fair amount of performance, IF they can smooth out the frame pacing and such to get rid of those 'flashy' frames and such.

Personally, I'm quite excited that they've been able to get it to work so far, assuming they can iron out the issues. Pairing a R7 GDDR5 discrete card with a Kaveri almost doubles the performance in the Hardware France BF4 bench! The discrete R7 GDDR5 card on it's own only manages a 45% increase over the iGPU on it's own, with an additional 35% increase over the dGPU score when the iGPU is added.

As noted, it doesn't help out in some titles currently (heavy multithreading seems to be needed), but it's nice to know the performance boost would be available for the optimized titles if you want it...



Having read most of the reviews today, I'm seeing performance increases on the gaming end with Kaveri in a number of benchmarks. An increase is an increase, and is especially important if that increase gets you closer to/past the 30 FPS threshold.

Single thread performance suffers due to the reduced clock speed, but y'all should have already known that going in. But when you consider AMD's heavily investing in Mantle and TrueAudio, well there are APU performance gains to be had there, by offloading/shifting some of the workload away from the CPU. Then there was that offhand remark I saw in one review (Anand?) about HSA not being well supported as of yet, due to it being so new...

I fully expect to see some more impressive performance numbers once the other features are finally fully implemented. So, if you are on the fence, I'd recommend holding off a month or two, to see if the suggested performance gains from these come even close to AMD's vision. If AMD is even close to right, well you'll be buying into some cutting edge technologies - Kaveri isn't just about the benchmarks.

I'd expect at least a few of the 'usual players' to do a performance update once Mantle, True Audio, Dual Graphics, etc. are finally supported. So it can't hurt to wait.

Besides, the Kaveri APUs will probably be a bit cheaper by then...

BTW, Thief will be the first title that fully incorporates Mantle and TrueAudio, so that'll be a good place to look for how these technologies perform. February 25th I believe is the launch date?

As for those looking for serious gaming performance, well you should be going the dGPU route anyways, and Intel is your chip of choice for high performance gaming (if you can afford it). Which Intel has been for quite some time now... nothing to see here, move along!

One of Kaveri's sweet spots will be those ITX boards... one of the reviews today (Guru3D) was on an ITX board... This will be especially true if TrueAudio and Mantle are able to deliver some gains.
http://www.guru3d.com/articles_pages/amd_a8_7600_apu_review,5.html

I'm sorry, but that ITX board is just too cute! You could fit what, 8 of those on an ATX board?!? It amuses me...

That being said, I'm quite pleased what I've seen so far, and have learned a bit more about the value of higher memory speeds with AMD APUs in the process!
:biggrin:

 

[DominionSeraph]

I dont think they are biased, just pretty slack for not taking time to test at playable settings. Nobody game on ultra settings on APU at 1080p for starters so what is the point of testing it.. it was a complete waste of my time to read that review I had to seek out other sites to see a more realistic test scenario.

Thankfully Anand doesn't test hardware only for you. He tests it for people who are actually relevant, like me. In order to put together builds for people I need to know where the performance lies in relation to alternatives, and nobody tests discrete GPUs on Low.

Honestly, your complaints don't hold up to the barest of scrutiny. If games can be scaled down to be playable for the notebook reviews, why do you need to see them scaled down to see if they're playable on a desktop APU?

Oh gee, I wonder if I there is any setting that would allow me to play Skyrim at better than 21.6 FPS on an A10-7870K? How ever could I figure this one out?

You're desperate to get one word that you can use in your internet fanboy arguments: "Playable." But the tailor-made settings you ask for are useless to everyone else because nothing else is benched at those settings.

 

[Shivansps]

People can buy a G3220 and a HD7790 for the same price of a 7850K, well, less, actually if consider mb and memory prices, yet no one has incluided that in comparison.

 

[Accord99]

APU are in another league that low end CPU/low end discrete cannot touch, much more power efficient and cheaper.

Cheaper yes but not more power efficient:

 

[AtenRa]

People can buy a G3220 and a HD7790 for the same price of a 7850K, well, less, actually if consider mb and memory prices, yet no one has incluided that in comparison.

Why people doesnt say the same thing when it comes to Intel Reviews ??

You can buy a G3220 + HD7790 for less than just the Core i5 4670K and have faster gaming performance vs HD4600. :whiste:

The Kaveri is not a CPU, it is an APU. You pay for having the fastest iGPU in a 95W TDP Single Chip that also has 4 powerful CPU Cores, you pay more for added features like TrueAudio etc.

 

[Silverforce11]

Thankfully Anand doesn't test hardware only for you. He tests it for people who are actually relevant, like me.

Yes if you buy an APU and game at ultra 1080p settings, its great Anand test it that way for you. Tell everyone who is "relevant" like you how awesome it is to game with a slideshow fps, because I have doubts they understand that.

Somehow, I don't think there's too many people like you who would do such a thing.

There's a target for HTPC gaming rigs, its often linked to the output TV, which these days are mostly 1080p. If you cannot figure this out and why its important to test at the best playable settings for this resolution, then you deserve to have Anand's very informative review.

I'm glad there are other reviewers who do understand the market better, an example of a gamer testing for other gamers: "This is IGP/APU performance hence you can't really flick open image quality settings. As such the DX11 tests have been set at 'medium quality' and admittedly that doesn't look bad." http://www.guru3d.com/articles_pages/amd_a8_7600_apu_review,10.html

 

[Shivansps]

Why people doesnt say the same thing when it comes to Intel Reviews ??

You can buy a G3220 + HD7790 for less than just the Core i5 4670K and have faster gaming performance vs HD4600. :whiste:

The Kaveri is not a CPU, it is an APU. You pay for having the fastest iGPU in a 95W TDP Single Chip that also has 4 powerful CPU Cores, you pay more for added features like TrueAudio etc.

Thats nonsence, the i5 4670K will slaughter the G3220 on CPU, im not so sure about G3220 vs 7850K, specially if it has a hard time beating the I3-4330.
If someone wants just cheap gaming the G3220+HD7790 must be considered, you just cant ignore it "because its an APU" thats pointless unless i really need it, HTPC, mini itx, etc.
Also you can place a real dgpu on the 4670K and no be CPU limited, unlike what happens on 7850K and G3220, and the HD7790 has everything but HSA.

 

[Silverforce11]

If someone wants just cheap gaming the G3220+HD7790 must be considered, you just cant ignore it "because its an APU" thats pointless unless i really need it, HTPC, mini itx, etc.

As its been said, let's not pretend these APU are something they are not. Clearly they are not a good option for people who want to use a discrete GPU, period. There are better options.

But as a package and when size, thermals, power is a constraint, they are unmatched in their own category.

 

[Shivansps]

performance/price is what matters in desktop, all other is secondary, its pointless to buy a APU just because its an APU, leave that for HTPC and mobiles where everything is important.