AMD sent Techreport an Intel system to test their GPU's

[Ferzerp]

If I remember correctly, the review guide for some radeon circulated back when the 7xxx series was new that used intel as well. I can't find a link right now though.


edit: here it is, [H] posted the review guide in their forum. http://hardforum.com/showthread.php?t=1677932 It lists the Intel system as the primary platform for their data.

All data, unless otherwise noted has been collected on the primary benchmark system below.

Primary Benchmark System
CPU
Intel Corei7 3960X (3.3GHz)
Motherboard
MSI X79A-GD65 8D
Memory
Corsair Vengeance (4x4GB) DDR3-1600 (9-9-9-24)
HDD
1TB Seagate Barracuda 7200.11 (ST3100033AS)
Display Driver Version
AMD Radeon™ HD 7800 Press Driver (8.95.5 RC1),
AMD Catalyst 12.1 Performance Driver,
NVIDIA ForceWare 295.73 WHQL

"primary benchmark system" is the intel one.

 

[ShintaiDK]

Abit wierd they claim the 8790M to be a successor of the 7690M. One would think it would atleast be a 7770M. But that might not have looked so good in the benchmarks.

 

[mrmt]

. I tend to think it was just a matter of convenience of putting the platform together rather than that an AMD CPU would have been limiting.

They didn't use a notebook platform chip, not even an low-wattage chip. They went straight for a 3770k, so we are already speaking of a non-common usage scenario where the FX would nicely fit.

And more important, this is a marketing initiative, not a validation program. They could use this time exactly to showcase how good, or at least how "good enough" FX is. Everyone that bothers to read reviews know that Intel is better for games, and the people that don't bother buy Intel anyway. They could use this small window to prop up FX, or at least not give more brand awareness to Intel.

And from a business perspective, no matter how much money they plan to make with these mid-tier GPU, they have a lot more at stake in their CPU business. Even if they can't make money with their CPU they must sell it or then incur in another take-or-pay charge with GLF, and you can bet that one more charge and they are done for. So whatever they do, they just can't hurt they CPU brand more than they already did.

 

[meloz]

Why would AMD need to send any cpu to anyone? Why cant they just send a card?

THIS.

Why the need to ship systems and CPUs? Something funny going on in the BIOS / UEFI of different OEMs?

 

[Rvenger]

THIS.

Why the need to ship systems and CPUs? Something funny going on in the BIOS / UEFI of different OEMs?

Thinking they knew they didn't have any consistent microstutter issues on that particular hardware platform.

 

[Idontcare]

Why would AMD need to send any cpu to anyone? Why cant they just send a card?

THIS.

Why the need to ship systems and CPUs? Something funny going on in the BIOS / UEFI of different OEMs?

If you want all the reviews of your GPU product to come out remotely comparable and without any eyebrow-raising outliers, you send out an entire platform to ensure fewer things go wrong at the test site.

 

[Ajay]

They didn't use a notebook platform chip, not even an low-wattage chip. They went straight for a 3770k, so we are already speaking of a non-common usage scenario where the FX would nicely fit.

And more important, this is a marketing initiative, not a validation program. They could use this time exactly to showcase how good, or at least how "good enough" FX is. Everyone that bothers to read reviews know that Intel is better for games, and the people that don't bother buy Intel anyway. They could use this small window to prop up FX, or at least not give more brand awareness to Intel.

And from a business perspective, no matter how much money they plan to make with these mid-tier GPU, they have a lot more at stake in their CPU business. Even if they can't make money with their CPU they must sell it or then incur in another take-or-pay charge with GLF, and you can bet that one more charge and they are done for. So whatever they do, they just can't hurt they CPU brand more than they already did.

At the rate they are going, AMD is going to make a good case study in Business School as an example of poor leadership and it's cascading long term effects. If I were an AMD investor, I'd want the BoD dragged behind a horse team over very rocky roads for about an hour D:

 

[WaitingForNehalem]

AMD really does have terrible management and marketing teams, their hardware engineers OTOH are superb. What a terrible decision that tarnishes their FX platforms.

 

[Red Hawk]

Abit wierd they claim the 8790M to be a successor of the 7690M. One would think it would atleast be a 7770M. But that might not have looked so good in the benchmarks.

It is weird, and counter-intuitive. I find it odd that they are releasing low end GCN chips at all, after refraining from doing so with the 7000 series. AMD may have lost market share over the past year but relying on VLIW5 parts for the low end isn't the reason, because Nvidia did the same thing -- the smallest Kepler part they released was GK107, a rough match for Cape Verde (Radeon HD 7700 and 7800 series). Nvidia did have an advantage though because they shrank Fermi down to 28nm for the 630m and other parts, while AMD stuck with 40nm for low end 7000 series. The logical response, then, is to provide a shrink of VLIW5 parts down to 28nm, or maybe even isolate the VLIW4 GPU from Trinity into a discrete chip. Not develop a whole new GCN chip. On the other hand, they may be doing something similar to that last one -- it might be the case that the 8790m is the same graphics chip that will be featured in Kaveri.

And yeah, the comparison to a 7770m wouldn't be favorable. The 7770m is a desktop 7750 clocked at 675 MHz. That's 512 GCN shaders, 128 more than the 8790m. 8790m does have an impressive clock speed of 900 MHz, but that likely wouldn't be enough to make up the difference.

 

[Torn Mind]

They didn't use a notebook platform chip, not even an low-wattage chip. They went straight for a 3770k, so we are already speaking of a non-common usage scenario where the FX would nicely fit.

And more important, this is a marketing initiative, not a validation program. They could use this time exactly to showcase how good, or at least how "good enough" FX is. Everyone that bothers to read reviews know that Intel is better for games, and the people that don't bother buy Intel anyway. They could use this small window to prop up FX, or at least not give more brand awareness to Intel.

And from a business perspective, no matter how much money they plan to make with these mid-tier GPU, they have a lot more at stake in their CPU business. Even if they can't make money with their CPU they must sell it or then incur in another take-or-pay charge with GLF, and you can bet that one more charge and they are done for. So whatever they do, they just can't hurt they CPU brand more than they already did.

They aren't going to convince skeptics by just giving an FX testbed without a suitable comparison testbed with Intel parts. For the actual review, I have no idea how good the 3770K is worth as a buy and I would expect the same reaction for an FX chip. Their reputation for CPUs would remain unchanged while they do risk obscuring the performance gains of their mobile GPU if the FX chip is indeed a bottleneck; I think they would know in-house if it were or not.

 

[pablo87]

The platform and the use of the ATI brand is bizarre for sure but red herrings aside, this is a pretty good performing part for the die size - seems to sit between a 640M and a 650M which are 50% larger.

My WAG BOM cost is $40 for an AMD platform, $60 for Intel

 

[Red Hawk]

I doubt it really competes with the 640m, which uses GK107. Maybe it has better memory bandwidth, since the 640m uses GDDR3, but the 640m should have much more raw shading power.

 

[Haserath]

I doubt it really competes with the 640m, which uses GK107. Maybe it has better memory bandwidth, since the 640m uses GDDR3, but the 640m should have much more raw shading power.

www.geforce.com/hardware/notebook-gpus/geforce-gt-640m/specifications

I think the 8790m should beat it by quite a bit, and the lowest 8xxxm should be close. They have equivalent shaders.

 

[Red Hawk]

www.geforce.com/hardware/notebook-gpus/geforce-gt-640m/specifications

I think the 8790m should beat it by quite a bit, and the lowest 8xxxm should be close. They have equivalent shaders.

GCN performance per shader =/= Kepler performance per shader. Otherwise the Radeon HD 7950 would beat the Geforce GTX 680.

 

[Lepton87]

GCN performance per shader =/= Kepler performance per shader. Otherwise the Radeon HD 7950 would beat the Geforce GTX 680.

Normalize the clocks and they are pretty close, as long as tahiti still have its bandwidth advantage. I bet that after 1.2GHz core clock 7950 would beat GTX680.

HIS 7950 is at 1210/1500

 

[IntelUser2000]

About the topic:

GPU marketing does their stuff, CPU marketing does their own stuff. It could be done better I guess, but its not a bad way to do things either. You aren't going to pair FX chips or even Trinity chips to these GPUs, you pair them with a mid-range or better Intel CPU.

And there's no real excuse by using the "mobile vs desktop chip" argument. The high end quad core Ivy Bridge mobile chip from Intel has performance on par or sometimes exceeding 2600K. 10 or so % gain you get from using 3770K does little in practice.

It is weird, and counter-intuitive. I find it odd that they are releasing low end GCN chips at all, after refraining from doing so with the 7000 series. AMD may have lost market share over the past year but relying on VLIW5 parts for the low end isn't the reason, because Nvidia did the same thing -- the smallest Kepler part they released was GK107, a rough match for Cape Verde (Radeon HD 7700 and 7800 series).

This isn't due to anything of malicious intent or cheating you out of your money. It's because they HAVE to. You can't deliver every product at every segments in the same time, different designs take time. Oh, and branding older generation as newer may feel like deception but it probably doesn't hurt money-wise, in fact it probably helps it to sell.

 

[Idontcare]

At the rate they are going, AMD is going to make a good case study in Business School as an example of poor leadership and it's cascading long term effects. If I were an AMD investor, I'd want the BoD dragged behind a horse team over very rocky roads for about an hour D:

I agree with the first part, not sure what more the BoD could do though versus what they already have done. They fired Dirk and spent 6 months looking for an acceptable replacement. It is too soon to say Rory isn't making the best of a very bad situation created by Dirk to begin with.

The BoD has to give management time to change course, if they whipsaw the top too much then they'll just have an HP on their hands.

The real criminal element here got away free though - Hector. He soaked up millions in bonuses while selling the AMD shareholder down the river with loaded asset-stealing bombs in the GloFo contract, let alone all the insider info stuff that was going on, and he's sitting rich somewhere laughing every day.

Everything Hector touched has turned out to be toxic afterwards. That includes Dirk IMO.

They aren't going to convince skeptics by just giving an FX testbed without a suitable comparison testbed with Intel parts. For the actual review, I have no idea how good the 3770K is worth as a buy and I would expect the same reaction for an FX chip. Their reputation for CPUs would remain unchanged while they do risk obscuring the performance gains of their mobile GPU if the FX chip is indeed a bottleneck; I think they would know in-house if it were or not.

I completely agree with the bold, chaining their GPU to an FX platform for the reviewers at this point would be the investing equivalent of throwing good money after bad.

The GPU team is literally fighting for their own jobs, no way they are going to tie their personal fates to the masts of S.S. Bulldozer. At least this way they stand a chance of saving their own necks in the next round of layoffs if they can keep GPU sales from further eroding.

In a sinking ship it is every man for themselves and the GPU guys are in no position to throw a lifeline to their CPU compatriots.

 

[meloz]

Thinking they knew they didn't have any consistent microstutter issues on that particular hardware platform.

Then they should address those microstutter issues rather than taking such desperado measures to hide them. Actually these kind of stunts draw more attention to any problem.

If you want all the reviews of your GPU product to come out remotely comparable and without any eyebrow-raising outliers, you send out an entire platform to ensure fewer things go wrong at the test site.

This is assuming your rivals and your own GPU will all get tested on the very same system. Did AMD just 'donate' this system to Techreport for future reviews of all AMD and nvidia GPUs? And what happens when nvidia provides one system, AMD another.

Reputable websites already use a standard hardware system for their reviews, specify which versions of BIOS, OS, drivers etc are used. This is particularly paranoid from AMD.

 

[TekDemon]

http://techreport.com/review/24086/a-first-look-at-amd-radeon-hd-8790m

Really AMD? Just outsource the rest of your marketing team, please.

It would be CPU limited by their own CPUs so what do you want them to do? Ship over a liquid nitrogen cooled AMD system for benching?

 

[frozentundra123456]

It would be CPU limited by their own CPUs so what do you want them to do? Ship over a liquid nitrogen cooled AMD system for benching?

Are you sure about this? I dont think a mid level mobile gpu would limit an FX or probably even an A10. If they were testing high end desktop cards that would be another story.

 

[tw33k]

They should never have stopped using the ATI brand. Moreons.

lol...I love the irony. I'm thinking about pots and kettles now

 

[dguy6789]

lol...I love the irony. I'm thinking about pots and kettles now

I can't believe there are people unable to distinguish between an unnoticed typo and a misspelling

 

[mrmt]

The real criminal element here got away free though - Hector. He soaked up millions in bonuses while selling the AMD shareholder down the river with loaded asset-stealing bombs in the GloFo contract, let alone all the insider info stuff that was going on, and he's sitting rich somewhere laughing every day

While Ruiz had its share of shady stuff, he could not have done everything he did by himself, in some decisions the BoD should at least be informed, in others the board should give explicit approval to Ruiz, and so they did. In fact, after the spin off they put Ruiz at the helm of GLF, not something you can do when your reputation is bad.

But there is more.

In 2006 AMD BoD was comprised of 10 members, being Hector as CEO and Chairman and Robert Rivet, CFO. The others members were:

- W. Michael Barnes
- Robert B. Palmer
- John E. Caldwell
- Bruce L. Claflin
- H. Paulett Eberhart
- James Fleck
- Morton Topfer
- Leonard Silverman

In 2012 what we have 12 members as following:

- Dr. W. Michael Barnes
- John E. Caldwell
- Henry WK Chow
- Bruce L. Claflin
- Craig A. Conway
- Nicholas M. Donofrio
- H. Paulett Eberhart
- Jack Harding
- Robert B. Palmer
- Ahmed Yahia Al Idrissi
- Waleed Al Muhairi
- Rory Read

So, almost half of the BoD, including the Chairman, is comprised of the same persons that reviewed and endorsed the streak of bad decisions that Ruiz and Dirk took, more if you discount that Ahmed and Waleed are Mubadala and ATIC men and the CEO has a mandatory place. Hector is surely criminal, but AMD corporate structure is being completely lenient with his accomplices.

 

[Haserath]

I can't believe there are people unable to distinguish between an unnoticed typo and a misspelling

What moarons.:awe:

/joking

 

[zebrax2]

My view on this is people generally don't care on what system you review the product in( much less where it came from) unless it bottlenecks the said product or have some other issues with the system.