GPU DRAM

Bit-Flip Vulnerability of Shared KV-Cache Blocks in LLM Serving Systems

Announce Type: replace Abstract: Rowhammer on GPU DRAM has enabled adversarial bit flips in model weights; shared KV-cache blocks in LLM serving systems present an analogous but previously unexamined target. In vLLM's Prefix Caching, these blocks exist as a single physical copy without integrity protection. Using software fault injection under ideal bit targeting, we characterize worst-case severity and identify three properties: (1) Silent divergence - 13 of 16 BF16 bit positions produce...

PlayStation Architecture

Supporting imagery A quick introduction Sony knew that 3D hardware could get very messy to develop for. Thus, their debuting console will keep its design simple and practical… Although this may come at a cost!

From Roofline to Ruggedness: Decomposing and Smoothing the GEMM Performance Landscape

arXiv:2605.29752v1 Announce Type: cross Abstract: Adjacent GEMM problems that differ by a single 128-element step in N can show 30% different throughput on the same GPU. This pervasive performance ruggedness - invisible to roofline analysis and peak-FLOPs intuition, yet dominant for every non-peak workload - is the subject of this paper. We propose performance ruggedness analysis as an analytical framework complementary to roofline: rather than summarizing GPU performance with a scalar...

Qualcomm picks bad time to pitch a $300 laptop platform

Upstart chipmakers keep challenging Nvidia. This time it's Microsoft-backed D-Matrix

In the increasingly competitive AI chip market, there's another startup in production that claims an advantage over Nvidia, the world's most valuable company. D-Matrix, located three miles away from Nvidia's Silicon Valley headquarters, says its chips can run inference workloads 10 times faster and using five times less energy than a standalone graphics processing unit from the market leader — as long as the workloads are small. The new inference chip, called Corsair, takes a novel approach...